diff --git a/bench/README.md b/bench/README.md index 75c3c2ac87c..e5cb2e84ea7 100644 --- a/bench/README.md +++ b/bench/README.md @@ -209,7 +209,7 @@ also run during timing. Every timed round reads and checks all payload bytes, returns lease credit, and fills each stream again. Throughput counts delivered payload once. Nightly's supplemental Rust benchmark leg records the series and keeps the memory output with its raw results. The broader MoQ session transport -and version sweep remains in [its quest](/quest/m4/bench-transport-sweep.md). +and version sweep remains in [its quest](/quest/m5/bench-transport-sweep.md). Compare one multi-threaded Tokio runtime with the same number of independent Tokio/epoll and io\_uring workers: diff --git a/doc/concept/socket.md b/doc/concept/socket.md index 8bc9f893936..53814468e83 100644 --- a/doc/concept/socket.md +++ b/doc/concept/socket.md @@ -148,7 +148,7 @@ every hop, so a slow reader slows the writer, as TCP backpressure does, and nothing is lost. A writer that doesn't wait sheds the front of its half, and a reader that falls behind it fails with an eviction error instead of skipping data. Pacing reaches across native QUIC and qmux hops; see -[unbounded groups](https://github.com/moq-dev/moq/tree/main/quest/m2/unbounded-groups) +[unbounded groups](https://github.com/moq-dev/moq/tree/main/quest/m3/unbounded-groups) for what remains, such as browsers, which read without pacing. Use `tcp` for a byte stream or a log of records that must all arrive in order, diff --git a/doc/lib/dart/index.md b/doc/lib/dart/index.md index 8779f6929bf..a5bc7f01950 100644 --- a/doc/lib/dart/index.md +++ b/doc/lib/dart/index.md @@ -11,7 +11,8 @@ The [`moq`](https://pub.dev/packages/moq) package on pub.dev wraps the generated [`moq_ffi`](https://pub.dev/packages/moq_ffi) bindings in Dart futures and streams. A Native Assets hook supplies the Rust core for Android (API 24+), iOS (16+), Linux, macOS, and Windows. Flutter web is not supported, -since it can't load a native library. +since it can't load a native library. The iOS library ships but has not yet +been run in an iOS app, so treat iOS as untested. Import `package:moq/media.dart` with a prefix for catalogs, encoded-media imports, and container consumers. `TrackProducer.audio` / `.video` take a broadcast, an init record, diff --git a/infra/README.md b/infra/README.md index a30686c1bf7..dd449a58d71 100644 --- a/infra/README.md +++ b/infra/README.md @@ -385,7 +385,7 @@ paths, with their old signatures, indefinitely. The next Cache push refills it. ## Bootstrapping the video corpus The nightly video-quality job, which [the harness -quest](../quest/m3/video-quality/harness.md) adds, reads its clips from two +quest](../quest/m9/video-quality/harness.md) adds, reads its clips from two buckets. `corpus-moq-dev` holds open clips and is public at `corpus.moq.dev`. `corpus-private-moq-dev` holds clips that cannot be redistributed. It has no diff --git a/quest/README.md b/quest/README.md index 70c5e250044..409074a8f0e 100644 --- a/quest/README.md +++ b/quest/README.md @@ -19,6 +19,24 @@ release execution, and bug fixes in shipped code; m4 holds tests, tooling, benchmarks, optimizations, and features nobody waits on yet. m5 holds work gated on the outside world, or speculative work with no named consumer. +[m9](/quest/m9/README.md) is the lab (decided 2026-11-12, maintainer, +/quest-audit; renamed from `lab` 2026-11-13): experiments and measurements +that touch no published API, wire format, or CLI surface. They start any +time, merge whenever, never gate a cut, and stay in place when the +milestones shift. `quest` orders milestones by name, so `lab` ranked ahead +of m0; as m9 it ranks last. + +Until the cut, [m4](/quest/m4/README.md) and [m5](/quest/m5/README.md) +require the [release cut](/quest/m3/release/cut.md) (decided 2026-11-13, +maintainer), so spare capacity goes to m0 to m3 and then to m9's lab work +instead of to later features. m3 stays open, since it holds the cut and its +reviews. The cut's PR deletes the cut quest, which lifts the gate. + +No quest requires one in a later milestone (decided 2026-11-13, maintainer): +a blocker ranks by its own milestone, so the earlier quest would stall behind +the later milestone's work. Additive work that builds on m2 joins m3, which +ranks right after m2, rather than m1. + A change is breaking when a published consumer or a deployed peer must adapt to it: the public Rust, JS, or bindings API, a frozen wire version, the hang catalog JSON, a stored or exported container format, or CLI flags, config diff --git a/quest/m0/cut-placement.md b/quest/m0/cut-placement.md new file mode 100644 index 00000000000..b27e822ae48 --- /dev/null +++ b/quest/m0/cut-placement.md @@ -0,0 +1,33 @@ +# [M] Place each cycle's cuts so no milestone waits on a later one + +## Goal + +The [release cut](/quest/m3/release/cut.md)'s steps re-create each later cut +in a place where no quest requires one in a later milestone, and re-create +the gate that holds later work until that cut while the lab, +[m9](/quest/m9/README.md), stays open. The current cut follows them when it +re-creates its successor. + +## Plan + +Planned 2026-11-13 (maintainer, /quest-plan, on +[moq-dev/moq#8637](https://github.com/moq-dev/moq/pull/8637)): the same day +settled that no quest requires one in a later milestone, and that +[m4](/quest/m4/README.md) and [m5](/quest/m5/README.md) require this cut +until it lands. The cut's steps predate both: they re-create later cuts in +m0's release epic, and the wire cut that ends a cycle requires m2 and m3. + +The constraints a placement must meet, all enforced by `quest check` once +moq-dev/quest's rule ships: + +- A cut lives in or after the last milestone it waits on. +- A cut can't require the milestone that holds it, or any epic above it. +- A milestone gated on a cut can't hold that cut, or the gate is a cycle. + +Settle with the maintainer where each phase's cut lives and which milestones +its gate covers, then rewrite the cut's steps and the root +[questline](/quest/README.md)'s cycle description to match. + +## Related + +- [Release cut](/quest/m3/release/cut.md) - holds the steps this rewrites diff --git a/quest/m0/nightly/README.md b/quest/m0/nightly/README.md index d2852608302..95fe8762473 100644 --- a/quest/m0/nightly/README.md +++ b/quest/m0/nightly/README.md @@ -32,7 +32,7 @@ This README's own work, after the children: fixtures were repaired in [moq-dev/moq#7940](https://github.com/moq-dev/moq/pull/7940). The receive-windows benchmark's overcommitted cells, skipped since [moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838), moved to - [m1](/quest/m1/receive-windows-evicted.md) on 2026-11-04 (maintainer). + [m1](/quest/m3/receive-windows-evicted.md) on 2026-11-04 (maintainer). Decided 2026-11-05 (maintainer, /quest-plan): catch the next moq-net bench break in Nightly's measured coverage only. Do not add a test-mode bench run diff --git a/quest/m1/README.md b/quest/m1/README.md index a151bbf6821..bd8fa80fa4a 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -25,6 +25,31 @@ because a downstream iGPU transcode trial needs it now. Decided 2026-11-13 (maintainer, /quest-run): stream-credit stays in m1 and no longer gates m2, since takeover, its only wire change, moved to m3. +Decided 2026-11-12 (maintainer, /quest-audit before the cut): + +- ✅ Additive and docs-only work that sat in m2 moved here, so the cut stops + waiting on it: the time-addressing guide, the JS fills that carry a time, + and outbound federation (rejected: leaving each in m2 beside the breaking + work it follows). Each still lists the m2 quest it builds on under + `Required`. Superseded 2026-11-13, below. +- ✅ The mesh libraries and byte streams moved here from m3 (rejected: + staying first in m3). The 2026-11-03 reason to keep them out of m1 was that + the cut would wait on them; since 2026-11-05 the cut waits on m2 alone, and + a downstream application builds on them from `main` now. Superseded + 2026-11-13, below. +- ✅ Also moved here: the two reconnect fixes whose bugs came from PRs that + are on `main` and not on `release`, so the cut would publish them; the Go + module proxy, since the mirror repository nears 1 GB; and the ashpd bump. + +Decided 2026-11-13 (maintainer, /quest-plan): no quest requires one in a later +milestone, so the 14 m1 quests that waited on m2 work moved to +[m3](/quest/m3/README.md): buffered-ascending, export-order, +receive-windows-evicted, route-trace-validation, routing-guide, +subscribe-fills-js, time-addressing-guide, and the cluster-routing and mesh +epics whole. ✅ m3 (rejected: m2, where the cut would wait on them; pulling +their 12 breaking m2 blockers into m1). m3 is not gated, and they rank right +after m2, which is when they can start. + Decided 2026-11-13 (maintainer, /quest-plan on what testing to add): tests that guard what production traffic reaches join m1 ahead of the questline's m4 default, so changes are proven before they ship: the diff --git a/quest/m4/ashpd-maintained.md b/quest/m1/ashpd-maintained.md similarity index 89% rename from quest/m4/ashpd-maintained.md rename to quest/m1/ashpd-maintained.md index f3361805ba2..8e7d35d663f 100644 --- a/quest/m4/ashpd-maintained.md +++ b/quest/m1/ashpd-maintained.md @@ -16,6 +16,8 @@ adapt the portal capture client to any 0.14 API changes, drop the three exceptions, then run `just rs audit` and, on Linux, `cargo check --locked -p moq-video --no-default-features --features pipewire` so the portal capture client compiles against the new API. +Confirmed 2026-11-12 (/quest-audit): `cargo info ashpd` lists 0.14.0 on +crates.io; the lock still holds 0.13.13. ## Related diff --git a/quest/m3/go-module-proxy.md b/quest/m1/go-module-proxy.md similarity index 93% rename from quest/m3/go-module-proxy.md rename to quest/m1/go-module-proxy.md index f13f6fa6e96..ceec8c773db 100644 --- a/quest/m3/go-module-proxy.md +++ b/quest/m1/go-module-proxy.md @@ -9,8 +9,9 @@ published version, old and new, verifies against sum.golang.org. The `moq-dev/moq-go-ffi` mirror stops receiving releases and is archived. Today each release commits about 61 MiB of compressed staticlibs to that -mirror. It reached 610 MB after 35 releases in 12 weeks, so it passes -GitHub's 1 GB guidance within weeks. +mirror. It reached 610 MB after 35 releases in 12 weeks, and about 819 MiB +with 44 tags on 2026-11-12 (`gh api repos/moq-dev/moq-go-ffi --jq .size`, +in KiB), so it passes GitHub's 1 GB guidance in about three more releases. ## Plan diff --git a/quest/m1/obs-js.md b/quest/m1/obs-js.md index 6631edf6daf..c8826a26a16 100644 --- a/quest/m1/obs-js.md +++ b/quest/m1/obs-js.md @@ -87,5 +87,5 @@ Recommended: - [OBS simulcast check](/quest/m3/obs-multitrack-check.md) - keeps the dock's Go Live layers; this quest runs the obs-websocket recipe -- [MoQ in obs-studio](/quest/m3/obs-studio/README.md) - a native output keeps +- [MoQ in obs-studio](/quest/m5/obs-studio/README.md) - a native output keeps the same obs-websocket surface diff --git a/quest/m1/quic-sim.md b/quest/m1/quic-sim.md index 1bd0a25f9ba..e92ef5be5eb 100644 --- a/quest/m1/quic-sim.md +++ b/quest/m1/quic-sim.md @@ -24,9 +24,13 @@ What remains: (`crate::bind::udp`), so a sim test cannot reach them yet. Pick how a caller hands either one a `driver::endpoint::Socket`, such as a documented `sim`-only hook (`just check` bans `#[doc(hidden)]`). That API - choice belongs to the maintainer. -- Route the session-level spawns that `server.rs`, `connection.rs`, and - `origin.rs` still send through `crate::runtime::spawn` through + choice belongs to the maintainer (see Waiting). `Client::with_network` + (`rs/moq-tokio/src/sim.rs`) already gives a client an in-memory network + under the `sim` feature, and the relay's integration tests run hops and + reconnects over it (`rs/moq-relay/tests/integration/memory.rs`), though + not over QUIC. +- Route the session-level spawns that `server.rs` and `origin.rs` still + send through `crate::runtime::spawn` through `runtime::Spawner`, as `transport.rs` and the WebTransport session already do. - Then run `moq_relay` between a publisher and a subscriber, and between two @@ -42,7 +46,6 @@ queue is not evidence about congestion loss. Decided 2026-11-13 (maintainer, /quest-plan on what testing to add): ✅ m1 (rejected: staying in m4). It is the base for [fault schedules](/quest/m4/cluster-fault-schedules.md). -Check first: `moq_tokio::Client::with_network` and the relay's -`cluster::Config::network` already exist under `sim`, so say which of "What -remains" has landed, and whether that `Network` carries QUIC or the in-memory -transport, before picking the server's hook. +## Waiting + +- Maintainer: how moq-tokio's `Client` and `Server` take a simulated QUIC socket. Options: (a) extend the `sim`-feature `Client::with_network` path, already documented as tracking internals that can change in any release, to carry `quic::driver::sim::Network`, with a matching hook on `Server`; (b) a new `sim`-only `with_socket(driver::endpoint::Socket)` on each; (c) drop this quest, since the memory network already covers relay hops and reconnects. Recommended: (a) if a failure only real QUIC would show is named, otherwise (c). diff --git a/quest/m3/reconnect-probe-redirect.md b/quest/m1/reconnect-probe-redirect.md similarity index 100% rename from quest/m3/reconnect-probe-redirect.md rename to quest/m1/reconnect-probe-redirect.md diff --git a/quest/m3/stream/README.md b/quest/m1/stream/README.md similarity index 100% rename from quest/m3/stream/README.md rename to quest/m1/stream/README.md diff --git a/quest/m3/stream/rust.md b/quest/m1/stream/rust.md similarity index 96% rename from quest/m3/stream/rust.md rename to quest/m1/stream/rust.md index ae449581d63..e2f33c50d08 100644 --- a/quest/m3/stream/rust.md +++ b/quest/m1/stream/rust.md @@ -11,7 +11,7 @@ unmodified client and server talk through a relay. ## Plan -See the [epic](/quest/m3/stream/README.md) for the mapping: a socket endpoint +See the [epic](/quest/m1/stream/README.md) for the mapping: a socket endpoint per forwarded service, a session per connection with one group each way, a frame per TCP read, no fill handler. diff --git a/quest/m1/token-count-signing-input.md b/quest/m1/token-count-signing-input.md new file mode 100644 index 00000000000..b74881965d6 --- /dev/null +++ b/quest/m1/token-count-signing-input.md @@ -0,0 +1,28 @@ +# [S] A malleated signature does not double a token's session cap + +## Goal + +The auth server counts sessions per token by the token's signing input +(header and payload), not by the whole compact JWS. Today the per-token count +hashes the full string (`Sessions::connect` in `rs/moq-auth/src/serve.rs`), +and an ECDSA signature can be malleated into a second valid string for the +same claims, so an ES256 or ES384 holder gets twice `--limit-token`. + +## Plan + +Found by the 2026-11-07 invariant audit (auth-6). Split 2026-11-12 +(maintainer, /quest-audit before the cut) from +[the O(1) session index](/quest/m4/auth-session-index.md): ✅ this is a fix +to shipped behavior, since token caps are in `release`, so it joins m1 and +can be backported; the index stays an optimization in m4 (rejected: leaving +the fix inside the optimization). + +- Key the count on the signing input or the decoded claims; pick whichever + the verifier already has in hand, and say which in the PR. +- Reproduce first: a test that connects with a token and its malleated twin + and exceeds the cap on `main`. +- Check the other places a token is used as a map key (attachment leases, + shared request-token grants) for the same assumption, and fix each in this + PR. + +Public API: none. Wire: none. diff --git a/quest/m1/wire-fuzz-coverage.md b/quest/m1/wire-fuzz-coverage.md index aed2b62fb71..5660dd7cd2c 100644 --- a/quest/m1/wire-fuzz-coverage.md +++ b/quest/m1/wire-fuzz-coverage.md @@ -2,9 +2,9 @@ ## Goal -The hand-written unsafe that peer input reaches in `moq-c`, `moq-uring`, and `moq-sock` runs under Miri or a fuzz target of its own, asserting more than "it did not panic". Other peer-facing decoders, such as the media importers and RTMP's AMF0, are outside this quest. +The hand-written unsafe that peer input reaches in `moq-uring` and `moq-sock` runs under Miri or a fuzz target of its own, asserting more than "it did not panic". Other peer-facing decoders, such as the media importers and RTMP's AMF0, are outside this quest. -The moq-net half is done: the `ingest` and `request` targets cover the incremental group ingest and request-stream framing, and the seeds include prefixes at and past every cap. What remains: `rs/moq-c`, `rs/moq-uring`, and `rs/moq-sock` have no fuzz target and no Miri run. Their cmsg walks bound every length by the buffer, with unit tests, but nothing searches for the next mistake. +The moq-net half is done: the `ingest` and `request` targets cover the incremental group ingest and request-stream framing, and the seeds include prefixes at and past every cap. What remains: `rs/moq-uring` and `rs/moq-sock` have no fuzz target and no Miri run. `moq-c`'s unsafe is the C ABI, reached by its C caller rather than by a peer, so it is not part of this quest. Their cmsg walks bound every length by the buffer, with unit tests, but nothing searches for the next mistake. ## Plan @@ -18,10 +18,7 @@ The point is the invariant, not the coverage number: a target that only checks f Each crate follows moq-net's shape (`rs/moq-net/fuzz/README.md`): the target body in a `fuzz` module, documented as test support behind the `fuzz` feature, so the crate's own tests replay `fuzz/regressions/` on stable, a one-line shim per target, an `--exclude` in the `no_fuzz` list of `rs/justfile`, and a row in `nightly.yml`'s fuzz matrix. Target names must stay unique across crates. -Ordering: - -- `moq-sock`'s `CMsg::decode` still reads `size_of::()` past a short message (only a `debug_assert_eq!` checks it), which [Bound cmsg payload reads](https://github.com/moq-dev/moq/pull/7394) fixes. A moq-sock target landing first would fail nightly on that known overread, so land it after, or with, that fix. The moq-uring target has no such dependency. -- `just rs miri` and its nightly job already exist ([moq-dev/moq#7315](https://github.com/moq-dev/moq/pull/7315)); this quest adds its crates to them. +`moq-sock`'s known overread is fixed: [moq-dev/moq#7394](https://github.com/moq-dev/moq/pull/7394) (merged 2026-11-01) bounds `CMsg::decode` (`rs/moq-sock/src/udp/cmsg/mod.rs`), which now returns `None` for a short message, so neither target waits on anything. `just rs miri` and its nightly job already exist ([moq-dev/moq#7315](https://github.com/moq-dev/moq/pull/7315)); this quest adds its crates to them. Decided 2026-11-13 (maintainer, /quest-plan on what testing to add): ✅ m1 (rejected: staying in m4). This unsafe is reachable by any peer. @@ -31,4 +28,3 @@ Public API: none. Wire: none. ## Related - [CI cost and latency](/quest/m0/ci/README.md) - where the nightly budget lives -- [Bound cmsg payload reads](https://github.com/moq-dev/moq/pull/7394) - fixes the moq-sock overread a cmsg target would hit diff --git a/quest/m2/README.md b/quest/m2/README.md index 53addd4a318..86bb7a1ffb7 100644 --- a/quest/m2/README.md +++ b/quest/m2/README.md @@ -27,10 +27,30 @@ fixes, and moq-c parity) live under the recurring cut gate, decided the same day; the [release cut](/quest/m3/release/cut.md) holds its template. -Held by open claims: the [probe epic](/quest/m2/probe/README.md) -stays here while its open claims last, so the cut waits on it until the -[re-cut follow-up](/quest/m2/recut-held.md) moves it to m3. Its padding and JS -rung disable quests already moved to m3 on 2026-11-06. +Decided 2026-11-12 (maintainer, /quest-audit before the cut): + +- ✅ Left m2, so the cut stops waiting on them: the + [probe epic](/quest/m3/probe/README.md) to m3 and the routes + [browser proof](/quest/m5/routes-browser-proof.md) to m5 (each waits on a + browser release and breaks nothing); the docs-only and additive quests and + outbound federation to [m1](/quest/m1/README.md); and + [unbounded groups](/quest/m3/unbounded-groups/README.md) to m3, whose byte + offset goes on the wire version after lite-07 (rejected for each: staying + in m2). The padding and JS rung disable quests are parked in m5. +- ✅ Joined m2, each a small break the cut would otherwise freeze: the mesh + [grant shape](/quest/m2/mesh-grant-shape.md) and + [key budget](/quest/m2/mesh-device-bounds.md), since `moq_auth::mesh` + publishes for the first time at this cut; + [auth key validation](/quest/m2/auth-key-validation.md); the + [RTMPS connect](/quest/m2/rtmp-rtmps-tcurl.md) signature; + [usage's derived settings](/quest/m2/usage-derive-settings.md); + [first-publish versions](/quest/m2/first-publish-versions.md); and three + API fences: [mesh enrollment](/quest/m2/api-mesh-enrollment.md), + [the capture selections](/quest/m2/api-capture-fences.md), and + [the acknowledgment hook's trait](/quest/m2/api-ack-hook-trait.md). None + is on the routes chain, which stays the critical path. +- ✅ Under the path rule a route keeps one sticky down-only flag beside its + path; see the [routes epic](/quest/m2/cluster-routes/README.md). Carried from m1's Plan before the re-cut, where "m1" means the whole pre-cut milestone, now m1 and m2: @@ -85,7 +105,7 @@ warmup rule for the shipping dual decoder, the binding migration guide fixes, and the API-diff tooling. Public routes now wait on cluster routing. Added 2026-10-20 (maintainer): the probe target, native probing, and the -native rung disable moved up from the [m2 probe epic](/quest/m2/probe/README.md), +native rung disable moved up from the [m2 probe epic](/quest/m3/probe/README.md), so Rust, FFI, and OBS publishers recover a stopped rendition in this release; padding and the browser's rung disable stayed in m2, with only padding's lite-07 stream type here. Both moved to m3 on 2026-11-06 (maintainer, diff --git a/quest/m2/api-ack-hook-trait.md b/quest/m2/api-ack-hook-trait.md new file mode 100644 index 00000000000..7d0c05b63f9 --- /dev/null +++ b/quest/m2/api-ack-hook-trait.md @@ -0,0 +1,31 @@ +# [S] The send-stream trait has its acknowledgment hook + +## Goal + +`moq_net::transport::poll::SendStream` gains the shape the acknowledgment +hook needs: an associated `Offset` type, `offset()`, and `poll_acked` with a +default that reports the hook as unsupported. Every transport in the tree +compiles against it, and none has to implement the hook yet. The trait is +public and unsealed, so adding these after the cut would break every +implementer. + +## Plan + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ land the +trait shape in m2 and leave the drivers to the +[acknowledgment hook](/quest/m3/quic-ack-hook.md) (rejected: an extension +trait after the cut, which puts a new bound on moq-net's generic session +code and is itself a break; accepting another moq-net break later). + +- Take the shape from the hook quest's Plan; the tokio driver already has an + `Offset` (`rs/moq-tokio/src/quic/driver/send.rs`, from + [moq-dev/moq#6950](https://github.com/moq-dev/moq/pull/6950)), and + moq-quic's `poll_acked` exists behind `#[cfg(test)]`. +- "Unsupported" is a typed answer a caller can branch on at setup, not an + error it must match at runtime: a session decides once whether its + transport reports acknowledgments. +- JS has no counterpart: WebTransport exposes no acknowledgment, so `js/net` + is unchanged; say so in the PR. + +Public API: breaking for transports implementing `poll::SendStream` outside +the tree only if a method lacks a default; aim for none. Wire: none. diff --git a/quest/m2/api-capture-fences.md b/quest/m2/api-capture-fences.md new file mode 100644 index 00000000000..de23e1b0734 --- /dev/null +++ b/quest/m2/api-capture-fences.md @@ -0,0 +1,35 @@ +# [XS] Capture listings and selections can grow + +## Goal + +moq-video's capture API takes new fields without a break: + +- `Camera`, `Display`, `Window`, and `App`, the structs `cameras()`, + `displays()`, `windows()`, and `apps()` return, are `#[non_exhaustive]`. +- `Source::Display` takes a `#[non_exhaustive]` selection struct in place of + `Option`, built by a constructor from the display id, so an + exclusion list or a capture option joins it later. + +## Plan + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ fence these +now (rejected: a second moq-video break after the cut; adding exclusion as a +separate setting outside `Source`). Two later quests add fields here: +[capture listing](/quest/m3/capture-list.md) (placement, kind, identity) and +[capture exclusion](/quest/m4/capture-exclude.md), which says the exclusion +belongs to the display selection. Neither is pulled into this cut; this quest +only keeps them additive. + +Check the other `Source` variants while here: a tuple variant holding +`Option` has the same problem wherever a later quest plans a second +field. Fence those the same way and list each in the PR. Mirror the selection +struct in moq-ffi and the bindings that expose `Source`. + +Public API: breaking for moq-video callers that build or destructure these +(`Source::Display(None)` becomes the constructor), in Rust and the bindings. +Wire: none. + +## Related + +- [Capture listing](/quest/m3/capture-list.md) - adds fields to the listing structs +- [Capture exclusion](/quest/m4/capture-exclude.md) - adds the exclusion to the display selection diff --git a/quest/m2/api-mesh-enrollment.md b/quest/m2/api-mesh-enrollment.md new file mode 100644 index 00000000000..3c9722bb737 --- /dev/null +++ b/quest/m2/api-mesh-enrollment.md @@ -0,0 +1,24 @@ +# [XS] A mesh enrollment reply can grow + +## Goal + +`moq_auth::mesh::Enrollment` and the other reply and request structs the mesh +module exports are `#[non_exhaustive]` with constructors where a caller builds +one, so a later field (a certificate, a revocation hint) is additive. The cut +publishes the module for the first time, and today `Enrollment` is a plain +struct with a public `token: String`. + +## Plan + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): + +- ✅ Fence the structs now (rejected: publishing them open and breaking + later). +- ✅ The mesh [relay node](/quest/m3/mesh/node.md) takes a credential handle + and the application enrolls however it likes (rejected: the coordinator and + `Enrollment` learning to issue certificates; deferring). So this quest adds + no certificate field; it only keeps one possible. + +Sweep `rs/moq-auth/src/mesh/` for every public struct and enum a caller +matches or builds, and fence each the same way. Public API: breaking for a +caller that builds or exhaustively destructures one. Wire: none. diff --git a/quest/m3/auth-key-validation.md b/quest/m2/auth-key-validation.md similarity index 89% rename from quest/m3/auth-key-validation.md rename to quest/m2/auth-key-validation.md index 89a38890570..2471137d1bb 100644 --- a/quest/m3/auth-key-validation.md +++ b/quest/m2/auth-key-validation.md @@ -11,6 +11,10 @@ ## Plan +Moved to m2 2026-11-12 (maintainer, /quest-audit before the cut): the fix +makes `@moq/auth`'s `Key.public` throw, and `KeySet.parse` and Rust's `Jwk` +parse reject keys they accept today, so it is a break the cut should carry. + Decided 2026-11-08 (maintainer, /quest-plan of an external code audit of `9f51fa0966`): milestone ✅ m3 (rejected: m1), audit scores 5, 3 and 3/10. Under the cross-package sync rule, Rust's `Jwk::import` also checks algorithm against key type eagerly. Today that check runs only when a key is diff --git a/quest/m2/cluster-routes/README.md b/quest/m2/cluster-routes/README.md index 2e7e945cab0..4f9c28e3070 100644 --- a/quest/m2/cluster-routes/README.md +++ b/quest/m2/cluster-routes/README.md @@ -59,6 +59,10 @@ The items below were decided 2026-11-10 (maintainer, /quest-plan on lite-07 rout 28. ✅ Decided 2026-11-12 (maintainer, /quest-plan on the quests still written for Babel), asked how item 27's removal is split and ordered: **the router first with no wire change, then a switch of the sessions, and the switch lands before the route stream.** [The router](https://github.com/moq-dev/moq/pull/8622) runs the path rule with a route per scope, sans-IO, beside today's router. [The switch](/quest/m2/cluster-routes/switch.md) puts it on the lite-07-wip wire, the draft, the Rust sessions, and `js/net`, and deletes Babel's router. [The interest set](/quest/m2/cluster-routes/interest-set.md) then edits the path router only, and [the route stream](/quest/m2/cluster-routes/route-stream.md) follows the switch with no echo or route request to carry. The switch's recommended wire puts the answer that ends a hold on the publisher half, behind the routes its sender sent first, so the path rule needs no order between streams that today's sessions lack for one prefix. [Acceptance](/quest/m2/cluster-routes/acceptance.md) is of the switched sessions, and the inductive invariant stays in m1, rewritten for the path, without `LoopFree` as an invariant. Rejected: the router quest in m1; one quest for router, wire, draft, sessions, and `js/net`; the route stream first, which builds the echo and ROUTE_REQUEST on the new stream and then deletes them; one wire change for the path rule and the one stream together; moving the inductive invariant to m4; deleting it. Reversed 2026-11-13 (maintainer, /quest-complete): ✅ the inductive invariant is deleted, since Apalache cannot step even one transition of the route model ([moq-dev/moq#8564](https://github.com/moq-dev/moq/pull/8564)), no Rust TLA+ checker qualifies ([moq-dev/moq#8600](https://github.com/moq-dev/moq/pull/8600), [moq-dev/moq#8669](https://github.com/moq-dev/moq/pull/8669)), and no more JVM model checking is wanted here (rejected: keeping it paused). +29. ✅ Decided 2026-11-12 (maintainer, /quest-audit before the cut), asked how the path rule keeps the `upstream` guarantee, which the prototype left out and [the switch](/quest/m2/cluster-routes/switch.md) had listed for removal with Babel's machinery: **a route keeps one sticky down-only flag beside its path.** A relay sets it on a route it learns over a link marked `upstream`, it is never cleared on a later hop, and a relay never sends a flagged route over an upstream link. The flag is export policy, not loop avoidance, so the path does not replace it: a relay that hears a CDN route from a LAN peer learned it on a link that is not upstream, and only the flag tells it the route came down from an upstream. That case is a local cluster with more than one uplink ([outbound federation](/quest/m3/cluster-routing/federation.md)), where a relay whose own uplink died reaches the CDN through a peer; an edge that dials only its cores needs just the local rule. This is the Only-To-Customer attribute of BGP roles (RFC 9234). The router (`rs/moq-net/src/route/path.rs`, built in [#8622](https://github.com/moq-dev/moq/pull/8622)) carries the flag and its export check, [the switch](/quest/m2/cluster-routes/switch.md) adds them to `RouteHandoff.tla` and keeps the flag on ROUTE_START and ROUTE_UPDATE, and [acceptance](/quest/m2/cluster-routes/acceptance.md)'s two-uplink mesh proves it. Sender side only, as today. Rejected: a mark on each hop of the path, which gives the same answer from more bytes and model state; the local rule with no flag, which either leaks through a LAN peer or gives up the dead-uplink fallback; also refusing a flagged route that arrives from a link whose grant marks the sender as a customer, which can be added later without a wire change. + +30. ✅ Decided 2026-11-12 (maintainer, /quest-audit before the cut), after a survey of BGP, EIGRP, and their successors against item 27: most of the path rule already matches that prior art (AS_PATH loop detection, root-cause withdrawal, graceful restart, End-of-RIB, outbound route filtering, and item 29's Only-To-Customer flag). Three gaps are taken, all in [the switch](/quest/m2/cluster-routes/switch.md) and [acceptance](/quest/m2/cluster-routes/acceptance.md): **a dead-hop message on a scoped session applies only to the routes heard on that session and casts no doubt**, since in BGP a neighbour withdraws only what it sent, and a scoped peer (a federated relay, later a browser) could otherwise name a core hop it only read in a path and drop healthy routes outside its scope; **holds are bounded in time**, a neighbour's missing answer counting as nothing after a deadline, and a session whose announce stream has stopped draining is closed, which avoids EIGRP's stuck-in-active and BGP's zombie routes (RFC 9687), and is safe because a hold protects presence, not loop freedom; and **three tests before any design change**: neighbours noticing a dead relay up to the 3 s keep-alive apart (to size the doubt timer, measured in hops of delay today), a three-relay priority dispute wheel (RFC 4264's wedgies, since priority ranks above cost), and a primary that fails and returns. No wire bytes change; the first defines what the dead-hop message means. Rejected: declaring the link relation in SETUP and refusing leaked routes on receipt (RFC 9234), since the accepting side already knows the relation from its grant and it adds wire for a misconfiguration; MRAI and route-flap damping, which exist because BGP delays announcements; link state; per-hop signatures. + The route layer's core landed in [moq-dev/moq#7676](https://github.com/moq-dev/moq/pull/7676): a crate-private, sans-IO `Router` in `rs/moq-net/src/route/`, generic over the @@ -94,7 +98,7 @@ Settled by the core, for review: session. Keep the 2026-10-01 two-layer direction, reconciled with the maintainer's -2026-10-13 [decisions](/quest/m2/cluster-routing/README.md) and the routing +2026-10-13 [decisions](/quest/m3/cluster-routing/README.md) and the routing contract below. The routing contract's remaining model fixtures live in `rs/moq-net/tests/integration/routing_contracts/` and run against a small reference model; port each topology to the implementation as acceptance, then delete the model. diff --git a/quest/m2/cluster-routes/acceptance.md b/quest/m2/cluster-routes/acceptance.md index 7f0cde75004..c938ceed927 100644 --- a/quest/m2/cluster-routes/acceptance.md +++ b/quest/m2/cluster-routes/acceptance.md @@ -98,6 +98,14 @@ benchmark smoke alone does not enforce complexity, so add operation-count regressions. Bound held routes, doubts, and remembered dead hops under same-clock churn. +Decided 2026-11-12 (maintainer, /quest-audit; epic Plan item 30): three +fixtures from routing prior art, as tests before any design change. A +relay vanishes and its neighbours notice up to the 3 s keep-alive apart, +which shows whether the doubt timer must follow the keep-alive or relay +keep-alives must shorten. Three relays whose local priorities form a dispute +wheel settle, or the test pins how they fail (RFC 4264). A primary route +fails and returns, and every relay ends back on it. + ## Required - [lite-07 sessions run the path router](/quest/m2/cluster-routes/switch.md) - the sessions and the route rule these fixtures accept diff --git a/quest/m2/cluster-routes/playwright-pin.md b/quest/m2/cluster-routes/playwright-pin.md index 1f683a3f8ad..e2b41f3d2f9 100644 --- a/quest/m2/cluster-routes/playwright-pin.md +++ b/quest/m2/cluster-routes/playwright-pin.md @@ -8,7 +8,7 @@ draft-ietf-webtrans-http3, so the route key lane (`test/interop/clients/js/route-key.ts`) proves the Chromium tab. The lane already requires a proof from Chromium 158 on, so the bump changes no expectation. Firefox's exporter departs from the draft in every build so far; -[its own quest](/quest/m2/cluster-routes/browser-proof.md) carries +[its own quest](/quest/m5/routes-browser-proof.md) carries that bump. The root flake's `pkgs.playwright-driver` supplies `PLAYWRIGHT_BROWSERS_PATH` diff --git a/quest/m2/cluster-routes/route-exhaustive-sim.md b/quest/m2/cluster-routes/route-exhaustive-sim.md index e9e6dbff8a4..bf96557f709 100644 --- a/quest/m2/cluster-routes/route-exhaustive-sim.md +++ b/quest/m2/cluster-routes/route-exhaustive-sim.md @@ -43,7 +43,7 @@ out. - A search that takes seconds joins the merge gate; a larger one runs in Nightly. -This sits beside [trace validation](/quest/m1/route-trace-validation.md), which +This sits beside [trace validation](/quest/m3/route-trace-validation.md), which keeps the TLA+ model honest for the design work done on the spec before code exists. It replaces neither that nor the seeded sweeps on larger meshes (`tests/integration/mesh_sweep.rs`). If the spike shows the @@ -61,4 +61,4 @@ Public API: none. Wire: none. ## Related -- [Route sweeps are checked against the TLA+ model](/quest/m1/route-trace-validation.md) - guards the model against the code; this checks the code itself +- [Route sweeps are checked against the TLA+ model](/quest/m3/route-trace-validation.md) - guards the model against the code; this checks the code itself diff --git a/quest/m2/cluster-routes/switch.md b/quest/m2/cluster-routes/switch.md index e71904da705..1963f7e07fc 100644 --- a/quest/m2/cluster-routes/switch.md +++ b/quest/m2/cluster-routes/switch.md @@ -60,19 +60,31 @@ else, the model wins: - ROUTE_START and ROUTE_UPDATE carry the path as hops, each a Node ID and the session that node learned the route on, and the prefix the route - covers as the model settled it. The seqno and the down-only bit go. The + covers as the model settled it. The seqno goes. The down-only flag stays as + a route attribute beside the path (epic Plan item 29, decided 2026-11-12: + set when a route is learned over an `upstream` link, never cleared, never + sent over one), which amends this quest's first draft, where it went. The metric stays as the cost that ranks routes, and no longer decides which are safe. The hop limit stays the wire's 32. - One message names a dead hop. A relay sends it once per session it loses, to every neighbour, covering every publisher, and it lists the routes that moved to another path, as the prototype has it. The same message carries DRAIN - entries, each naming a node that sent GOAWAY, covering every publisher. + entries, each naming a node that sent GOAWAY, covering every publisher. On a + scoped session it applies only to the routes heard on that session and casts + no doubt (epic Plan item 30), so a peer cannot drop routes it only read in a + path. - The answer that ends a hold rides the publisher half: the stream that carries the answering side's routes to the relay that waits. It is then ordered behind every route its sender sent first, on today's streams. A relay waits only on a neighbour it listens to, since no other can hand a - stale route back. + stale route back. A neighbour that has not answered by a deadline counts + as answering with nothing, and a session whose announce stream has stopped + draining for a set time is closed (epic Plan item 30); a hold protects + presence, not loop freedom, so either is safe. - ROUTE_HOLD stays as the model has it (epic Plan items 14 and 19). +- `RouteHandoff.tla` gains the down-only flag and its export rule (epic Plan + item 29), which the router already has, so the model checks the rule the + wire carries. - ROUTE_REQUEST and the echo are removed with their types. The Node ID a request names (epic Plan item 2) stays. - The doubt is local: nothing on the wire carries it. Its timer is the @@ -135,7 +147,9 @@ Work: and no relay moves onto a route through it early; a hold ends behind a neighbour that answers and behind one whose session closes; an unknown hop named in a withdrawal changes nothing; a path at the hop limit is - refused. + refused; a scoped peer's dead-hop message leaves routes heard on other + sessions alone; a hold ends at its deadline behind a neighbour that never + answers; and a peer that stops reading its announce stream is closed. - Measure route and claim messages per change with `rs/moq-net/tests/integration/mesh_sweep.rs`'s benchmark before and after, and put both in the PR. diff --git a/quest/m2/first-publish-versions.md b/quest/m2/first-publish-versions.md new file mode 100644 index 00000000000..495ccd35d5b --- /dev/null +++ b/quest/m2/first-publish-versions.md @@ -0,0 +1,28 @@ +# [XS] New experimental packages first publish at 0.0.x + +## Goal + +`moq-rtsp` and `@moq/stats` are at `0.0.1` and `moq-bench` sets +`publish = false`, before the cut publishes them for the first time. At the +cut `sh/version.py` moves every package not at `0.0.x` onto the shared version +line for good, and all three are new on `main` at `0.1.0`. + +## Plan + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ experiments +start at `0.0.x` as [Versions](/RELEASE.md#versions) says, and moq-bench stays +off crates.io since it is a binary with no dependents (rejected: only +unpublishing moq-bench; leaving all three on the shared line). `moq-quic` and +`moq-congestion` are also new at `0.1.0` and are core, so they join the +shared line. + +- Check each dependent's version requirement after the change, and that + `just version` leaves the three alone. +- [Claim new npm names](/quest/m3/release/npm-first-publish.md) still claims + `@moq/stats`: a `0.0.x` package publishes too. +- Sweep for any other package first publishing at this cut above `0.0.x` + that the list above misses (compare `main`'s packages against the + registries), and bring each to the maintainer rather than deciding it. + +Public API: none. Wire: none. A version on the shared line cannot go back, +which is why this lands before the cut. diff --git a/quest/m2/group-drops.md b/quest/m2/group-drops.md index c46847b7eeb..1d74fb8506b 100644 --- a/quest/m2/group-drops.md +++ b/quest/m2/group-drops.md @@ -55,18 +55,16 @@ every run; the 256-stream cell sometimes does. for a named consumer (for example a player or container reader that skips gaps), per the API audit rule that new surface needs one; name it in the PR or leave the mirror out. -- **Order with the drop set ([moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476)).** The drop set - replaces `poll_drop`'s `Delivered` argument and the skipped record, and - edits the lite-07 SUBSCRIBE_DROP wording. Publish the API against its - final disposition record, so the public shape doesn't change twice. +- **After the drop set.** This waited (2026-11-05) for Drop set, which + rewrote the lite-07 SUBSCRIBE_DROP wording and the drop signal, and + finished in [moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476) on + 2026-11-12. `poll_drop` now takes no `Delivered` argument and yields the + given-up run with its `Hole` (`rs/moq-net/src/model/track.rs`, + `track/hole.rs`). Publish the API against that record. - The drop set owns the per-cursor drop runs and their reasons. This quest reports what a cursor gives up; it doesn't change how a cursor judges a missing group. -Decided 2026-11-05 (maintainer, /quest-audit): ✅ this waits for -Drop set ([moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476)), since both rewrite the lite-07 -SUBSCRIBE_DROP draft wording and the drop signal's `Delivered` argument. - Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): ✅ key the drop signal by group start and land after [time addressing's model](/quest/m2/time-addressing/model.md), so its public API never ships sequence-keyed (rejected: landing first and rekeying with @@ -100,4 +98,4 @@ Wire: possibly lite-07 (SUBSCRIBE_DROP for a stopped group), decided here. - [Address groups by timestamp](/quest/m2/time-addressing/model.md) - the drop range is keyed by start ## Related -- [The receive-windows benchmark](/quest/m1/receive-windows-evicted.md) - its overcommitted cells wait on this (skipped by [moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838) until it lands) +- [The receive-windows benchmark](/quest/m3/receive-windows-evicted.md) - its overcommitted cells wait on this (skipped by [moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838) until it lands) diff --git a/quest/m2/lite07-finalize.md b/quest/m2/lite07-finalize.md index 069aeb32d36..ca4fb04e4ed 100644 --- a/quest/m2/lite07-finalize.md +++ b/quest/m2/lite07-finalize.md @@ -24,7 +24,7 @@ until the maintainer calls the cut. This quest is the mechanical rename and checks done at that call; the Required list is the work already known to land first, not a closed freeze set. -The route layer of [Cluster routing](/quest/m2/cluster-routing/README.md) +The route layer of [Cluster routing](/quest/m3/cluster-routing/README.md) (ROUTE_START/UPDATE/END, path-less ANNOUNCE, the native hop list dropped) lands in lite-07, decided 2026-10-05. That removes lite-07's `Hop Base`/`Hop Keep` announce compression along with the hop list, and moq.pro's lite-07 rollout @@ -79,7 +79,7 @@ parameter, has landed, and the rest is API work that does not block the rename. The archive line's TRACKS request (a per-broadcast track listing that replaced SUBSCRIBE_ALL) landed in [#6158](https://github.com/moq-dev/moq/pull/6158) and [#6167](https://github.com/moq-dev/moq/pull/6167), and the padding stream -type for the browser [padding probes](/quest/m3/probe-padding.md) landed in +type for the browser [padding probes](/quest/m5/probe-padding.md) landed in [#6233](https://github.com/moq-dev/moq/pull/6233). Decided 2026-10-30 (maintainer, on [moq-dev/moq#7390](https://github.com/moq-dev/moq/pull/7390)): those probes diff --git a/quest/m3/mesh/device-bounds.md b/quest/m2/mesh-device-bounds.md similarity index 100% rename from quest/m3/mesh/device-bounds.md rename to quest/m2/mesh-device-bounds.md diff --git a/quest/m2/mesh-grant-shape.md b/quest/m2/mesh-grant-shape.md new file mode 100644 index 00000000000..12e234606d1 --- /dev/null +++ b/quest/m2/mesh-grant-shape.md @@ -0,0 +1,43 @@ +# [S] Mesh grants scope nodes and services apart + +## Goal + +A mesh member's grant publishes `mesh/node//**` and the two service +patterns `mesh/service/*/.server/` and `mesh/service/*/.client/`, in +place of `mesh//**`. A service named after another member can no longer +put `.server/` under its own grant and answer as that member. The +shape lands before the cut publishes `moq_auth::mesh` for the first time, so +no enrolled device ever holds the old grant from a release. + +## Plan + +Split from the [Endpoint API](/quest/m3/mesh/endpoint.md) on 2026-11-12 +(maintainer, /quest-audit before the cut): ✅ the grant shape lands in m2 on +its own (rejected: landing with the Endpoint after the cut, with the +coordinator issuing both shapes for one release). `rs/moq-auth/src/mesh` +exists on `main` and not on `release`, so there is nothing to migrate and no +dual-grant rollout. + +The shape was decided 2026-11-07 (maintainer, /quest-complete, from the +review of [moq-dev/moq#8214](https://github.com/moq-dev/moq/pull/8214)): +members own `mesh/node//**` and each service lives at +`mesh/service/` (rejected: `mesh/.service/`; keeping +`mesh/` and refusing service names shaped like an endpoint ID). A +service name is one path segment, so `*` in the grant covers it. + +- `moq_auth::mesh`'s coordinator issues the new patterns + (`rs/moq-auth/src/mesh/coordinator.rs`), with its docs + (`rs/moq-auth/src/mesh/mod.rs`) and contract tests. +- `doc/bin/relay/auth.md`'s mesh enrollment section names the new grant. +- moq-crypto's and `@moq/crypto`'s sign examples and tests that use + `mesh//agent` paths move to `mesh/node//agent`. +- A hosted coordinator implementing the JSON contract issues the same grant; + the PR's Impact section says so. Applications built on `main` re-enroll. + +Public API: the grant a mesh coordinator issues changes; no Rust or JS +signature does. Wire: none. + +## Related + +- [Mesh enrollment is bounded](/quest/m2/mesh-device-bounds.md) - the other mesh contract change before the cut +- [Mesh enrollment can grow](/quest/m2/api-mesh-enrollment.md) - fences the enrollment reply diff --git a/quest/m2/recut-held.md b/quest/m2/recut-held.md deleted file mode 100644 index 039c9f6930c..00000000000 --- a/quest/m2/recut-held.md +++ /dev/null @@ -1,30 +0,0 @@ -# [XS] Finish the milestone re-cut - -## Goal - -The epics that open claims held in place during the 2026-11-05 milestone -re-cut sit where the re-cut's rules put them: the probe epic moves from m2 to -m3. The "held" notes in the m2 and cut gate READMEs are removed, along with -each held entry's extra link from the gate's `Required`. Stream-credit stays -in m1 (maintainer, 2026-11-13): with takeover moved to m3 it breaks no wire, -and m2 no longer requires it. - -## Plan - -`quest mv` refuses to move a claimed quest, since its branch is its path, and -a child must stay listed by its own epic, so each epic waited for its claimed -children. Move each epic with `quest mv` once nothing in it is claimed; an -epic whose claim finished all its work is already gone. Before moving, -re-check what remains against the root [questline](/quest/README.md)'s rules: -an epic with no breaking work left stays in m1, and one with no gating work -left leaves the cut gate for m1. If a child is claimed again before this runs, -leave its epic in place and keep this quest for it. -Drop-set is a single quest, not an epic, so it has no `Required` entry here: -when this runs, move it if it still exists and nothing claims it. - -Prose that names a held epic's milestone becomes stale when it moves; update -it in the same PR. The browser padding probes already moved to m3 on 2026-11-06. - -## Required - -- [Browser send estimate](/quest/m2/probe/browser-send-estimate.md) - claim held probe diff --git a/quest/m3/rtmp-rtmps-tcurl.md b/quest/m2/rtmp-rtmps-tcurl.md similarity index 86% rename from quest/m3/rtmp-rtmps-tcurl.md rename to quest/m2/rtmp-rtmps-tcurl.md index f03cff13025..9d3981310c5 100644 --- a/quest/m3/rtmp-rtmps-tcurl.md +++ b/quest/m2/rtmp-rtmps-tcurl.md @@ -20,6 +20,13 @@ quest in m3, since moq-dev/studio's [destinations](https://github.com/moq-dev/studio/blob/main/quest/m0/destinations.md) wait on it to push to RTMPS ingests. +Moved to m2 2026-11-12 (maintainer, /quest-audit before the cut): changing +`Client::connect`'s signature breaks moq-rtmp's public API, so it lands before +the cut. ✅ The new `connect` takes a `#[non_exhaustive]` options value built +from the URL (rejected: positional arguments again), so a later +[dial policy](/quest/m3/webrtc-dial-policy.md) joins it without a second +break. + Today (`rs/moq-rtmp/src/dial.rs`): - `Client::connect(addr, app)` takes a resolved `SocketAddr`, dials plaintext diff --git a/quest/m2/subscribe-ranges/README.md b/quest/m2/subscribe-ranges/README.md index 05d6654b06c..7a56021b112 100644 --- a/quest/m2/subscribe-ranges/README.md +++ b/quest/m2/subscribe-ranges/README.md @@ -1,106 +1,26 @@ -# Subscribe ranges +# [S] Subscribe ranges end to end ## Goal -A moq-lite-07 SUBSCRIBE replaces FETCH for anything wider than one group. One subscription asks for any number -of group ranges, past or live, bounded by group or by time, in ascending or -descending order, and the -publisher serves their union from one cursor without sending a group twice. -Every sequence in range either arrives or is named by a SUBSCRIBE_SKIP or -SUBSCRIBE_DROP, so -nothing blocks on a cache miss. Relays fill misses upstream with ranges, not -one request per group, including over moq-transport where group numbers may be -sparse. Lite-07 FETCH is single-group only: one group, or a slice within one -group, which SUBSCRIBE cannot express (maintainer, 2026-11-03, replacing the -earlier removal). +One test holds the finished subscribe-ranges work together: a relay with a +sparse cache answers a wide past range plus live, in both orders, over lite-07 +and moq-transport, and every sequence in range is delivered or dropped. ## Plan -Decided in planning (09-29), after the fetch-span agent (#4558) found that a -relay bounds a sparse FETCH only if the model can request ranges: - -- **Wire (lite-07, unpublished).** SUBSCRIBE carries a list of group ranges - (an open end means live; ends may be relative, below) in place of Group Start/End, plus - `order: asc | desc` (desc, newest first, is the default). Order landed in - #5385; over moq-transport the upstream order is fixed at SUBSCRIBE, since no - draft lets an update change it. - Subscriber Max Age (the model's `max_delay`) stays as a cap alongside the - ranges, since an untimed track's groups have no timestamp to bound. Decided 2026-10-10: - historical requests also obey that cap and publisher max-age. An omitted - FETCH budget means publisher max-age alone; translation to a range must - not accidentally give it the live SUBSCRIBE zero-delay default. SUBSCRIBE_UPDATE - replaces the list. Lite-07 FETCH is single-group only (decided 2026-11-03, - replacing its removal); published versions keep it, answered through the same model. -- **Holes.** Lite-07 names every sequence in the requested ranges the - publisher won't deliver: SUBSCRIBE_SKIP one that never existed, and - SUBSCRIBE_DROP (restored in lite-07) one expired or refused upstream. On lite-01..06 datagram groups are always dropped in a past range (#4982); on lite-07 ranges never cover datagrams, which have their own track kind. The - publisher delivers what it has and fills the rest as upstream answers. -- **Model.** `Subscription` gains `ranges` beside its `order`, mirroring the wire; - `request_groups` as a separate API was rejected. A publisher's - `track::Requests` receives a `track::Fill` per run of misses (landed). - Single-group FETCH stays (2026-11-03), so `fetch_group` stays too. This - breaks published moq-net APIs. -- **moq-transport.** Some deployments use non-contiguous group numbers, so a - relay never probes upstream one group at a time. Upstream, each run of - locally missing groups becomes one standalone range FETCH, capped below the - upstream's live group (which arrives via SUBSCRIBE), so it never blocks. - moq-transport caps a FETCH at the Largest Object, so it can return the live - group's existing prefix but never wait for its future objects. Downstream, - our publishers answer only a one-group FETCH or fill and refuse wider ones - (maintainer, 2026-10-28, - [moq-dev/moq#7204](https://github.com/moq-dev/moq/pull/7204)), and the - IETF scope freeze (`doc/concept/standard.md`) adds no hole-marker - extension that would let a cache answer a range. - [Range FETCH over known holes](/quest/m4/ietf-range-fetch.md) later answers - a range whose every group is cached or proven never produced. A FETCH of the track's newest live group is refused - unless its range ends below that group's live edge - ([moq-dev/moq#7944](https://github.com/moq-dev/moq/pull/7944), - `poll_settle` in `rs/moq-net/src/ietf/publisher.rs`), replacing the - 2026-10-25 rule that it reads the open group whole; an older group still - arriving is waited on until it finishes or its edge passes the range. -- **Relative ends** (maintainer, 2026-10-11, superseding the 10-07 `Live` - flag and closing #5085). Each range end is `Absolute(group)` or - `Relative(n)`, naming the group n before the latest; either end may be - relative, and an omitted end is live. An offset past the track's start - clamps to group 0. Before any group exists, relative ends resolve when the - first one arrives, so `Relative(0)..` starts at it. The cross-language and - bridge tests cover both edges. Ends are inclusive of whole groups: - a relative start resolves to frame 0 of that group, and a relative end to - just after its last frame (the model's exclusive `Position::after_group`), - so `Absolute(0)..Relative(0)` includes the latest group. Live is `Relative(0)..`, so no separate `Live` field exists. - Merging subscribers is the union of their ranges, so a range above the - live edge never hides the latest group from a merged live subscriber, - which is the starvation the flag existed to prevent. Subscriber Max Age - still caps every range, so `Absolute(0)..` means everything not stale. - A relay that cannot resolve a relative end locally forwards it upstream as - relative over lite-07; a moq-transport bridge resolves it to an absolute - Group ID first, from its cache or the Largest Object of a live join - (decided 2026-10-27). -- **Time ends** (maintainer, 2026-10-17, for DVR seek in the - [archive design](/doc/concept/archives.md)). A range end may also be - `Time(pts)`, in the track's timescale. Whoever holds the track resolves it: - the cache from its groups' timestamps, or a `track::Requests` handler such as - an archive consumer. A start resolves to the group containing pts, and an - end to just after that group. A lite-07 relay that cannot resolve one - forwards it upstream unchanged. An untimed track refuses a time end, and a - moq-transport bridge resolves it locally or refuses it, since moq-transport - has no time filter. It is part of lite-07 and must land before - [lite-07 finalizes](/quest/m2/lite07-finalize.md), which this line gates. This amends the 09-29 - premise that ranges are group-only because a group's timestamp isn't always - known: timed tracks know it, and untimed ones refuse. The cache half landed in its own PR after #5929. - A datagram was judged on its own (maintainer, 2026-10-20). Superseded - 2026-11-01 (maintainer): datagrams become their own track kind, so on lite-07 - a range never covers a datagram. -- Ranges are whole groups: a range always starts at a group's first frame - (maintainer, on [moq-dev/moq#6465](https://github.com/moq-dev/moq/pull/6465)). The IETF joining FETCH for the - live group's uncached prefix stays, because today's bridge relies on it - (maintainer, 09-29). -- The relay half of fetch-span moves here; #4558 landed only the no-handler - skip. - -This line owns the end-to-end test: a relay with a sparse cache answers a -wide past range plus live in both orders, over lite-07 and moq-transport, with -every sequence delivered or dropped. +Every child of the old epic has landed (the last, lite-07 FETCH as a +single-group request, in [moq-dev/moq#8616](https://github.com/moq-dev/moq/pull/8616)), +and the audit checked what is left against the integration suite on 2026-11-14. +Covered in pieces: `subscribe_ranges.rs` (sparse and merged ranges, widening, +time ends, over lite and moq-transport), `range_read.rs` (contiguous history and +live-edge ranges, ascending), `group_order.rs` (both orders through a relay on +every wire, contiguous live groups), `range_fill.rs` (lite-07 recovery), and +`drop_record.rs` (every sequence delivered or dropped through a relay). + +Missing is the combination: a relay cache with holes, a past range wider than +what it holds plus the live edge, read in ascending and descending order, on +lite-07 and on moq-transport. Add it beside those tests, reusing their +fixtures, and delete this quest in that PR. ## Related diff --git a/quest/m2/time-addressing/model.md b/quest/m2/time-addressing/model.md index 43a8d3bb0f1..02065c0c8d9 100644 --- a/quest/m2/time-addressing/model.md +++ b/quest/m2/time-addressing/model.md @@ -57,8 +57,9 @@ a producer would need per-track memory of its last number. wire, Lite-07 ranges ([moq-dev/moq#7872](https://github.com/moq-dev/moq/pull/7872), merged 2026-11-05) and Delete the old start bookkeeping (finished in [moq-dev/moq#8613](https://github.com/moq-dev/moq/pull/8613)), - not the whole epic. Drop set ([moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476)) lands first too: its - runs keyed by group sequence convert mechanically to starts. + not the whole epic. Drop set landed first, as planned + ([moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476), 2026-11-12): + its runs keyed by group sequence convert mechanically to starts. - **Old relays** (decided 2026-10-26, maintainer): a released relay fills a hole over lite-05/06 with one FETCH per integer ID, so sparse IDs cost it one FETCH per tick in the hole. Accepted: "This was a lingering problem. @@ -140,10 +141,11 @@ recorded only sequencing. and moq-net-tests. - **One anchor rule.** Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): ✅ collapse `Anchor` to `{Oldest, Relative(n), Time(ts)}` with every end inclusive (rejected: keep - `Absolute` and make it inclusive; only add `Oldest`). Today an `Absolute` - end is exclusive while `Relative` and `Time` ends include their group - (`rs/moq-net/src/model/subscription.rs`), and "oldest" is the sentinel - `Relative(u64::MAX)`. Once keys are starts, `Absolute(key)` and `Time(ts)` + `Absolute` and make it inclusive; only add `Oldest`). `Oldest` landed in + [moq-dev/moq#8204](https://github.com/moq-dev/moq/pull/8204), replacing the + `Relative(u64::MAX)` sentinel, so what remains is `Absolute`: today an + `Absolute` end is exclusive while `Relative` and `Time` ends include their + group (`rs/moq-net/src/model/subscription.rs`). Once keys are starts, `Absolute(key)` and `Time(ts)` name the same group, so `Absolute` goes; a legacy exclusive end `E` maps to `Time(E - 1)`, since legacy keys pass through. A legacy inclusive start `S` keeps its bound: it resolves to the first group keyed at or @@ -153,13 +155,11 @@ recorded only sequencing. `Subscription::end()`) selects nothing, so it is dropped as an empty range before conversion, never saturated to `Time(0)`, which would include group 0. Regressions cover `E = 0` and `E = 1` beside the legacy round - trips. `Oldest` is explicit and - encoded the same on every wire (lite-07, lite-06 and earlier, IETF), where - today the lite encoder treats any `back > MAX_QUIC` as oldest and the - model does not. `Range::groups(RangeBounds)` and + trips. `Oldest` stays explicit on every wire, as #8204 left it: lite-07 + sends it as no anchor and every relative count as it is, and older lite + and IETF carry only `Oldest` and `Relative(0)` as starts. + `Range::groups(RangeBounds)` and `Subscription::with_range` go. JS's omitted start keeps meaning `Oldest`. - This absorbs the deleted anchor-oldest quest (from the 2026-11-07 - invariant audit, model-4). - **Fence what grows after the cut.** Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): `Position` becomes `#[non_exhaustive]`, so byte-holes can add a byte offset without a break. - **Wires without a start.** Check which moq-lite versions can encode a diff --git a/quest/m2/usage-derive-settings.md b/quest/m2/usage-derive-settings.md new file mode 100644 index 00000000000..1e676f8b773 --- /dev/null +++ b/quest/m2/usage-derive-settings.md @@ -0,0 +1,39 @@ +# [S] Settings and the unknown-env check come from usage's derive + +## Goal + +`rs/moq-tokio/src/settings.rs` and `rs/moq-relay/src/settings.rs` (about 600 +lines declaring every merged flag a second time) are deleted, and +`moq_tokio::cli::check_env` with its `UnknownEnv` error is replaced by usage's +`unknown_env` node. Each flag is declared once, on its `#[derive(Cli)]` field. +Source presence, precedence, deprecated-name rejection, provenance, and the +refusal of an unknown `MOQ_*` variable keep working. CLI behavior and wire are +unchanged. + +## Plan + +Both halves shipped upstream in usage-lib +[7.2.0](https://github.com/jdx/usage/releases/tag/v7.2.0) on 2026-11-03 +(checked 2026-11-04, /quest-unblock), and the workspace is on `usage-rs` +7.3.0: + +- Derive flags declare their own settings + ([jdx/usage#1610](https://github.com/jdx/usage/pull/1610)), which replaces + the hand-written registries and the `Registry::drift` test that kept the two + declarations in step. +- `#[usage(unknown_env(prefix = "MOQ_", allow(...)))]` + ([jdx/usage#1609](https://github.com/jdx/usage/pull/1609)) refuses unknown + variables under a prefix; `allow` takes the names read outside the spec, + such as `moq_video::ENV`. + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ one quest in +m2, merging the two upstream-wait quests (rejected: adopting after the cut +behind deprecated shims; keeping the drift test and both modules). +`moq_tokio::settings` is a `pub mod` with public structs, and +`cli::check_env` and `UnknownEnv` are public and used by moq-cli, moq-relay, +and moq-bench, so deleting them breaks moq-tokio's API. That is cheap only +before the cut. + +Public API: removes `moq_tokio::settings`, `moq_tokio::cli::check_env`, and +`moq_tokio::cli::UnknownEnv`; the PR's Impact section says what replaces +each. Wire: none. diff --git a/quest/m3/README.md b/quest/m3/README.md index d6876dd4315..8b91777dfea 100644 --- a/quest/m3/README.md +++ b/quest/m3/README.md @@ -3,8 +3,7 @@ ## Goal Work that follows the next breaking release and has someone waiting on it: -byte streams and the mesh libraries first, then release execution after the -cut, quests a named consumer needs, including downstream deployments, and bug fixes in +release execution after the cut, quests a named consumer needs, including downstream deployments, and bug fixes in shipped code. Additive, API, and wire work mix here; the cut after m2's sorts them by kind into m1, m2, and m3. @@ -26,10 +25,44 @@ Dependencies on m1 and m2 remain explicit. An implementation that discovers a new necessary API break brings that decision to the maintainer before proceeding. -Decided 2026-11-03 (maintainer, Kixel's quest audit): [byte streams](/quest/m3/stream/README.md) +Decided 2026-11-03 (maintainer, Kixel's quest audit): [byte streams](/quest/m1/stream/README.md) and the [mesh libraries](/quest/m3/mesh/README.md) rank first here, ahead of [release execution](/quest/m3/release/README.md). Kixel's daemon builds on them from `main` through a git submodule, so it needs them merged, not released. They stay out of m1 and m2 so the cut does not wait on them; each starts as soon as its prerequisites (sockets, fastest-reader pacing, outbound -federation) land. +federation) land. Superseded 2026-11-12 (maintainer, /quest-audit): both moved +to [m1](/quest/m1/README.md), since the cut no longer waits on m1. The mesh +libraries came back 2026-11-13, below; byte streams stay in m1. + +Decided 2026-11-12 (maintainer, /quest-audit before the cut), from a read of +every quest against the code and the downstream quest trees: + +- ✅ Quests with no named consumer and no bug in shipped code moved to + [m4](/quest/m4/README.md) (rejected: keeping them here with a reason + each): the intra-refresh epic, the vendored WebRTC cleanups, JS byte + streams, the watch worker, the moq-transport refactors, and smaller + cleanups. +- ✅ Quests gated on an outside party or shelved moved to + [m5](/quest/m5/README.md): the ALSA runtime-loading pair, the OBS upstream + epic, padding and the JS rung disable, the two GPU vendor proofs, and + others. +- ✅ Promoted here from m4, each with a downstream deployment or application + now waiting: Windows capture verification, the stock HLS player test, the + Kotlin, Swift, and Dart interop clients, VA-API render tests on real + hardware, and the WHIP/WHEP browser test. + +Decided 2026-11-13 (maintainer, /quest-plan): no quest requires one in a later +milestone, since a blocker ranks by its own milestone and the earlier quest +would stall behind it. + +- ✅ The 14 m1 quests that waited on m2 work moved here, the cluster-routing + and mesh epics whole (rejected: m2, where the cut would wait on them; + pulling their 12 breaking m2 blockers into m1). The + [m1 README](/quest/m1/README.md) lists them. +- ✅ [cli-packaging](/quest/m3/cli-packaging.md)'s blockers, + [capture-alsa-link](/quest/m3/capture-alsa-link.md) and the + [cpal wait](/quest/m3/cpal-alsa-runtime.md), moved here from m5 (rejected: + moving cli-packaging and nixpkgs-cli to m5, which waits for the cut). +- m3 is not gated, unlike m4 and m5, because it holds the + [release cut](/quest/m3/release/cut.md) and its reviews. diff --git a/quest/m3/audio-encode-flush.md b/quest/m3/audio-encode-flush.md index 391c6d8c3be..4164cd46ff5 100644 --- a/quest/m3/audio-encode-flush.md +++ b/quest/m3/audio-encode-flush.md @@ -16,4 +16,4 @@ non-consuming `Encoder::drain` is `pub(super)` (`rs/moq-audio/src/encode/encoder.rs`). A studio media file's sound ([moq-dev/studio#44](https://github.com/moq-dev/studio/pull/44)) loses its last ~26 ms when a clip ends but the source stays live. Related: [An audio track's final flush joins its open -group](/quest/m3/audio-flush-group.md). +group](/quest/m4/audio-flush-group.md). diff --git a/quest/m3/audio-lang.md b/quest/m3/audio-lang.md index ee758da8688..3a40341ca69 100644 --- a/quest/m3/audio-lang.md +++ b/quest/m3/audio-lang.md @@ -14,10 +14,12 @@ Reported in [moq-dev/moq#7945](https://github.com/moq-dev/moq/issues/7945). Today audio renditions have no language field in either catalog (`js/hang/src/catalog/audio.ts`, `rs/hang/src/catalog/audio/mod.rs`), while text renditions already have `lang` (`js/hang/src/catalog/text.ts`). Watch -ranks every enabled audio rendition together (`ranked()`), so two languages -compete on bitrate. `controls.audio.target` can already pin a rendition by -name. The hang-to-MSF export sets `altGroup = 1` on all audio -(`rs/moq-mux/src/catalog/msf/producer.rs`), marking languages as switchable +picks from every enabled audio rendition together (`#select` in +`js/watch/src/audio/source.ts`, the first by container kind), so with two +languages it takes whichever comes first. `controls.audio.target` can already +pin a rendition by name. The hang-to-MSF export sets `altGroup = 1` on all +audio once a broadcast has more than one audio rendition +(`rs/moq-mux/src/catalog/producer.rs`), marking languages as switchable bitrates, and the MSF-to-hang consumers keep `lang` for text only. Decided 2026-11-05 (maintainer, /quest-import): diff --git a/quest/m3/audio-latest-reach-back.md b/quest/m3/audio-latest-reach-back.md index 7cef2998013..69adea73b63 100644 --- a/quest/m3/audio-latest-reach-back.md +++ b/quest/m3/audio-latest-reach-back.md @@ -6,7 +6,7 @@ ## Plan -Decided 2026-10-30 (maintainer, follow-up planning): m2. Use the same start resolution the container consumer uses rather than a second one. With `min` and warmup at 0, the start is unchanged. Add a test of a cold join with a 2-group-old start. [Audio join budget](https://github.com/moq-dev/moq/pull/7342) changes the same `joining_latest` path in `rs/moq-audio/src/decode/consumer.rs`; whichever lands second builds its start on the other's. +Decided 2026-10-30 (maintainer, follow-up planning): use the same start resolution the container consumer uses rather than a second one. With `min` and warmup at 0, the start is unchanged. Add a test of a cold join with a 2-group-old start. [Audio join budget](https://github.com/moq-dev/moq/pull/7342) has merged; the `Start::Latest` join in `rs/moq-audio/src/decode/consumer.rs` still narrows to `live_edge.group..`, while moq-video's decoder subtracts its lookback (`rs/moq-video/src/decode/consumer.rs`), so build on that. Public API: none. Wire: none. diff --git a/quest/m3/audit-code.md b/quest/m3/audit-code.md index e7fe0f9eb9c..e30769d7025 100644 --- a/quest/m3/audit-code.md +++ b/quest/m3/audit-code.md @@ -52,7 +52,7 @@ lints [moq-dev/moq#6576](https://github.com/moq-dev/moq/pull/6576). - Still open from m0: `ietf/session.rs` `start()` builds both halves twice across the draft arms; lite and IETF `GroupServe` share their open, flush, - and FIN-settle driver; `ietf/adapter.rs` reimplements `kio::Queue`. + and FIN-settle driver. - The codec importers each forward `demand`, `flush`, `finish`, `abort`, `cut`, `discontinuity`, and `seek` to their container producer. Dropping them means a shared track handle, which is an API proposal. diff --git a/quest/m3/audit-cost.md b/quest/m3/audit-cost.md index b0782ff9252..417e8da058d 100644 --- a/quest/m3/audit-cost.md +++ b/quest/m3/audit-cost.md @@ -116,8 +116,8 @@ tests [moq-dev/moq#7600](https://github.com/moq-dev/moq/pull/7600), build time `rs/moq-quic/README.md` (decided 2026-10-28). Only a [vendor sync](/quest/m3/vendor-sync/quinn.md), which reviews the full upstream range, advances the recorded upstream commit. -- Waiting on upstream: [ashpd](/quest/m4/ashpd-maintained.md) and - [askama](/quest/m4/askama-maintained.md); the root Bun overrides for glob, +- Waiting on upstream: [ashpd](/quest/m1/ashpd-maintained.md) and + [askama](/quest/m5/askama-maintained.md); the root Bun overrides for glob, minimatch, chokidar, postcss-selector-parser, and Vite, each removable once its parent allows a safe range; `deny.toml`'s proc-macro-error2 ignore, until core-models moves to hax-lib 0.4; paste through tikv-jemalloc-ctl; diff --git a/quest/m3/audit-test-coverage.md b/quest/m3/audit-test-coverage.md index 7eca1714560..fe3bf7c19cd 100644 --- a/quest/m3/audit-test-coverage.md +++ b/quest/m3/audit-test-coverage.md @@ -37,7 +37,7 @@ Previous run, with its per-crate and per-package baselines: workflow reaches; the local-only ones are listed in `sh/gh/test-inventory.allow` with their owning quests. - Sample `fix` commits for regression tests, counting `#[kio::test]`. -- Whether [interop bindings](/quest/m4/interop-bindings.md) and +- Whether [interop bindings](/quest/m3/interop-bindings.md) and [GPU CI](/quest/m3/gpu-ci.md) landed, before re-auditing the tests they own. - Each TLA+ model's `verify-*` skill against the code it covers: run `/maintain-formal-verification-skill` on every one whose Coverage files diff --git a/quest/m1/buffered-ascending.md b/quest/m3/buffered-ascending.md similarity index 91% rename from quest/m1/buffered-ascending.md rename to quest/m3/buffered-ascending.md index 2c03fd41ae6..699abdb3145 100644 --- a/quest/m1/buffered-ascending.md +++ b/quest/m3/buffered-ascending.md @@ -16,6 +16,11 @@ ordering for the Python or FFI audio consumers. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + Motivated by the moq.dev Voice demo: each TTS reply is a fresh track whose head arrives newest-first, which shifted Chrome's decoder timeline and overflowed the ring (moq-dev/moq#6523), then clipped each reply's opening @@ -38,7 +43,7 @@ Decided 2026-10-23 by the maintainer: max-delay release rules go. Not chosen: delete the hold outright; keep #6549's rules unchanged. - ✅ The hold stays audio-only. Video's startup hold is the poster quest's. -- ✅ Placed in m1, requiring [Finalize moq-lite-07](/quest/m2/lite07-finalize.md): +- ✅ Placed in m1 (superseded 2026-11-13: m3, above), requiring [Finalize moq-lite-07](/quest/m2/lite07-finalize.md): lite-06 and earlier serve an ascending request newest-first, so this is a no-op on default sessions until lite-07 negotiates by default. `release` has no `order` field, so it ships in the next minor, not a 0.6.x patch. diff --git a/quest/m3/c/README.md b/quest/m3/c/README.md index ee1d4c2982a..39f425a9142 100644 --- a/quest/m3/c/README.md +++ b/quest/m3/c/README.md @@ -11,14 +11,15 @@ scope; they already sit on moq-ffi through the generated C++. ## Plan Decided 2026-10-25 (maintainer): this line ships after the cut, once moq builds -its bindings from rust2's UniFFI fork. The hand-written crate ships at the cut +its bindings from [rust2's exporter](/quest/m3/ffi-exporter.md) (amended +2026-11-02, when moq chose the exporter over rust2's UniFFI fork). The hand-written crate ships at the cut instead, with the binding features moq-ffi gained in m1 and m2 ([moq-dev/moq#8060](https://github.com/moq-dev/moq/pull/8060)). Features moq-c never had, such as single-group fetch and group-level consumption, wait for this -line. The generator fix for the C name -clash landed in rust2's fork -([moq-dev/rust2#66](https://github.com/moq-dev/rust2/pull/66)), not in a -`kixelated/uniffi-bindgen-cpp` release, and reaches moq with the switch. +line. The C generator's GLib-style names, which fix +the C name clash, come from rust2 +([moq-dev/rust2#66](https://github.com/moq-dev/rust2/pull/66)), not from a +`kixelated/uniffi-bindgen-cpp` release, and reach moq with the switch. Decided earlier, still standing: diff --git a/quest/m3/cache-routing/README.md b/quest/m3/cache-routing/README.md index 0c244a6744d..ec9dae6a653 100644 --- a/quest/m3/cache-routing/README.md +++ b/quest/m3/cache-routing/README.md @@ -24,7 +24,7 @@ Decided 2026-11-08 (maintainer, /quest-plan): 1. ✅ **Node routes are the cold cost, and optional per-broadcast warm routes are the warm cost.** The per-origin ROUTE metric of - [Cluster routing](/quest/m2/cluster-routing/README.md) stays the cold + [Cluster routing](/quest/m3/cluster-routing/README.md) stays the cold cost, static and one per origin. A relay carrying a broadcast may advertise a warm route for it. This revises Cluster routing's non-goal of cache-dependent route prices. (rejected: per-broadcast forwarding with a @@ -214,9 +214,9 @@ This README's own work, once its children land: broadcasts by the hash. - Docs: warm routes and how to price skip links in `doc/bin/relay/cluster.md`, and a section on the routing concept page once the - [routing guide](/quest/m1/routing-guide.md) has added it. + [routing guide](/quest/m3/routing-guide.md) has added it. ## Related -- [Cluster routing](/quest/m2/cluster-routing/README.md) - the node routes and feasibility distances this ranks on +- [Cluster routing](/quest/m3/cluster-routing/README.md) - the node routes and feasibility distances this ranks on - [Relays split horizon per session](/quest/m3/route-split-horizon.md) - per-session suppression of node routes, beside warm routes' per-neighbor rule diff --git a/quest/m3/cache-routing/select.md b/quest/m3/cache-routing/select.md index d271cbfaeb0..96f4a3d4ba0 100644 --- a/quest/m3/cache-routing/select.md +++ b/quest/m3/cache-routing/select.md @@ -44,6 +44,11 @@ change the implementation needs updates the draft's Warm Routes section in the same PR, in lite-07 while it is unpublished and in the next version after. +Noted 2026-11-12 (/quest-audit): this plan predates the path rule (epic Plan +item 8, [moq-dev/moq#8533](https://github.com/moq-dev/moq/pull/8533)). Its +one-hop warm routes and feasibility-distance eligibility are rewritten once +[moq-dev/moq#8457](https://github.com/moq-dev/moq/pull/8457) settles. + ## Required - [Prove cache-aware routing](/quest/m2/cache-routing-proof.md) - the rule for routes that propagate, its fixtures, and the wire this implements diff --git a/quest/m3/capture-alsa-link.md b/quest/m3/capture-alsa-link.md index aa2d78fb920..a5240363bce 100644 --- a/quest/m3/capture-alsa-link.md +++ b/quest/m3/capture-alsa-link.md @@ -9,6 +9,11 @@ libraries. ## Plan +Moved from m5 to m3 2026-11-13 (maintainer, /quest-plan): +[cli-packaging](/quest/m3/cli-packaging.md) in m3 requires it, and no quest +requires one in a later milestone. ✅ m3 (rejected: moving cli-packaging and +nixpkgs-cli to m5, which waits for the cut). + Nix already covers the dev-time side (alsa-lib, alsa-plugins, and pipewire are in the dev shell). What remains is the runtime linkage: cpal 0.18's `alsa` dependency is non-optional on Linux, so libasound becomes a load-time @@ -24,6 +29,8 @@ release offers it, and the in-tree routes are hacks), so this waits on enable whatever it needs, and update the moq-audio and moq-cli feature comments that call libasound a build-time link. +Moved to m5 2026-11-12 (maintainer, /quest-audit): it waits only on an upstream cpal release. Superseded 2026-11-13: it is back in m3, per the decision above. + Verify by building in the Nix shell, then running the shipped binary on a host without libasound: it starts, lists devices, and captures where a backend exists. Inspect the binary to prove no load-time requirement on diff --git a/quest/m3/capture-linux.md b/quest/m3/capture-linux.md index 6bbfbe05472..0578b34a721 100644 --- a/quest/m3/capture-linux.md +++ b/quest/m3/capture-linux.md @@ -19,13 +19,15 @@ unimplemented, because xdg-desktop-portal owns source selection: - Portal-picked windows reach the CLI: #5089 added `capture::Source::Portal(portal::Selection)`, and #5566 sends a bare `--window` to the portal (`doc/bin/cli.md`). -- **Display selection stays picker-driven.** On Wayland a bare `--display` - goes to the portal and a non-X11 id is refused. Decided 2026-10-26 - (maintainer): the token becomes persistable (API plus CLI), so a - previously-approved source is reused without the picker. #8562 added the API - (`Selection::restore_token` watches it, `with_restore_token` seeds it), but - the CLI does not persist it yet, so a new `moq` run still shows the picker. - [Wayland studio sources](https://github.com/moq-dev/studio/blob/main/quest/m0/wayland-sources.md) in moq-dev/studio consume it. +- **Display selection stays picker-driven in the CLI.** On Wayland a bare + `--display` goes to the portal and a non-X11 id is refused. Decided + 2026-10-26 (maintainer): the token becomes persistable (API plus CLI), so a + previously-approved source is reused without the picker. The API half is + done: [moq-dev/moq#8562](https://github.com/moq-dev/moq/pull/8562) added + `Selection::restore_token` and `with_restore_token` + (`rs/moq-video/src/capture/portal.rs`), which a downstream application + consumes. What remains is the CLI: nothing in `rs/moq-cli` persists or + seeds the token, so a new `moq` run still shows the picker. - **System audio** needs a PipeWire monitor node. `moq_audio::capture::Source::System` exists but is macOS-only and returns `Unsupported` elsewhere. This is independent of the screencast portal. @@ -41,6 +43,10 @@ unimplemented, because xdg-desktop-portal owns source selection: Windows process tree and the macOS application. The Windows and macOS halves share the same `moq_audio::capture` source. +## Waiting + +- Maintainer: whether app capture exists off macOS, here and in [Windows capture parity](/quest/m3/capture-windows.md). (A) Linux and Windows offer window capture only: `Source::App` stays macOS-only and fails with its reason elsewhere, and a listing off macOS uses apps only to group windows. (B) Build app capture by composing an app's windows. I recommend A, since no consumer asks for B and it matches the capture listing's rule that a listed app must not hand out a source that fails at open. + ## Related - [Windows capture parity](/quest/m3/capture-windows.md) - the same gaps, diff --git a/quest/m3/capture-list.md b/quest/m3/capture-list.md index e7fcc4018f1..2f43ab66683 100644 --- a/quest/m3/capture-list.md +++ b/quest/m3/capture-list.md @@ -161,6 +161,16 @@ Tests, sized to what each runner has: Public API: breaking for `capture::{Camera, Display, Window, App}` and the listing calls. Wire: none. +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ this quest +stays after the cut, and +[capture listings and selections can grow](/quest/m2/api-capture-fences.md) +fences the listing structs it adds fields to in m2 (rejected: pulling this quest into m2; a second moq-video +break). What this adds is then additive. + +## Required + +- [Capture listings and selections can grow](/quest/m2/api-capture-fences.md) - the fenced structs this adds fields to + ## Related - [macOS capture listing futures are Send](/quest/m3/capture-shareable-send.md) - the ScreenCaptureKit listing this rewrites; whatever replaces it stays `Send` diff --git a/quest/m4/capture-wgc.md b/quest/m3/capture-wgc.md similarity index 72% rename from quest/m4/capture-wgc.md rename to quest/m3/capture-wgc.md index f5be091f5a5..5ae9ff09520 100644 --- a/quest/m4/capture-wgc.md +++ b/quest/m3/capture-wgc.md @@ -28,7 +28,16 @@ result in the PR: Full-display capture with the cursor on was verified on Windows 11 25H2 with one monitor. +Moved to m3 2026-11-12 (maintainer, /quest-audit): a downstream application +now runs display and window capture on Windows. + +Note: the Windows host on hand has one monitor and one discrete GPU. Run +every item it can; the multi-monitor, hybrid-GPU laptop, and Windows 10 +items need hardware nobody has, so record them as not run and leave them to +[video hardware validation](/quest/m5/video-hardware.md) rather than holding +this quest open for them. + ## Related - [Windows capture parity](/quest/m3/capture-windows.md) - app capture and system audio, which WGC does not answer -- [Direct3D11 render import](/quest/m3/render-d3d11.md) - the other end of keeping Windows frames on the GPU +- [Direct3D11 render import](/quest/m4/render-d3d11.md) - the other end of keeping Windows frames on the GPU diff --git a/quest/m3/capture-windows.md b/quest/m3/capture-windows.md index dd93a41746c..1e9d6ffe138 100644 --- a/quest/m3/capture-windows.md +++ b/quest/m3/capture-windows.md @@ -3,8 +3,9 @@ ## Goal App capture and system audio work on Windows. Display and window capture, -including the cursor, move to -[Windows.Graphics.Capture](/quest/m4/capture-wgc.md). +including the cursor, already run on the Windows.Graphics.Capture backend +(`rs/moq-video/src/capture/wgc/native.rs`), which +[its own quest](/quest/m3/capture-wgc.md) verifies on real hardware. ## Plan @@ -27,11 +28,15 @@ including the cursor, move to WGC does not answer app capture. It captures one monitor or one window per item and has no per-process source, so the per-window-composition or -window-only decision still stands after that backend lands. +window-only decision still stands now that the backend has landed. + +## Waiting + +- Maintainer: whether app capture exists off macOS, here and in [Linux capture parity](/quest/m3/capture-linux.md). (A) Windows and Linux offer window capture only: `Source::App` stays macOS-only and fails with its reason elsewhere, and a listing off macOS uses apps only to group windows. (B) Build app capture by composing a process tree's windows. I recommend A, since no consumer asks for B and it matches the capture listing's rule that a listed app must not hand out a source that fails at open. ## Related -- [Windows.Graphics.Capture](/quest/m4/capture-wgc.md) - display and window capture with the cursor, which this quest no longer owns +- [Windows.Graphics.Capture](/quest/m3/capture-wgc.md) - display and window capture with the cursor, which this quest no longer owns - [A capture listing a picker can draw](/quest/m3/capture-list.md) - lists apps on Windows and gives them the identity an app capture would name - [Linux capture parity](/quest/m3/capture-linux.md) - the same gaps, through the portal and PipeWire diff --git a/quest/m3/cli-packaging.md b/quest/m3/cli-packaging.md index 5ae81868996..c07f012e420 100644 --- a/quest/m3/cli-packaging.md +++ b/quest/m3/cli-packaging.md @@ -68,6 +68,10 @@ and `moq-audio/capture` explicitly, and `full` enables `capture`. Renamed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): the moq-cli feature that enables everything is named `full`, not `release`, so this quest and its followers say `full`. +## Waiting + +- Maintainer: what ships while cpal has no release that loads libasound at runtime (RustAudio/cpal#1421 is open; the last release is 0.18.2). (A) Wait: this quest stays blocked on [ALSA without a load-time link](/quest/m3/capture-alsa-link.md) until the cpal release ships. (B) Split: ship `full` now on macOS and Windows, which link no system audio library, and leave only Linux waiting on the cpal release. (C) Linux release binaries take libasound as a load-time requirement. (D) Fork cpal, alsa, and alsa-sys. I recommend B. + ## Required - [Audio capture without runtime system libraries](/quest/m3/capture-alsa-link.md) - the microphone path must start without system audio libraries before every distribution can ship it diff --git a/quest/m2/cluster-routing/README.md b/quest/m3/cluster-routing/README.md similarity index 95% rename from quest/m2/cluster-routing/README.md rename to quest/m3/cluster-routing/README.md index 262227065bc..0a259809ddb 100644 --- a/quest/m2/cluster-routing/README.md +++ b/quest/m3/cluster-routing/README.md @@ -14,6 +14,11 @@ plan was deleted 2026-11-09 for want of a consumer. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + Decided with the maintainer, 2026-10-13: - Keep per-origin ROUTEs and path-less ANNOUNCEs on the ordered announce @@ -95,7 +100,7 @@ Decided with the maintainer, 2026-10-13: sent routes learned from other connections. Relays default on; applications choose per P2P connection (on for redundancy, off for conference offload). Browser tabs stay off until the - [Rust-to-TypeScript](/quest/m5/rs2ts-js.md) port. Applications + [Rust model runs in the browser](/quest/m5/rs2ts-js.md). Applications explicitly select P2P connections; discovery does not dial everyone. - Each CDN broadcast enters a local network once by configuration: local peers outrank a backup uplink, and origin aggregation joins the rest. @@ -143,7 +148,7 @@ are the cold cost that [cache-aware routing](/quest/m3/cache-routing/README.md)' per-broadcast warm routes rank against (decided 2026-11-08, maintainer). Each implementation updates the docs it makes stale. The separate -[routing guide](/quest/m1/routing-guide.md) explains the combined design and +[routing guide](/quest/m3/routing-guide.md) explains the combined design and setup after the capabilities land. ## Required @@ -153,7 +158,7 @@ setup after the capabilities land. ## Related - [P2P](/quest/m4/p2p/README.md) - application-selected peer sessions feed the shared routing model -- [Routing guide](/quest/m1/routing-guide.md) - concept page and federation setup; P2P extends it in m4 +- [Routing guide](/quest/m3/routing-guide.md) - concept page and federation setup; P2P extends it in m4 - [Route wakes](/quest/m4/route-wakes.md) - local work follows affected fronts rather than the whole covered table - [Link quality](/quest/m5/link-quality.md) - optional measured link cost, distinct from delivery-health failover - [Cache-aware routing](/quest/m3/cache-routing/README.md) - per-broadcast warm routes ranked against these node routes diff --git a/quest/m2/cluster-routing/federation.md b/quest/m3/cluster-routing/federation.md similarity index 98% rename from quest/m2/cluster-routing/federation.md rename to quest/m3/cluster-routing/federation.md index ff7cbc6802b..d9239984dd7 100644 --- a/quest/m2/cluster-routing/federation.md +++ b/quest/m3/cluster-routing/federation.md @@ -79,4 +79,4 @@ Rust/JS parity. Run `just check`, `just test interop --all`, and `just drafts ch ## Related -- [Routing guide](/quest/m1/routing-guide.md) - runnable operator setup after integration +- [Routing guide](/quest/m3/routing-guide.md) - runnable operator setup after integration diff --git a/quest/m3/cpal-alsa-runtime.md b/quest/m3/cpal-alsa-runtime.md index e98cfaeedf8..0fcc65f7061 100644 --- a/quest/m3/cpal-alsa-runtime.md +++ b/quest/m3/cpal-alsa-runtime.md @@ -21,19 +21,30 @@ release is out, delete this quest. ## Plan +Moved from m5 to m3 2026-11-13 (maintainer, /quest-plan): +[cli-packaging](/quest/m3/cli-packaging.md) in m3 requires it, and no quest +requires one in a later milestone. ✅ m3 (rejected: moving cli-packaging and +nixpkgs-cli to m5, which waits for the cut). + Decided 2026-10-08: the capture chain (this quest, [capture-alsa-link](/quest/m3/capture-alsa-link.md), and -[Ship capture and playback](/quest/m3/cli-packaging.md)) moves to m2 and parks -until cpal ships runtime ALSA loading. Nothing is posted upstream and no fork -is taken meanwhile. +[Ship capture and playback](/quest/m3/cli-packaging.md)) parks until cpal +ships runtime ALSA loading, with no fork taken meanwhile. + +Moved to m5 2026-11-12 (maintainer, /quest-audit): it waits only on an upstream cpal release. Superseded 2026-11-13: it is back in m3, per the decision above. Filed 2026-11-02 (maintainer-approved, /quest-unblock): [RustAudio/cpal#1421](https://github.com/RustAudio/cpal/issues/1421), recommending runtime loading first (most naturally in alsa-sys) and an optional ALSA host second. Related, not duplicates: RustAudio/cpal#1404 (FreeBSD-only `alsa` feature), cpal#929, and diwic/alsa-sys#10 (static linking). -State as of 2026-10-05: cpal 0.18.2 (pinned here) and cpal master (0.19.0) +State as of 2026-11-12: #1421 is open. A cpal maintainer replied on +2026-11-04 that an optional ALSA host would break existing builds and that +`default_host()` would then have no host to return on a headless Linux +build, and asked whether static linking would do; nothing is decided. cpal's +latest release is still 0.18.2 (2026-08-16), the version pinned here. + +State as of 2026-10-05: cpal 0.18.2 and cpal master (0.19.0) both depend on `alsa` unconditionally on Linux and the BSDs. `alsa` 0.12.1 depends on `alsa-sys` 0.6.1, whose build script links libasound through -pkg-config and offers no runtime-loading mode. No upstream issue or PR asks -for either. +pkg-config and offers no runtime-loading mode. Two upstream shapes, either of which unblocks [Audio capture without runtime system libraries](/quest/m3/capture-alsa-link.md): diff --git a/quest/m3/cpp-ffi-reuse.md b/quest/m3/cpp-ffi-reuse.md index 7c542787ebd..d0543de0fa0 100644 --- a/quest/m3/cpp-ffi-reuse.md +++ b/quest/m3/cpp-ffi-reuse.md @@ -56,3 +56,14 @@ Keep this quest open for the published release input: Follow-up to [#5549](https://github.com/moq-dev/moq/pull/5549), which shared the default FFI build across Go, Kotlin and Swift. + +Checked 2026-11-12 (/quest-audit): `cpp/moq/FFI_VERSION` is still unset and +`cpp/moq/FFI_SHA256.json` is `{}`. moq-ffi releases still come from `release` +(moq-ffi-v0.4.25 on 2026-11-11 ships per-target libraries, not the four +archives), and `main`'s `release-shared-ffi.yml`, which runs +`just cpp ffi archive`, differs from `release`'s, so the first release that +can satisfy the pin is the cut's. + +## Required + +- [Release cut](/quest/m3/release/cut.md) - the first moq-ffi release that publishes the four default-feature archives diff --git a/quest/m3/crates-trusted-publisher.md b/quest/m3/crates-trusted-publisher.md index 5072291a2b0..1e1b11d0363 100644 --- a/quest/m3/crates-trusted-publisher.md +++ b/quest/m3/crates-trusted-publisher.md @@ -7,9 +7,13 @@ for `moq-dev/moq`, workflow `release-rs.yml`, and environment `publish`, so [Rust releases publish through OIDC](/quest/m3/release-rs-oidc.md) can start. Delete this quest once every crate has one. +Corrected 2026-11-12 (/quest-audit): moq-bench leaves the list, since +[first-publish versions](/quest/m2/first-publish-versions.md) sets it +`publish = false`; seven crates remain. + ## Waiting -- Maintainer: after [the cut](/quest/m3/release/cut.md) publishes kio-macros, moq-bench, +- Maintainer: after [the cut](/quest/m3/release/cut.md) publishes kio-macros, moq-c, moq-congestion, moq-crypto, moq-quic, moq-rtsp, and moq-webrtc by token, add their trusted publishers the same way. Settings pages need a crates.io login, so check with a `release-rs.yml` run that publishes through OIDC. @@ -17,8 +21,8 @@ Delete this quest once every crate has one. ## Plan Decided 2026-11-05 (maintainer left the choice to the agent, /quest-unblock): -✅ move this quest and [Rust releases publish through OIDC](/quest/m3/release-rs-oidc.md) -to m2 behind [the cut](/quest/m3/release/cut.md). The seven crates the cut +✅ this quest and [Rust releases publish through OIDC](/quest/m3/release-rs-oidc.md) +wait behind [the cut](/quest/m3/release/cut.md). The seven crates the cut first-publishes by token cannot get a trusted publisher before it, and the cut runs after m1, so neither quest could finish inside m1. Not chosen: keeping both in m1 and treating the cut as part of finishing m1. @@ -49,7 +53,8 @@ Settings pages need a crates.io login, so the check is a `release-rs.yml` run publishing through OIDC. Done 2026-11-04 (maintainer, /quest-unblock): all 31 published crates have -the trusted publisher. Only the 7 unpublished crates remain. +the trusted publisher. Only the unpublished crates remain: as of 2026-11-12 +crates.io has none of the seven in Waiting. Decided 2026-11-05 (maintainer, /quest-audit): moq-e2ee, published but retired in favor of moq-crypto, gets no trusted publisher; diff --git a/quest/m3/dash-reconfigure.md b/quest/m3/dash-reconfigure.md index 72cc36c818f..5ea0178dad2 100644 --- a/quest/m3/dash-reconfigure.md +++ b/quest/m3/dash-reconfigure.md @@ -38,4 +38,4 @@ Public API: none. Wire: none. ## Related -- [HLS and DASH browser playback](/quest/m4/hls-player-test.md) - dash.js survives reconfiguration nightly +- [HLS and DASH browser playback](/quest/m3/hls-player-test.md) - dash.js survives reconfiguration nightly diff --git a/quest/m3/export-live-start.md b/quest/m3/export-live-start.md index 7cbed7f11c7..21c0240db35 100644 --- a/quest/m3/export-live-start.md +++ b/quest/m3/export-live-start.md @@ -10,12 +10,15 @@ track. Decided 2026-11-05 (maintainer, on moq-dev/studio's /quest-plan): m3, since the studio is a named consumer that is not urgent. -`with_max_delay` (`rs/moq-mux/src/container/fmp4/export.rs`, and -`container/mp4.rs`) sets both how far an export may fall behind and where it -starts, since it reads from the oldest group within that window. The studio +`with_delay` (`rs/moq-mux/src/container/fmp4/export.rs`, and +`container/mp4.rs`) takes a `Delay { min, max }`, but the start still follows +the max: the container consumer starts at the lowest group that arrived and +skips only groups trailing the newest content by more than `max` +(`rs/moq-mux/src/container/consumer.rs`), so it reads from the oldest group +within that window whatever `min` says. The studio recorder ([moq-dev/studio#18](https://github.com/moq-dev/studio/pull/18)) sets 2 s so a disk stall loses nothing, and so every recording also starts up to 2 s before it was started (`rs/moq-studio/src/record.rs`). Only TS export reports stats today; fMP4 and MP4 report no sample counts, which the recorder needs to say what it wrote. -Changed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): [moq-dev/moq#8298](https://github.com/moq-dev/moq/pull/8298) gives fMP4, MP4, MKV, and FLV a `moq_net::Delay { min, max }` before the cut, so splitting the start from the skip budget is an implementation fix, not a break. +Changed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): [moq-dev/moq#8298](https://github.com/moq-dev/moq/pull/8298) (merged) gives fMP4, MP4, MKV, and FLV a `moq_net::track::Delay { min, max }` before the cut, so splitting the start from the skip budget is an implementation fix, not a break. diff --git a/quest/m1/export-order.md b/quest/m3/export-order.md similarity index 90% rename from quest/m1/export-order.md rename to quest/m3/export-order.md index d0569e63419..f2f27c7f0e4 100644 --- a/quest/m1/export-order.md +++ b/quest/m3/export-order.md @@ -11,6 +11,11 @@ an 11.7 s delay, where SRT loses about 0. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + Decided 2026-11-12 (maintainer, `/quest-import` of [moq-dev/moq#8559](https://github.com/moq-dev/moq/issues/8559)): ✅ measure Ascending against Descending, then pick (rejected: Ascending for @@ -38,7 +43,7 @@ The maintainer's reasoning, to keep in view: - The right order may depend on the maximum latency. The starting hypothesis to measure mirrors -[Buffered playback subscribes in ascending order](/quest/m1/buffered-ascending.md): +[Buffered playback subscribes in ascending order](/quest/m3/buffered-ascending.md): an export with a nonzero playout delay (`Delay::min`) wants its window from the start, and goes Ascending; one that chases the live edge stays Descending. diff --git a/quest/m3/ffi-exporter.md b/quest/m3/ffi-exporter.md index b14d3cc3d19..149b93a8630 100644 --- a/quest/m3/ffi-exporter.md +++ b/quest/m3/ffi-exporter.md @@ -24,39 +24,14 @@ exporter. Until the switch, moq keeps building from its pinned [the vendor sync](/quest/m3/vendor-sync/README.md). Not chosen: publish the fork and switch to it first; decide later. -Carried over from the fork plan, where they still apply to the exporter's -generators (each needs re-checking against the exporter's design): +The bullets below carried over from the fork plan; each still applies to the +exporter's generators and needs re-checking against its design. The fork-era +choices (UniFFI as the base, a hard fork in rust2 published as the +`rust2-uniffi*` crates) no longer apply: the fork is not published, and +moq-ffi builds on the exporter's pinned revision with no `[patch]` and no git +dependency, so crates.io moq-ffi still builds (see the +[condition quest](/quest/m3/rust2-exporter.md)). - -Decided with the maintainer on 2026-10-13, after a review of how moq uses -UniFFI: - -- **Base**: UniFFI. [The BoltFFI spike](https://github.com/moq-dev/rust2/blob/main/bench/ffi/README.md#boltffi-comparison) - recommended it. -- **Why fork**: every UniFFI minor release breaks external generators (0.31 - changed `module_path`, 0.32 changed the metadata encoding without bumping - the contract version, and `main` removes the old library-mode API). No - external generator has released for 0.32, so moq carries three forks, each - pinned in about five places. The core also blocks work moq needs: byte - copies, zero-copy returns, and namespaced external types - (mozilla/uniffi-rs#3026, auto-closed by our - [#5498](https://github.com/moq-dev/moq/pull/5498) and still unfixed - upstream). -- **Free to diverge**: the fork may change metadata, checksums, and the FFI - contract when a moq quest needs it. A diverged contract must never mix - silently with upstream UniFFI. - -Decided with the maintainer on 2026-10-23: the hard fork lives in -`moq-dev/rust2`, not a new `moq-dev/uniffi`. rust2 groups Rust-to-other-language -tools (bindings, a TypeScript translator, build-speed work) under one setup. -moq's generator code moves with it, including `rs/uniffi-bindgen`. The import -and crates.io publishing are rust2's -[UniFFI import quest](https://github.com/moq-dev/rust2/blob/main/quest/m1/uniffi.md), -which publishes `rust2-uniffi`, `rust2-uniffi-core`, `rust2-uniffi-macros`, and -`rust2-uniffi-bindgen`. This quest is the moq side: - -- moq-ffi depends on the `rust2-uniffi*` crates directly, with no `[patch]` and - no git dependency, so crates.io moq-ffi still builds. - **Generated code stays identical**: Kotlin package names, library symbol prefixes, and every wrapper's API do not change, except Dart's `MoqRoute.instanceId()` below. Diff the regenerated bindings for every @@ -100,18 +75,13 @@ which publishes `rust2-uniffi`, `rust2-uniffi-core`, `rust2-uniffi-macros`, and `moq::Subscription{}` or generated C `moq_subscription_default()` means a max delay of 0, as every hand-written binding's does. Until then those two leave both bounds unset, and their docs say to start from - `subscription_defaults()` (accepted docs-only on #7697). Options shown: - m0; m1; ✅ m2, in this quest. + `subscription_defaults()` (accepted docs-only on #7697). It lands in + this quest. - Negative durations (2026-10-25, maintainer): the Dart negative-duration patch moq carries goes upstream into the fork's Dart generator, and the Go generator stops panicking on a negative `time.Duration` (the Go wrapper guards it today; delete the guard in the PR that pins a generator with the fix, with a test passing a negative duration through the Go binding). -- Cancelled Go results ([moq-dev/moq#7397](https://github.com/moq-dev/moq/pull/7397)): - uniffi-bindgen-go `v0.10.0-kixelated.2` destroys a result whose context was - cancelled as the call completed, and completes a future cancelled while - pending as `context.Canceled`. The exporter's Go bindings must do both, tested - like the fork's `destroy` fixture, which counts live handles. - Python running-loop preference and Dart buffer ownership (2026-10-31, from [moq-dev/moq#7461](https://github.com/moq-dev/moq/pull/7461)): [kixelated/uniffi-rs#1](https://github.com/kixelated/uniffi-rs/pull/1) @@ -126,7 +96,10 @@ which publishes `rust2-uniffi`, `rust2-uniffi-core`, `rust2-uniffi-macros`, and ([Generated C bindings](/quest/m3/c/README.md)) to point at rust2, then archive the three forks. - Cancelled Go results (2026-10-31, from - [moq-dev/moq#7397](https://github.com/moq-dev/moq/pull/7397), open): the + [moq-dev/moq#7397](https://github.com/moq-dev/moq/pull/7397)): + uniffi-bindgen-go `v0.10.0-kixelated.2` destroys a result whose context was + cancelled as the call completed, and completes a future cancelled while + pending as `context.Canceled`; the exporter's Go bindings must do both. The Go generator's fix that destroys a cancelled async call's result lives on `kixelated/uniffi-bindgen-go`'s `cancelled-result` branch and tag `v0.10.0-kixelated.2+v0.32.0`, not on the fork's `main`, like the earlier @@ -141,6 +114,10 @@ CamelCase, and the generated record `MoqAudioEncoderFinished` becomes `MoqAudioEncoderFinish` in every language; wrapper APIs do not change except Dart's `MoqRoute.instanceId()`, which takes the exporter's name. Wire: none. +## Waiting + +- Maintainer: whether the generated binding renames may land after the cut. The switch renames C types to CamelCase, `MoqAudioEncoderFinished` to `MoqAudioEncoderFinish` in every generated binding, and Dart `moq_ffi`'s `instanceId`, which would force a binding-breaking release. (A) Accept one binding-breaking release when the switch lands. (B) Count only the hand-written wrappers as public API, since they already alias these names (for example `instance()` in `dart/moq/lib/src/client.dart`), so the generated renames ship as non-breaking. I recommend B; C costs nothing either way, since no `c-v*` tag ships before the cut. + ## Required - [rust2's exporter is ready for moq](/quest/m3/rust2-exporter.md) - the exporter generates every language moq ships and moq can pin it diff --git a/quest/m3/follow-refused-retry.md b/quest/m3/follow-refused-retry.md index fc12b8c410a..19b1e0606df 100644 --- a/quest/m3/follow-refused-retry.md +++ b/quest/m3/follow-refused-retry.md @@ -14,7 +14,6 @@ ends and restarts the path, which a refused follower re-asks today. Planned 2026-11-04 with the maintainer, from a review thread declined on [moq-dev/moq#7836](https://github.com/moq-dev/moq/pull/7836) ([`follow.rs`](https://github.com/moq-dev/moq/pull/7836#discussion_r4471135777)). -Milestone m2 (maintainer). The gap: since #7836, a refused `Following` (its `refused` flag in `rs/moq-net/src/model/follow.rs`) waits for an announcement change before diff --git a/quest/m3/go-ffi-platform-modules.md b/quest/m3/go-ffi-platform-modules.md index 613ef4191e5..c3e0aece059 100644 --- a/quest/m3/go-ffi-platform-modules.md +++ b/quest/m3/go-ffi-platform-modules.md @@ -32,4 +32,4 @@ static GOPROXY layout: ## Required -- [Go module proxy](/quest/m3/go-module-proxy.md) - only our proxy can add modules under `moq.dev/moq-ffi` without another git mirror +- [Go module proxy](/quest/m1/go-module-proxy.md) - only our proxy can add modules under `moq.dev/moq-ffi` without another git mirror diff --git a/quest/m3/goaway-migration-followups.md b/quest/m3/goaway-migration-followups.md index cbbd79b9c30..ebd8e867f96 100644 --- a/quest/m3/goaway-migration-followups.md +++ b/quest/m3/goaway-migration-followups.md @@ -2,17 +2,17 @@ ## Goal -After a GOAWAY migration, moq-tokio's `Connection` closes the idle predecessor only once the replacement's announcements have synced, so a broadcast nobody subscribes to never reads as ended and started again during the switch. The JS `Connection` closes an idle predecessor early the same way, instead of holding it for `goaway.handover`. +After a GOAWAY migration, the Rust connect policy (moq-net's `Reconnect`, which moq-tokio's `Connection` drives) closes the idle predecessor only once the replacement's announcements have synced, so a broadcast nobody subscribes to never reads as ended and started again during the switch. The JS `Connection` closes an idle predecessor early the same way, instead of holding it for `goaway.handover`. ## Plan -Follow-ups of [moq-dev/moq#7462](https://github.com/moq-dev/moq/pull/7462) (merged), which added moq-net's `Session::poll_idle` / `Session::idle` (resolve once the session carries no subscription or fetch in either direction) and has moq-tokio close a migrated predecessor once it is idle (`rs/moq-tokio/src/connection.rs`, `poll_idle`), instead of waiting out `connect.goaway.handover` (10 s by default). +Follow-ups of [moq-dev/moq#7462](https://github.com/moq-dev/moq/pull/7462) (merged), which added moq-net's `Session::poll_idle` (now crate-private; ready once the session carries no subscription or fetch in either direction) and closes a migrated predecessor once it is idle, instead of waiting out `connect.goaway.handover` (10 s by default). Since [moq-dev/moq#7904](https://github.com/moq-dev/moq/pull/7904) that close lives in moq-net's connect policy: `Predecessor::poll_idle` in `rs/moq-net/src/reconnect.rs` (about line 1190 as of a86433043). -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, both parts in one quest, since the JS half should copy the Rust half's final rule rather than #7462's. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): both parts in one quest, since the JS half should copy the Rust half's final rule rather than #7462's. -- **Hold for the replacement's announce sync.** Today the idle predecessor can close as soon as the replacement connects, before the replacement's announcements reach either side, so an unsubscribed broadcast briefly reads as ended then started. Gate the close on both idle and the replacement's announce sync. That needs a signal moq-net doesn't expose yet: a per-session "announcements synced" readiness (lite-07's announce sync, and the IETF equivalent or a bound where the version can't say). Keep the handover cap as the upper bound. -- **JS early close.** `js/net/src/connection/reload.ts` holds the predecessor to `goaway.handover` (`handover(...)`, about line 670). Add an idle signal on `Connection.Session` mirroring `Session::idle` (subscriptions and fetches in either direction; announcements, track info, PROBE, and AUTH don't count), the same announce-sync gate, and a retire in `Reload` that closes the predecessor when both hold. -- Test both on simulated time over the mock transport, as #7462's `rs/moq-net/tests/integration/session_idle.rs` does: a broadcast announced on both sessions and subscribed by nobody never shows an end across the migration, and a busy predecessor still waits for its readers. +- **Hold for the replacement's announce sync.** Today the idle predecessor can close as soon as the replacement connects, before the replacement's announcements reach either side, so an unsubscribed broadcast briefly reads as ended then started. Gate the close on both idle and the replacement's announce sync. First check whether [moq-dev/moq#8375](https://github.com/moq-dev/moq/pull/8375), which reports a let-go as offline rather than END, already hides the flap from listeners; if it does, narrow this half to what still shows. That needs a signal moq-net doesn't expose yet: a per-session "announcements synced" readiness (lite-07's announce sync, and the IETF equivalent or a bound where the version can't say). Keep the handover cap as the upper bound. +- **JS early close.** `js/net/src/connection/reload.ts` holds the predecessor to `goaway.handover` (the `Draining` class and its `handover(...)` cap, about lines 815 and 1061 as of a86433043). Add an idle signal on `Connection.Session` mirroring `Session::poll_idle` (subscriptions and fetches in either direction; announcements, track info, PROBE, and AUTH don't count), the same announce-sync gate, and a retire in `Reload` that closes the predecessor when both hold. +- Test both on simulated time over the mock transport, as `rs/moq-net/src/session/tests.rs` does for `poll_idle`: a broadcast announced on both sessions and subscribed by nobody never shows an end across the migration, and a busy predecessor still waits for its readers. Public API: a moq-net readiness for the announce sync, `Connection.Session`'s idle signal in `@moq/net`, and their mirrors. Wire: none. diff --git a/quest/m3/gpu-ci.md b/quest/m3/gpu-ci.md index a3395cd81c8..e2e44ca64b9 100644 --- a/quest/m3/gpu-ci.md +++ b/quest/m3/gpu-ci.md @@ -59,15 +59,17 @@ skipping inside the Nix shell. guest, and keep the runner registered only to this repository with approval required for every external contributor's workflow, as the README requires. -- Open (2026-10-25): the [CI epic](/quest/m0/ci/README.md) dropped the VM - for the CI host in favor of kernel sandboxing, which conflicts with VFIO - passthrough. Recommendation: a container runner with the NVIDIA device - nodes and the host driver's libraries bound in. Confirm with the - maintainer before starting. +- Open since 2026-10-25: the [CI epic](/quest/m0/ci/README.md) dropped the + VM for the CI host in favor of kernel sandboxing, which conflicts with the + VFIO passthrough above. See `## Waiting`. Public API: none. Wire: none. +## Waiting + +- Maintainer: pick the GPU runner's shape, since the 2026-10-20 VM decision conflicts with the CI epic's later no-VM decision. Options: (a) a sandboxed runner container on the CI host with the NVIDIA device nodes and the host driver's libraries bound in, as [GPU runners](/quest/m9/video-quality/gpu-runners.md) already plans for the quality job; (b) a VFIO-passthrough VM from `github-runner-vm`. Recommended: (a), then drop the VFIO text above. + ## Related - [Video hardware validation](/quest/m5/video-hardware.md) - hardware paths nothing runs yet -- [Video quality hardware rows](/quest/m3/video-quality/hardware.md) - the nightly quality job that also needs this runner's GPU +- [Video quality hardware rows](/quest/m9/video-quality/hardware.md) - the nightly quality job that also needs this runner's GPU diff --git a/quest/m3/gpu-health.md b/quest/m3/gpu-health.md index c61f1a855d9..87ae2cdfa78 100644 --- a/quest/m3/gpu-health.md +++ b/quest/m3/gpu-health.md @@ -18,14 +18,14 @@ Those are the caller's. Decided 2026-10-05 in moq.pro's quest audit: vendor-neutral now, not NVIDIA-scoped. Reason: the external GPU surface (#4975), -[Vulkan encode](/quest/m3/vulkan-encode.md), and -[VA-API import](/quest/m3/vaapi-vulkan-import.md) make AMD and Intel encode +[Vulkan encode](/quest/m5/vulkan-encode.md), and +[VA-API import](/quest/m5/vaapi-vulkan-import.md) make AMD and Intel encode first-class, so an NVML-shaped API would be rewritten once they land. moq.pro's transcode admission and its benchmarks assume NVML today; they adopt this instead. Decided 2026-10-05: one device identity wherever a render node exists, the -render node's `dev_t`, shared with [VA-API import](/quest/m3/vaapi-vulkan-import.md). +render node's `dev_t`, shared with [VA-API import](/quest/m5/vaapi-vulkan-import.md). NVIDIA's device UUID maps to its node through `VK_EXT_physical_device_drm`. The external GPU surface (#4975) carries that identity as `frame::vulkan::Device { device_uuid, driver_uuid, render_node: Option }`. @@ -103,7 +103,7 @@ Other guidance: - A snapshot call, not a callback; cheap enough to poll about once a second. - The snapshot is additive and keys by the surface's identity. Publish it after the vendor implementations it reports are validated. It does not gate - [the multi-vendor release](/quest/m3/gpu-release.md) or the next cut. + [the multi-vendor release](/quest/m5/gpu-release.md) or the next cut. ✅ Decided 2026-10-16, support floor: ship per vendor with every value optional. NVIDIA's device-wide adapter lands with the values every vendor has @@ -130,5 +130,5 @@ Public API: a device capacity and health snapshot in moq-video. Wire: none. ## Related -- [VA-API encodes an external Vulkan image](/quest/m3/vaapi-vulkan-import.md) - picks a render node by the same `dev_t` -- [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - the AMD backend whose sessions this counts +- [VA-API encodes an external Vulkan image](/quest/m5/vaapi-vulkan-import.md) - picks a render node by the same `dev_t` +- [Vulkan Video encode on AMD](/quest/m5/vulkan-encode.md) - the AMD backend whose sessions this counts diff --git a/quest/m3/h265-min-width.md b/quest/m3/h265-min-width.md index 97ad64bcf8a..ca619253c26 100644 --- a/quest/m3/h265-min-width.md +++ b/quest/m3/h265-min-width.md @@ -43,4 +43,4 @@ Public API: the encode error names the minimum size; the probe stays private. Wi ## Related -- [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - may report a different minimum +- [Vulkan Video encode on AMD](/quest/m5/vulkan-encode.md) - may report a different minimum diff --git a/quest/m4/hls-player-test.md b/quest/m3/hls-player-test.md similarity index 100% rename from quest/m4/hls-player-test.md rename to quest/m3/hls-player-test.md diff --git a/quest/m3/hls-rendition-source.md b/quest/m3/hls-rendition-source.md index 64a46450c18..174a320a9e3 100644 --- a/quest/m3/hls-rendition-source.md +++ b/quest/m3/hls-rendition-source.md @@ -9,7 +9,7 @@ actual source rather than on the requesting viewer's grants. ## Plan -Decided 2026-11-03 (maintainer, `/quest-plan`): m2. A downstream HLS cache +Decided 2026-11-03 (maintainer, `/quest-plan`). A downstream HLS cache keys entries on a digest of each viewer's subscribe scope, because it can't tell which source a referenced rendition reads. Viewers holding different grants for the same broadcast then never share a cached entry. Correctness is diff --git a/quest/m3/ietf-absolute-join-race.md b/quest/m3/ietf-absolute-join-race.md index 87ebea66f40..ad81f9c0ab5 100644 --- a/quest/m3/ietf-absolute-join-race.md +++ b/quest/m3/ietf-absolute-join-race.md @@ -10,8 +10,8 @@ shows that main can't reach the collision, and this quest is deleted. ## Plan Decided 2026-11-03 (maintainer, /quest-plan of the follow-ups from -[moq-dev/moq#7680](https://github.com/moq-dev/moq/pull/7680)): milestone m2, -reproduce first. +[moq-dev/moq#7680](https://github.com/moq-dev/moq/pull/7680)): reproduce +first. - #7680's local review, while backporting the joined-head FETCH fix, flagged a possible pre-existing collision on main. An Absolute joining @@ -22,7 +22,7 @@ reproduce first. that interleaves an older-history FETCH with an Absolute join on drafts 18, 20, and 22. If the test passes, delete the quest with the test kept as coverage. -- Check [the fill redesign](/quest/m3/ietf-fill-refill.md) first. It deletes +- Check [the fill redesign](/quest/m4/ietf-fill-refill.md) first. It deletes the `Fill` handoff and may make this moot. If it lands first, rerun the test against it. @@ -30,4 +30,4 @@ Public API: none expected. Wire: none. ## Related -- [A moq-transport group head fills its hole in place](/quest/m3/ietf-fill-refill.md) - the fill redesign that may remove the race +- [A moq-transport group head is a flight](/quest/m4/ietf-fill-refill.md) - the fill redesign that may remove the race diff --git a/quest/m3/ietf-finish.md b/quest/m3/ietf-finish.md index 9109cbd1bcd..d2c12cf7046 100644 --- a/quest/m3/ietf-finish.md +++ b/quest/m3/ietf-finish.md @@ -17,8 +17,12 @@ Observed on 2026-10-22 with `just test interop --finish --all --versions ietf` - Draft 18: late readers see the finished catalog end empty, a short tail, or a refused joining FETCH (`publish done: internal error`). This is the pre-draft-22 join after a finish, which - [moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306) fixes for Rust publishers; rerun - these drafts. + [moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306) (merged) fixes + for Rust publishers. The rerun of these drafts since #7306 is still owed + (2026-11-12 audit); start there, and leave whatever still fails with a Rust + publisher once the head is held to + [the kept join](/quest/m3/ietf-join-pin-head.md), which owns the draft-18 + `tail` and `finish` cells with a Rust publisher. - Draft 22: several readers time out after the withdrawal (C and C++ on the health broadcast), JS readers see `publish done: internal error`, and native JS refuses a "malformed request stream message" from the relay. diff --git a/quest/m3/ietf-join-pin-head.md b/quest/m3/ietf-join-pin-head.md index bc4bcd4f6d8..f11a7446644 100644 --- a/quest/m3/ietf-join-pin-head.md +++ b/quest/m3/ietf-join-pin-head.md @@ -12,7 +12,7 @@ draft-18 interop `tail` and `finish` cells pass with a Rust publisher. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306): m2, as +[moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306), as recommended. [moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306) (merged) kept the join after finish. [moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306) keeps the subscription's request open after PUBLISH_DONE but does not diff --git a/quest/m4/interop-bindings.md b/quest/m3/interop-bindings.md similarity index 100% rename from quest/m4/interop-bindings.md rename to quest/m3/interop-bindings.md diff --git a/quest/m3/interop-shutdown-warnings.md b/quest/m3/interop-shutdown-warnings.md index 3fbd40ce319..7dc9ffab2e1 100644 --- a/quest/m3/interop-shutdown-warnings.md +++ b/quest/m3/interop-shutdown-warnings.md @@ -11,7 +11,7 @@ cause or shown to be expected, in which case the harness stops printing it. ## Plan Decided 2026-11-03 (maintainer, /quest-plan of the follow-ups from -[moq-dev/moq#7701](https://github.com/moq-dev/moq/pull/7701)): milestone m2. +[moq-dev/moq#7701](https://github.com/moq-dev/moq/pull/7701)). - Publisher linger: `stop_publisher` in `test/interop/interop.sh` (around lines 1138–1152) warns "publisher '$lang' was still running 10 s diff --git a/quest/m3/js-announced-coalesce.md b/quest/m3/js-announced-coalesce.md index bca61294af1..93d67ab4c2f 100644 --- a/quest/m3/js-announced-coalesce.md +++ b/quest/m3/js-announced-coalesce.md @@ -2,15 +2,19 @@ ## Goal -`@moq/net`'s `Announced` holds at most one pending event per path, as Rust's -announce consumer does, so an application that reads slowly while routes churn +`@moq/net`'s route stream (`Routes` in `js/net/src/announced.ts`, which +`origin.announced()`'s `Announce.Consumer` and `origin.routes()` both read) +holds at most one pending event per path, as Rust's announce consumer does, so an application that reads slowly while routes churn holds memory proportional to distinct paths, not to the number of changes. ## Plan Decided 2026-11-01 (maintainer, announcement backpressure interview): today -`js/net/src/announced.ts` appends every start, end, restart, and update to a -`Signal`, and `#runAnnounced` in `js/net/src/origin.ts` pushes a +`Producer.append` in `js/net/src/announced.ts` pushes every start, end, +restart, update, and offline onto `AnnounceState.queue`, a +`Signal` (as of a86433043, after +[moq-dev/moq#8443](https://github.com/moq-dev/moq/pull/8443) renamed the +`Announced` stream to `Routes`), and `#runAnnounced` in `js/net/src/origin.ts` pushes a diff on each wake. Match Rust's `OriginConsumerState`: a pending change for a path overwrites the earlier one, and a retraction of a start the reader never saw cancels it. The JS lite publisher already coalesces this way diff --git a/quest/m3/js-warmup-trim.md b/quest/m3/js-warmup-trim.md index 9765aacb1e5..30940a880e0 100644 --- a/quest/m3/js-warmup-trim.md +++ b/quest/m3/js-warmup-trim.md @@ -11,7 +11,7 @@ and a short finished track is not empty. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7384](https://github.com/moq-dev/moq/pull/7384): m2, as +[moq-dev/moq#7384](https://github.com/moq-dev/moq/pull/7384), as recommended. [moq-dev/moq#7384](https://github.com/moq-dev/moq/pull/7384) (merged) stopped `moq_audio::decode::Consumer` from arming the warmup when diff --git a/quest/m3/json-snapshot-resync.md b/quest/m3/json-snapshot-resync.md index 2e91d66e9cf..4a340944fe6 100644 --- a/quest/m3/json-snapshot-resync.md +++ b/quest/m3/json-snapshot-resync.md @@ -9,11 +9,7 @@ instead of ending. `rs/moq-json`'s snapshot consumer does the same. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: - -- m0 -- m1 -- ✅ m2 +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs). From [moq-dev/moq#7693](https://github.com/moq-dev/moq/pull/7693)'s follow-ups, where both automated reviewers asked moq-boy to keep reading diff --git a/quest/m3/lite-tail-drop-count.md b/quest/m3/lite-tail-drop-count.md index d516641e28e..699db91d210 100644 --- a/quest/m3/lite-tail-drop-count.md +++ b/quest/m3/lite-tail-drop-count.md @@ -2,15 +2,22 @@ ## Goal -From lite-05 on, a moq-lite session's tail is complete once every group up to the end has arrived or been reported in SUBSCRIBE_SKIP or SUBSCRIBE_DROP. The 1 s `ledger::GRACE` (`rs/moq-net/src/ledger.rs:19`) then no longer decides it, in Rust and JS. +From lite-05 on, a moq-lite session's tail is complete once every group up to the end has arrived or been reported in SUBSCRIBE_SKIP or SUBSCRIBE_DROP. The grace (`tail::grace`, the subscription's max delay or else the 1 s `ledger::GRACE` in `rs/moq-net/src/ledger.rs`) then no longer decides it, in Rust and JS. ## Plan -Decided 2026-10-30 (maintainer, follow-up planning): m2. Found by the audit on [moq-dev/moq#7403](https://github.com/moq-dev/moq/pull/7403). The maintainer dislikes grace periods. +Decided 2026-10-30 (maintainer, follow-up planning). Found by the audit on [moq-dev/moq#7403](https://github.com/moq-dev/moq/pull/7403). The maintainer dislikes grace periods. + +Checked 2026-11-12 at a86433043: the counting half already exists. `Tail::account` (`rs/moq-net/src/tail.rs`) records groups a SUBSCRIBE_SKIP or SUBSCRIBE_DROP accounts for, and `lite/subscriber.rs` calls it for skips, drops, and served fill ranges, so a tail whose every group is accounted for completes without waiting. The grace still decides only the gaps nothing accounts for: a group stream reset before its header arrived (the `tail.rs` module docs) and, on lite-05/06, a lost datagram. That remainder is this quest. - Count per version. lite-01..04 have no drop message, so they keep today's behaviour. - On lite-05/06, account for holes left by lost datagrams. Lite-07 stream tracks carry no datagrams, which have their own track kind. -- Guard against a peer that never accounts for a group: decide with the maintainer whether that ends the tail with an error or keeps waiting. Don't add a new grace. +- Account for a stream reset before its header, so no gap is left only to the grace. Check whether the publisher already reports such a group in a SUBSCRIBE_DROP; if it needs a new signal, that is a lite-07-wip draft change, and the Wire line below changes with it. +- Guard against a peer that never accounts for a group as the `Waiting` decision settles. Don't add a new grace. - The IETF tail keeps its grace until the moq-transport gap-signal extension (planned in [moq-dev/moq#7425](https://github.com/moq-dev/moq/pull/7425)) lands, since moq-transport can't signal gaps. Public API: none. Wire: none. + +## Waiting + +- Maintainer: decide what a lite-05+ tail does when the peer never accounts for a group. Options: (a) end the tail with an error; (b) keep waiting until the session closes; (c) wait out the subscription's max delay where one is set. Recommended: (c), and (a) on lite-07 when no max delay is set, so the 1 s constant goes away without a buggy peer hanging the tail forever. diff --git a/quest/m3/memory-bench-hang.md b/quest/m3/memory-bench-hang.md deleted file mode 100644 index 7eee2124f3e..00000000000 --- a/quest/m3/memory-bench-hang.md +++ /dev/null @@ -1,13 +0,0 @@ -# [XS] The moq-net memory bench's allocation mode finishes at 1 MiB frames - -## Goal - -`MEMORY_ALLOCS=1 cargo bench -p moq-net --features fuzz --bench memory` runs every case to completion, including `size=1048576`, and the cause of today's hang is fixed where it lives (the bench, its counting allocator, or moq-net). If the cause is in moq-net, a regression test covers it. - -## Plan - -Found by [moq-dev/moq#7522](https://github.com/moq-dev/moq/pull/7522): on its base [`c1a1412820`](https://github.com/moq-dev/moq/commit/c1a1412820) (`main` at the time), the allocation mode parks with no CPU after `streams=1/size=16384`, at `size=1048576`. A park with no CPU points at a lost wake or a flow-control stall, not slow work. - -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2. Root-cause before fixing: take a stack of every thread while parked (`gdb -batch -p -ex 'thread apply all bt'` on the parked bench process), check whether the counting allocator's own state is involved (compare a run without `MEMORY_ALLOCS`), and whether a 1 MiB frame exceeds a window or cache bound the bench's in-memory transport waits on. A lost wake in moq-net gets its own regression test; a bench-only cause is fixed in `rs/moq-net/benches/memory.rs`. - -Public API: none. Wire: none. diff --git a/quest/m3/mesh/README.md b/quest/m3/mesh/README.md index 5023c3e27d2..268b1d25a54 100644 --- a/quest/m3/mesh/README.md +++ b/quest/m3/mesh/README.md @@ -24,6 +24,11 @@ coordinator cannot extend, and Windows. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + Decided 2026-11-02 with the maintainer (Kixel's planning interview, which moved moq.pro's hosted mesh and T3 Code lines to Kixel): **moq for libraries, Kixel for the binary and UI.** This epic keeps the `Endpoint` API, @@ -36,7 +41,7 @@ moq library, media included, can point at one. Decided 2026-10-25 with the maintainer (planning interview), still in force: -- Traffic is app-level [byte streams](/quest/m3/stream/README.md) and named +- Traffic is app-level [byte streams](/quest/m1/stream/README.md) and named tracks, never IP packets. - The coordinator contract is generic and lives here: enrollment issues a relay URL and a token, and self-hosters run it with `moq auth`. A hosted @@ -78,7 +83,7 @@ Decided 2026-10-25 with the maintainer (planning interview), still in force: proving its key. A hosted coordinator may put its own sign-in in front of minting that key. - The API is iroh-shaped (not wire-compatible with iroh). -- It builds on [outbound federation](/quest/m2/cluster-routing/federation.md): +- It builds on [outbound federation](/quest/m3/cluster-routing/federation.md): a node's local relay dials its relays as `upstream` links and ranks LAN peers above them. @@ -101,13 +106,32 @@ only, enrolled through `moq auth`; an `Endpoint` on each dials the others' services, taking the LAN path between the first two and the relay path to the third, and survives a relay restart. -## Required +Decided 2026-11-12 (maintainer, /quest-audit before the cut): -- [Outbound federation](/quest/m2/cluster-routing/federation.md) - a local relay dials its relays and prefers LAN peers -- [Byte streams](/quest/m3/stream/README.md) - lossless TCP-level streams over MoQ +- ✅ (Superseded 2026-11-13: this epic is back in m3, above.) This epic and [byte streams](/quest/m1/stream/README.md) moved to m1 + from the head of m3 (rejected: staying in m3). The cut no longer waits on + m1, and a downstream application builds on them from `main` now. +- ✅ The epic no longer lists federation and the stream epic under + `Required` (rejected: keeping them). An epic's `Required` holds back every + child, so certificate credentials and the Endpoint waited on federation and + on JS streams they do not use. The relay node and the Endpoint each list + their own blocker. +- ✅ The relay node takes a credential handle and the application enrolls + however it likes (rejected: the coordinator issuing certificates; + deferring). `moq_auth::mesh`'s token enrollment stays one way to get a + handle. [Mesh enrollment can grow](/quest/m2/api-mesh-enrollment.md) + fences the reply before the cut publishes it. +- ✅ The contract changes land in m2, because the cut publishes + `moq_auth::mesh` for the first time: the + [grant shape](/quest/m2/mesh-grant-shape.md) and the + [key budget](/quest/m2/mesh-device-bounds.md). +- The JS enrollment client and the guide are parked in m5: the application + they were planned for enrolls with certificates and documents itself. ## Related +- [Outbound federation](/quest/m3/cluster-routing/federation.md) - the relay node requires it; the other children do not +- [Rust byte streams](/quest/m1/stream/rust.md) - the Endpoint requires it; the other children do not - [P2P](/quest/m4/p2p/README.md) - internet peer sessions, which the m3 mesh does not use - [Mesh over ICE](/quest/m5/mesh-ice.md) - deferred direct paths between mesh nodes on different networks - [End-to-end encryption](/quest/m5/e2ee/README.md) - the AEAD the encryption level will reuse diff --git a/quest/m3/mesh/addr.md b/quest/m3/mesh/addr.md index 66cbbc4e98f..23b82172d6d 100644 --- a/quest/m3/mesh/addr.md +++ b/quest/m3/mesh/addr.md @@ -40,7 +40,7 @@ peers do. Amend that sentence of the docs with the mesh exception. Tests: namespaces with multicast off take the direct path; a forged hint (signed by another key) is never dialed; a colliding private address fails the pin and falls back to the relay; a peer that stops answering is backed off. -Update `doc/bin/relay/cluster.md` and the [guide](/quest/m3/mesh/guide.md) +Update `doc/bin/relay/cluster.md` and the [guide](/quest/m5/mesh-guide.md) if it exists. ## Required diff --git a/quest/m3/mesh/cert-credentials.md b/quest/m3/mesh/cert-credentials.md index 2b43a2a51be..d2d85838372 100644 --- a/quest/m3/mesh/cert-credentials.md +++ b/quest/m3/mesh/cert-credentials.md @@ -62,6 +62,6 @@ Added 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and rem - Chains are 2 to 4 KB in base64, close to common 8 KB URL limits; cache verified chains by fingerprint, and say in the PR how a long chain is carried. - The shared token corpus planned in [moq-dev/moq#8151](https://github.com/moq-dev/moq/pull/8151) includes this credential form. -## Required +## Related -- [Mesh relay node](/quest/m3/mesh/node.md) - the node that checks these credentials for its LAN peers +- [Mesh relay node](/quest/m3/mesh/node.md) - checks these credentials for its LAN peers, and only uses them, so this lands first diff --git a/quest/m3/mesh/endpoint.md b/quest/m3/mesh/endpoint.md index 7d359254e9b..1d9e719c8b0 100644 --- a/quest/m3/mesh/endpoint.md +++ b/quest/m3/mesh/endpoint.md @@ -25,7 +25,7 @@ Decided 2026-10-25 (maintainer): iroh-shaped, not drop-in compatible; tickets, `connect` dials with the caller's own endpoint ID, and `accept` reads the remote ID from the dial, which the socket's answer check proves and the grant pins (publish `.server/` and `.client/` only). -- Streams are [byte streams](/quest/m3/stream/rust.md) at the protection +- Streams are [byte streams](/quest/m1/stream/rust.md) at the protection level the service declares: identity reads `moq-net` directly, and authentication, the mesh default, reads through `moq_crypto::sign` against the `PublicKey` parsed from the peer's endpoint id, which names the endpoint @@ -34,7 +34,8 @@ Decided 2026-10-25 (maintainer): iroh-shaped, not drop-in compatible; tickets, - A mesh daemon (Kixel's, outside this repository, decided 2026-11-02) is built on this API, so a process either embeds an `Endpoint` or talks to a local [mesh relay node](/quest/m3/mesh/node.md). -- Write a short porting note into the [guide](/quest/m3/mesh/guide.md). +- Write a short porting note into `doc/` beside the Endpoint's own page; the + mesh [guide](/quest/m5/mesh-guide.md) is parked. Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): one socket endpoint per service, with a grant to publish on it under the member's own id, so `moq_net::socket` ships unchanged and needs no service field. @@ -51,26 +52,13 @@ member A put `.server/B` under A's own grant, so A could answer as B; the one path segment, validated when the `Endpoint` is built, so `*` in the grant covers it (`video/h264` would fall outside). -Member grants change shape, which breaks every enrolled device's paths: - -- `moq_auth::mesh`'s coordinator issues publish `mesh/node//**` and the - two service patterns (`rs/moq-auth/src/mesh/coordinator.rs`), with its - docs (`rs/moq-auth/src/mesh/mod.rs`), and contract tests; the PR's Impact section - replaces the `Pattern::subtree(&format!("mesh/{endpoint}"))` migration note - (`MIGRATE.md` is written at the release cut). -- `doc/bin/relay/auth.md`'s mesh enrollment section names the new grant. -- moq-crypto's and `@moq/crypto`'s sign examples and tests that use - `mesh//agent` paths move to `mesh/node//agent`, so the examples - match the grant. -- A hosted coordinator implementing the JSON contract issues the same - grant; the PR's Impact section says so. -- Rollout: an old client with a new grant, or a new client with an old - grant, can't publish. Open, for whoever starts this, with the maintainer: - recommended, the coordinator issues both the old and new patterns for one - release while clients upgrade and refresh their tokens, then drops the old - one, with the PR's Impact section giving that order; the old pattern keeps the - service-name collision open until it is dropped. +The grant change itself is its own quest, in m2, so it lands before the cut +publishes `moq_auth::mesh` (split 2026-11-12, maintainer, /quest-audit): +[mesh grants scope nodes and services apart](/quest/m2/mesh-grant-shape.md). +This quest builds the `Endpoint` on the new grant, and the dual-grant rollout +question it carried is gone, since no release ever issued the old shape. ## Required -- [Rust byte streams](/quest/m3/stream/rust.md) - the stream this API returns +- [Rust byte streams](/quest/m1/stream/rust.md) - the stream this API returns +- [Mesh grant shape](/quest/m2/mesh-grant-shape.md) - the grant that lets a member publish its service endpoints diff --git a/quest/m3/mesh/node.md b/quest/m3/mesh/node.md index 9b296129849..3247c82548b 100644 --- a/quest/m3/mesh/node.md +++ b/quest/m3/mesh/node.md @@ -28,6 +28,19 @@ CLI verbs, service install, and UI moved to Kixel. Carried over from it: embedding application supplies a trust store (decided 2026-11-07, maintainer). +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ the node +takes a credential handle, and the embedding application decides how it +enrolls (rejected: `moq_auth::mesh`'s coordinator and `Enrollment` learning +to issue certificates, which puts one consumer's contract in a public crate; +deferring). The handle yields what the node presents to the mesh's relays +and to LAN peers and says when to renew; `moq_auth::mesh::Client`'s token +enrollment is the implementation this repository ships and the `[mesh]` +config section uses, and a +[certificate credential](/quest/m3/mesh/cert-credentials.md) is another. The +Goal's "enrolls with an enrollment key" describes that default, not the only +path. Keep the handle a small trait or enum in moq-relay's mesh module, private +to the crate until an embedder outside this repository needs to implement it. + Tests in network namespaces: two relay nodes enroll through `moq auth`, take the LAN link when they share a segment and the relay path otherwise, refuse a LAN peer without an enrollment, and survive a relay restart. Update @@ -35,4 +48,4 @@ a LAN peer without an enrollment, and survive a relay restart. Update ## Required -- [Outbound federation](/quest/m2/cluster-routing/federation.md) - upstream links and LAN preference +- [Outbound federation](/quest/m3/cluster-routing/federation.md) - upstream links and LAN preference diff --git a/quest/m3/mesh/overlay.md b/quest/m3/mesh/overlay.md index f7ad341d6c5..02f41c9ed23 100644 --- a/quest/m3/mesh/overlay.md +++ b/quest/m3/mesh/overlay.md @@ -81,7 +81,7 @@ first. unset by default; the ring alone decides when it is unset. Over the cap, drop the longest fingers first. Small meshes come out close to full mesh on their own. -- Forwarding and route choice are [cluster routing](/quest/m2/cluster-routing/README.md)'s: +- Forwarding and route choice are [cluster routing](/quest/m3/cluster-routing/README.md)'s: overlay links join relay nodes as LAN-cost links, and relay uplinks stay `upstream` links costed above them. Nothing routes outside the shared priority/cost rules. @@ -94,7 +94,7 @@ follows the ring (and never exceeds the cap when set); every pair connects without crossing the relay; killing a node heals the overlay; splitting the segment sends cross-half traffic through the relay and healing restores the direct path. Report the link counts and hop counts the test measures. -Update `doc/bin/relay/cluster.md` and the [guide](/quest/m3/mesh/guide.md) if it exists. +Update `doc/bin/relay/cluster.md` and the [guide](/quest/m5/mesh-guide.md) if it exists. ## Required diff --git a/quest/m3/mf-encode-adapter.md b/quest/m3/mf-encode-adapter.md index 7a009cd2db6..69a4752efc3 100644 --- a/quest/m3/mf-encode-adapter.md +++ b/quest/m3/mf-encode-adapter.md @@ -11,16 +11,16 @@ for a readback encoder because Media Foundation picked another GPU. ## Plan Builds on `moq_video::render::export` from -[moq-dev/moq#6805](https://github.com/moq-dev/moq/pull/6805); start after it -merges. That PR's exporter probes each candidate encoder with a frame +[moq-dev/moq#6805](https://github.com/moq-dev/moq/pull/6805) (merged). That PR's exporter probes each candidate encoder with a frame exported from the drawing device, and passes over a Media Foundation encoder that sits on a different adapter. Enumerate with `MFTEnum2` instead of -`MFTEnumEx`, passing an `IMFAttributes` store whose `MFT_ENUM_ADAPTER_LUID` +`MFTEnumEx` (`rs/moq-video/src/encode/backend/mediafoundation.rs`, about +line 758 as of a86433043), passing an `IMFAttributes` store whose `MFT_ENUM_ADAPTER_LUID` is the texture's adapter, so the probe finds a hardware encoder that reads it. An explicitly named encoder that still cannot read the device fails at setup, as it does today. -Decided 2026-10-26 (maintainer): m2, beside Direct3D11 import. It changes no -public API, so it does not need to land before the cut, and it needs a +Decided 2026-10-26 (maintainer): it changes no public API, so it does not +need to land before the cut, and it needs a Windows machine with two GPUs to prove: an ignored hardware test plus an entry in [Hardware validation](/quest/m5/video-hardware.md) if CI cannot run it. @@ -29,5 +29,5 @@ Public API: none. Wire: none. ## Related -- [Direct3D11 render import](/quest/m3/render-d3d11.md) - the other Windows half of keeping rendered pictures on the GPU +- [Direct3D11 render import](/quest/m4/render-d3d11.md) - the other Windows half of keeping rendered pictures on the GPU - [Studio](https://github.com/moq-dev/studio) - moq-dev/studio, the app whose program this encodes diff --git a/quest/m3/mobile-capture-android.md b/quest/m3/mobile-capture-android.md index feec51d4a18..f58f0e2c418 100644 --- a/quest/m3/mobile-capture-android.md +++ b/quest/m3/mobile-capture-android.md @@ -21,7 +21,7 @@ so the mechanism exists. Capture sets the catalog `rotation` from the sensor and display orientation, like browser capture, so a phone held in portrait plays upright. -Deferred to m2; mobile capture does not gate the next breaking cut. +Mobile capture does not gate the next breaking cut. ## Related diff --git a/quest/m3/mobile-capture-ios.md b/quest/m3/mobile-capture-ios.md index be09d915a6d..b21f51523c5 100644 --- a/quest/m3/mobile-capture-ios.md +++ b/quest/m3/mobile-capture-ios.md @@ -30,7 +30,7 @@ same everywhere. Capture sets the catalog `rotation` from the device orientation, like browser capture, so a phone held in portrait plays upright. -Deferred to m2; mobile capture does not gate the next breaking cut. +Mobile capture does not gate the next breaking cut. ## Related diff --git a/quest/m3/mobile-completion.md b/quest/m3/mobile-completion.md index 833eda26943..c1552a243e4 100644 --- a/quest/m3/mobile-completion.md +++ b/quest/m3/mobile-completion.md @@ -25,8 +25,11 @@ Promoted from m2 on 2026-10-09 by maintainer priority, with both capture quests. Deferred back to m2 in #5427 (`8f0dfa760`) with those capture quests; mobile completion does not gate the next breaking cut. +Corrected 2026-11-12 (/quest-audit): Dart on iOS left `Required`. The mobile +SDK is the Swift and Kotlin bindings, so the Dart asset proof does not gate +it; that quest moved to m4. + ## Required -- [Dart on iOS](/quest/m3/dart-ios.md) - the Dart iOS asset proof - [iOS capture](/quest/m3/mobile-capture-ios.md) - deliver the selected iOS capture path - [Android capture](/quest/m3/mobile-capture-android.md) - deliver the selected Android capture and codec path diff --git a/quest/m3/mp4-repair-lock.md b/quest/m3/mp4-repair-lock.md index 09dd2babab7..dcd6210d162 100644 --- a/quest/m3/mp4-repair-lock.md +++ b/quest/m3/mp4-repair-lock.md @@ -8,7 +8,7 @@ Found by [moq-dev/moq#7560](https://github.com/moq-dev/moq/pull/7560), which adds `moq repair mp4`: it can't tell a running `export mp4` still writes the file, and repairing one corrupts it. #7560 merged 2026-11-02. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2. An advisory lock taken by the recorder and checked by repair, as #7560 suggested. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): an advisory lock taken by the recorder and checked by repair, as #7560 suggested. - The CLI owns the lock, not moq-mux: `repair` in moq-mux takes `impl Read + Seek` and stays file-agnostic. Use an exclusive `flock`-style lock on the open file (`std::fs::File::try_lock` is stable in Rust 1.89), held for the recording's life and released by the OS if the recorder dies, so a crashed recording stays repairable. - Supported or refused (`AGENTS.md`), on both sides. The recorder fails at startup if it can't take the lock, a filesystem without lock support included, since its recording could never be repaired safely. `repair` takes the same lock non-blocking before scanning and refuses if it can't: held means a recorder still has the file open, and any other lock error (no lock support) refuses with that error rather than repairing unlocked. diff --git a/quest/m3/mux-source-cut.md b/quest/m3/mux-source-cut.md index b6630d419e3..c9627fe8548 100644 --- a/quest/m3/mux-source-cut.md +++ b/quest/m3/mux-source-cut.md @@ -11,7 +11,7 @@ the broadcast over, exactly as one built through `Server::broadcaster` does. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from [moq-dev/moq#7399](https://github.com/moq-dev/moq/pull/7399): a separate S -quest in m2 for the `moq_mux::Source` API, alongside reporting stalled +quest for the `moq_mux::Source` API, alongside reporting stalled segments. [moq-dev/moq#7399](https://github.com/moq-dev/moq/pull/7399) (merged) detects takeover in the server's follower (`Followed::start` diff --git a/quest/m3/nixpkgs-relay.md b/quest/m3/nixpkgs-relay.md index 0e0062a6c82..b4307b3de07 100644 --- a/quest/m3/nixpkgs-relay.md +++ b/quest/m3/nixpkgs-relay.md @@ -36,8 +36,8 @@ maintainer opens the PR rather than an agent. Two stages, each its own PR here: -1. Prepare. Fork NixOS/nixpkgs to kixelated/nixpkgs (none exists as of - 2026-11-05) and push a branch with two commits: kixelated's +1. Prepare. Use the existing kixelated/nixpkgs fork (it exists as of + 2026-11-12, with no moq-relay branch yet) and push a branch with two commits: kixelated's `maintainers/maintainer-list.nix` entry (GitHub `kixelated`, id 432854), then `pkgs/by-name/mo/moq-relay/package.nix`. Both commits carry `Assisted-by:`. Package from the latest `moq-relay-v*` tag with diff --git a/quest/m3/obs-decode-latest.md b/quest/m3/obs-decode-latest.md index 2d5af02d934..17012e74d86 100644 --- a/quest/m3/obs-decode-latest.md +++ b/quest/m3/obs-decode-latest.md @@ -4,8 +4,7 @@ The OBS MoQ source's audio decoder starts at the newest cached group instead of replaying the track's backlog, so a reconnect or rendition change plays -live audio at once. The Go wrapper no longer re-exports the raw ffi decoder -output records. +live audio at once. ## Plan @@ -19,17 +18,15 @@ Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 - ✅ Switch to `DecodeStart::Latest`. Reason: the source is a live player, so replaying the cache only adds latency. -- `cpp/obs/src/moq-source.cpp:1068` (as of be39188000): set `output.start` to +- `cpp/obs/src/moq-source.cpp:1081` (as of a86433043): set `output.start` to `moq::DecodeStart::kLatest` on the `moq::AudioDecoderOutput`. Video reads raw frames through `MediaTrackConsumer` and its own decoder, so it is unaffected. -- Delete the `AudioDecoderOutput` and `VideoDecoderOutput` aliases in - `go/wrapper/types.go:15,81`: no wrapper API accepts them - (`audio.DecoderOutput`/`video.DecoderOutput` are the real inputs). Their - migration note is gathered by the - [migration audit](/quest/m3/release/migrate-audit.md). - If cheap, an e2e check that an audio decoder with `Latest` starts at the newest group, beside [moq-dev/moq#7288](https://github.com/moq-dev/moq/pull/7288)'s `video_decode_start`. -Public API: Go `wrapper.AudioDecoderOutput`/`VideoDecoderOutput` removed -(breaking). Wire: none. +The Go half of the original quest, deleting the wrapper's +`AudioDecoderOutput`/`VideoDecoderOutput` aliases, landed in +[moq-dev/moq#8303](https://github.com/moq-dev/moq/pull/8303). + +Public API: none. Wire: none. diff --git a/quest/m3/obs-moq-video/README.md b/quest/m3/obs-moq-video/README.md index f3e64f00b06..b14f8908637 100644 --- a/quest/m3/obs-moq-video/README.md +++ b/quest/m3/obs-moq-video/README.md @@ -16,8 +16,6 @@ Publishing remains opt-in, with one **Use MoQ encoders** choice for video and au The quests separate portable decoding, audio, and publishing so each can land and be validated independently. moq-ffi's decoded frames own their surface and convert to CPU pixels only on request (#4094); a surface decode exposes the platform surface as a borrowed view, which each platform GPU quest extends to its own surface type. -Decided in the 2026-09-30 audit: the Windows and Linux GPU decode paths and the macOS and Windows GPU encoder inputs moved to m2 (listed under Related), because each needs physical-hardware proof, and Linux GPU input is already m2/m3. - ## Related - [Video hardware validation](/quest/m5/video-hardware.md) - physical hardware evidence is required for each claimed GPU path diff --git a/quest/m3/obs-moq-video/adapter.md b/quest/m3/obs-moq-video/adapter.md index fd0a756ccbc..ebe3bbefebc 100644 --- a/quest/m3/obs-moq-video/adapter.md +++ b/quest/m3/obs-moq-video/adapter.md @@ -13,5 +13,3 @@ One opt-in Use MoQ encoders choice publishes OBS video and audio through moq-vid - Establish bounded submission/packet queues, explicit raw-frame drop behavior, thread confinement, cancellation, late completion, device loss, resize and color metadata. Never block OBS's graphics thread on network backpressure. The CPU path is a correctness/fallback baseline; platform quests establish accelerated input. - The audio side is the internal `moq_opus` encoder, whose settings are `bitrate` (kbps) and `preset` (`low_latency`, `balanced`, `quality`). Use MoQ encoders creates it in place of `ffmpeg_opus`; it is the only audio choice the combined mode offers until AAC encoding lands. - Test rejection, saturation, drain, stop during encode, delayed completion, and repeated start/stop. Validate real decoded pixels and audio continuity, matched timestamps, preset reporting, and frame-to-packet latency. Validate the new binding docs, feature combinations, package dependencies and native plugin linking. - -Changed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): the Go `*DecoderOutput` alias deletion lands before the cut in the api-shape bindings batch ([moq-dev/moq#8303](https://github.com/moq-dev/moq/pull/8303)); `DecodeStart::Latest` stays here. diff --git a/quest/m3/one-port/README.md b/quest/m3/one-port/README.md index e229b202850..3918a191b5c 100644 --- a/quest/m3/one-port/README.md +++ b/quest/m3/one-port/README.md @@ -96,4 +96,4 @@ so the shared port never answers public Binding requests. ## Related - [One port on the io_uring workers](/quest/m3/uring-demux.md) - the io_uring workers host the UDP demux -- [Stream transports](/quest/m4/stream-transports/README.md) - the io_uring workers that would host the TCP acceptor later +- [Stream transports](/quest/m9/stream-transports/README.md) - the io_uring workers that would host the TCP acceptor later diff --git a/quest/m3/opus-surround-export.md b/quest/m3/opus-surround-export.md index c570307ad56..1395337517c 100644 --- a/quest/m3/opus-surround-export.md +++ b/quest/m3/opus-surround-export.md @@ -25,7 +25,7 @@ and [moq-dev/moq#7308](https://github.com/moq-dev/moq/pull/7308). a plain table, so it imports as plain channels or fails as an invalid table. [moq-dev/moq#7336](https://github.com/moq-dev/moq/pull/7336) refuses family 3 in fMP4 only. Refuse it in `parse` itself (recommended, since no consumer can render a demixing matrix today), or - carry it as `ambisonic` if [VR and ambisonic fMP4](/quest/m4/mux-vr-ambisonic-fmp4.md) + carry it as `ambisonic` if [VR and ambisonic fMP4](/quest/m5/mux-vr-ambisonic-fmp4.md) wants it. Cover both with unit tests beside [moq-dev/moq#7336](https://github.com/moq-dev/moq/pull/7336)'s diff --git a/quest/m2/probe/README.md b/quest/m3/probe/README.md similarity index 88% rename from quest/m2/probe/README.md rename to quest/m3/probe/README.md index e4d84843236..391dc267bf6 100644 --- a/quest/m2/probe/README.md +++ b/quest/m3/probe/README.md @@ -12,6 +12,8 @@ by default; an application can override or disable them. ## Plan +Moved to m3 2026-11-12 (maintainer, /quest-audit): nothing left in it breaks, its one child waits on Chrome, and its own shaped-uplink test needs the parked padding probes, so the cut does not wait on it. + The estimators stop learning once a connection is app-limited. moq-quic's BBR3 only raises `max_bw` from an app-limited sample above its current value; while app-limited it ages its window every 3 s, so the estimate falls to what @@ -45,10 +47,11 @@ publisher-local padding handle: resends in-flight data during BBR3's Startup and one ProbeBW_UP per probe wait ([moq-dev/moq#6635](https://github.com/moq-dev/moq/pull/6635)). A browser over WebTransport can't reach its controller, so a lowest-priority padding stream on an ALR probe - schedule does ([padding](/quest/m3/probe-padding.md)). Parked 2026-11-06 + schedule does ([padding](/quest/m5/probe-padding.md)). Parked 2026-11-06 (maintainer, /quest-unblock): the maintainer doubts padding is the right - approach, so padding and the [JS rung disable](/quest/m3/probe-rung-disable-js.md), - which recovers through it, moved to m3 for a rethink. + approach, so padding and the [JS rung disable](/quest/m5/probe-rung-disable-js.md), + which recovers through it, left this epic for a rethink (both are in m5 + since 2026-11-12). Decided with the maintainer on 2026-10-20: @@ -69,7 +72,7 @@ Decided with the maintainer on 2026-10-20: native rung recovery needs, and viewer handles, viewer grants, the peer's PROBE target, and `Increase` came back here as viewer targets, beside their consumers, and landed in [moq-dev/moq#7090](https://github.com/moq-dev/moq/pull/7090). - The `js/net` send target moved to [padding](/quest/m3/probe-padding.md), its + The `js/net` send target moved to [padding](/quest/m5/probe-padding.md), its only sender. Decided with the maintainer on 2026-11-03, in review of @@ -99,4 +102,4 @@ This README's own work, once the children land: ## Related -- [Realtime uplink congestion control](/quest/m3/quic-realtime/README.md) - a delay-based alternative to BBR3 on native uplinks; it must implement native probing to ship (decided 2026-10-25) +- [Realtime uplink congestion control](/quest/m4/quic-realtime.md) - a delay-based alternative to BBR3 on native uplinks; it must implement native probing to ship (decided 2026-10-25) diff --git a/quest/m2/probe/browser-send-estimate.md b/quest/m3/probe/browser-send-estimate.md similarity index 94% rename from quest/m2/probe/browser-send-estimate.md rename to quest/m3/probe/browser-send-estimate.md index 8df877a25b8..99a80567665 100644 --- a/quest/m2/probe/browser-send-estimate.md +++ b/quest/m3/probe/browser-send-estimate.md @@ -8,6 +8,8 @@ The WebSocket fallback is out of scope: it has no browser estimate to read, so i ## Plan +Moved to m3 2026-11-12 (maintainer, /quest-audit), with the probe epic: no API or wire change, and it waits on Chrome, so the cut does not wait on it. + Found by [moq-dev/moq#7524](https://github.com/moq-dev/moq/pull/7524): `js/net`'s `Allocator` only reads `WebTransport.getStats()`'s `estimatedSendRate`, which shipping Chromium lacks (it is behind `--enable-blink-features=WebTransportStats`). qmux reports no counters, and WebRTC's `getStats` throws. Found 2026-11-05 ([moq-dev/moq#8016](https://github.com/moq-dev/moq/pull/8016)), replacing the 2026-11-01 plan to feed the peer's PROBE estimate to the send allocator: @@ -25,7 +27,7 @@ Reason: one rule decided at setup fixes the RTT, the polling, and the counters. What remains once the waits clear: -- In stable Chrome (no flag) and Firefox, verify through a shaped link that a publisher's bitrate follows a capacity drop. Reuse the [probe epic](/quest/m2/probe/README.md)'s shaped-uplink harness if it has landed. +- In stable Chrome (no flag) and Firefox, verify through a shaped link that a publisher's bitrate follows a capacity drop. Reuse the [probe epic](/quest/m3/probe/README.md)'s shaped-uplink harness if it has landed. - The WebSocket fallback has no estimate whatever the browsers ship, so its publishers still hold their rate, as decided below. Decided 2026-11-07 (maintainer, /quest-unblock): Safari publishers and the WebSocket fallback get no estimate before WebKit reports one. @@ -44,5 +46,5 @@ Public API: none expected. Wire: none. ## Related -- [Padding probes](/quest/m3/probe-padding.md) - raises the estimate once one exists +- [Padding probes](/quest/m5/probe-padding.md) - raises the estimate once one exists - [Media stats](/quest/m3/stats/README.md) - its JS stats read PROBE for the downlink diff --git a/quest/m3/publish-cancel-classify.md b/quest/m3/publish-cancel-classify.md index 1a2e4e7d98d..c5177027874 100644 --- a/quest/m3/publish-cancel-classify.md +++ b/quest/m3/publish-cancel-classify.md @@ -9,20 +9,23 @@ a clean probe cancel at the `Control` boundary. ## Plan -Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 +Planned 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7350](https://github.com/moq-dev/moq/pull/7350): m2, as -recommended. +[moq-dev/moq#7350](https://github.com/moq-dev/moq/pull/7350). -`is_cancel` (`rs/moq-net/src/lite/publisher.rs:1559` as of be39188000, -"TODO better classify WebTransport errors") matches all `Transport(_)` and -`IdleTimeout`. -[moq-dev/moq#7350](https://github.com/moq-dev/moq/pull/7350) left it alone and wrote its own three-variant match for probe ends in -`lite/publisher.rs` and `lite/subscriber.rs`. +`is_cancel` (`rs/moq-net/src/lite/publisher.rs:1662`, "TODO better classify +WebTransport errors") counts every `Error::Transport(_)` and `IdleTimeout` as +a cancel. The classification it asks for already happens at the transport +boundary: `Error::from_transport` (`rs/moq-net/src/error.rs`) reads a stream +reset's or session close's code, and otherwise the transport's `CloseKind`, +so a cancel code becomes `StreamError::Cancel` or `SessionError::Cancel`, a +local close becomes `SessionError::Cancel`, and only an unclassified failure +is left as `Error::Transport`. [moq-dev/moq#7350](https://github.com/moq-dev/moq/pull/7350)'s probe ends in `lite/publisher.rs` and +`lite/subscriber.rs` already match only the three cancel variants. -- Classify `Transport` errors by what the transport says (a stream reset or - stop with the cancel code, or a closed session, is a cancel; anything else - is a fault), then use one helper at every log site, including [moq-dev/moq#7350](https://github.com/moq-dev/moq/pull/7350)'s. +- Drop `Error::Transport(_)` from `is_cancel`, decide whether `IdleTimeout` + stays (a peer that went silent, not a fault of ours), and use that one + helper at both probe-end sites too, rather than separate matches. - Test: a subscriber cancelling a probe at the `Control` boundary logs no warning (or the classifier returns cancel for it), and an injected transport error is not a cancel. diff --git a/quest/m3/qos-publisher-timeliness.md b/quest/m3/qos-publisher-timeliness.md index 2ae1a252d45..7068e545d18 100644 --- a/quest/m3/qos-publisher-timeliness.md +++ b/quest/m3/qos-publisher-timeliness.md @@ -11,10 +11,14 @@ cooperation. ## Plan This is the ingress mirror of the egress `lag` histogram ([viewer lag -histogram](/quest/m3/qos/lag-histogram.md), which lands first so this reuses -its histogram type, decided 2026-10-08) and needs no acknowledgments: the relay is the receiver, and a frame is measured +histogram](/quest/m3/qos/lag-histogram.md)) and needs no acknowledgments: the relay is the receiver, and a frame is measured when its last byte arrives, because a partial frame is not useful. +Changed 2026-11-12 (/quest-audit): this quest no longer waits on the lag +histogram, which waits on the ACK hook. It defines the cumulative histogram +type and its bucket edges, and the lag histogram reuses them. This replaces +the 2026-10-08 order, where the egress histogram landed first. + Frame timestamps are relative and jittered with no epoch, so measure each track against itself. Per track compute each frame's raw offset `arrival_instant - timestamp` (both converted to one unit) and keep the @@ -30,7 +34,7 @@ Document the window length beside the bucket edges. Aggregate as a byte-weighted cumulative histogram of drift on the ingress (`Role::Subscriber`) side of the stats totals and prefix tracks (`/prefix.json`) rather than the per-path rows (decided in the 2026-10-06 audit), with the same -bucket edges and the same monotonic contract as the egress `lag` histogram. Beside it keep two cumulative +bucket edges and the same monotonic contract the egress `lag` histogram will share. Beside it keep two cumulative counters: `timestamp_regressions`, frames whose timestamp is below the previous group's newest timestamp, and `stalls`, gaps between complete frames longer than a documented threshold. Compare only across group boundaries for @@ -52,6 +56,6 @@ stall and no drift for frames it never sent; a regression across groups is counted while intra-group reordering is not; a track without a timescale is excluded. -## Required +## Related -- [Viewer lag histogram](/quest/m3/qos/lag-histogram.md) - the egress histogram whose type and edges this reuses +- [Viewer lag histogram](/quest/m3/qos/lag-histogram.md) - the egress histogram that reuses the type and edges this defines diff --git a/quest/m3/qos/README.md b/quest/m3/qos/README.md index 87688438394..50ad639b510 100644 --- a/quest/m3/qos/README.md +++ b/quest/m3/qos/README.md @@ -45,13 +45,18 @@ histogram and dropped counters across. Decided (2026-10-08): the egress histogram is its own quest, [viewer lag histogram](/quest/m3/qos/lag-histogram.md), since every other -child reads it. The line's PR, #4133, closed without landing (2026-10-09). +child reads it. Since 2026-11-12 the histogram type and bucket edges are +defined by [publisher timeliness](/quest/m3/qos-publisher-timeliness.md), +which needs no acknowledgments, and the lag histogram reuses them. The line's PR, #4133, closed without landing (2026-10-09). Its port, #6791, closed unmerged on 2026-10-26 to wait for the ACK hook, since `lag` and `dropped` are measured per acknowledged frame; its branch is the histogram quest's starting point. Decided (2026-09-28): the line's moq-stats changes break the published -crate. Client stats left the line (2026-09-29) +crate. No longer so since +[moq-dev/moq#8266](https://github.com/moq-dev/moq/pull/8266): `stats::Registry::report` +already takes the drain's instant, so the +[lag histogram](/quest/m3/qos/lag-histogram.md) adds fields only. Client stats left the line (2026-09-29) when media stats moved out of moq-stats onto hang tracks, which are additive on `main`. diff --git a/quest/m3/quic-ack-hook.md b/quest/m3/quic-ack-hook.md index fcd916b2315..6348c6116af 100644 --- a/quest/m3/quic-ack-hook.md +++ b/quest/m3/quic-ack-hook.md @@ -111,3 +111,14 @@ close. Decided in the 2026-09-30 audit: moved to m2 with its consumers, the [viewer lag histogram](/quest/m3/qos/lag-histogram.md) and a delivery-delay histogram, whose plan was deleted on 2026-11-09. + +Split 2026-11-12 (maintainer, /quest-audit before the cut): the trait shape +(`Offset`, `offset()`, and a defaulted `poll_acked`) lands in m2 as +[its own quest](/quest/m2/api-ack-hook-trait.md), because +`poll::SendStream` is public and unsealed and the cut would freeze it. This +quest keeps the driver half: moq-quic's `poll_acked` leaves `#[cfg(test)]`, +and the moq-tokio and moq-uring adapters implement the hook. + +## Required + +- [The send-stream trait has its acknowledgment hook](/quest/m2/api-ack-hook-trait.md) - the trait methods the drivers implement diff --git a/quest/m3/quic-emsgsize-classify.md b/quest/m3/quic-emsgsize-classify.md index 0755eb38e13..41990ec4e82 100644 --- a/quest/m3/quic-emsgsize-classify.md +++ b/quest/m3/quic-emsgsize-classify.md @@ -9,11 +9,8 @@ logged and counted, so it can't be dropped silently again. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: - -- m0 -- m1 -- ✅ m2 +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs): ✅ later work, which is m3 since the 2026-11-05 re-cut (rejected: m0; +m1). From [moq-dev/moq#7649](https://github.com/moq-dev/moq/pull/7649), which capped a GSO train at what one UDP send carries after full trains above a diff --git a/quest/m3/quic-path-return.md b/quest/m3/quic-path-return.md index 73db2f9a2cd..7744e6ac7cc 100644 --- a/quest/m3/quic-path-return.md +++ b/quest/m3/quic-path-return.md @@ -54,5 +54,5 @@ Public API: none. Wire: none. ## Related -- [Buffered exports pick their group order by measurement](/quest/m1/export-order.md) - the send-order half of the same outage +- [Buffered exports pick their group order by measurement](/quest/m3/export-order.md) - the send-order half of the same outage - [Per-stream and per-datagram deadlines](/quest/m4/quic-deadline.md) - faster recovery below a deadline diff --git a/quest/m1/receive-windows-evicted.md b/quest/m3/receive-windows-evicted.md similarity index 94% rename from quest/m1/receive-windows-evicted.md rename to quest/m3/receive-windows-evicted.md index 2e2b292cbe1..aadefaebfa9 100644 --- a/quest/m1/receive-windows-evicted.md +++ b/quest/m3/receive-windows-evicted.md @@ -17,6 +17,11 @@ alongside throughput and latency, instead of failing. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + Moved to m1 2026-11-04 with the maintainer, after [moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838) landed the fitting cells: what remains waits on diff --git a/quest/m3/release/c-publish.md b/quest/m3/release/c-publish.md index 40557e496ac..44504faca90 100644 --- a/quest/m3/release/c-publish.md +++ b/quest/m3/release/c-publish.md @@ -11,10 +11,13 @@ archive instead of the hand-written one. The generated C line proves the same cases with the built archive before publication. This quest proves registry/release delivery after publication -and records the artifact versions and results. The hand-written `moq-c` 0.7 -ships at the next cut (decided 2026-10-25), so the generated 0.8 supersedes a -published package; its source stays until -[c-retire](/quest/m3/release/c-retire.md). +and records the artifact versions and results. The hand-written `moq-c` +ships at the next cut (decided 2026-10-25), so the generated package +supersedes a published one; its source stays until +[c-retire](/quest/m3/release/c-retire.md). On `main` the two read 0.7.0 +(`rs/moq-c/Cargo.toml`) and 0.8.0 (`c/moq/VERSION`), but the cut's +`just version` sets both to the shared minor (`sh/version.py`), so neither +number survives it and both `moq-c` lines carry the same version. The smoke client moves here rather than with [C consumers](/quest/m3/c/consumers.md) because it checks a published archive, diff --git a/quest/m3/release/c-retire.md b/quest/m3/release/c-retire.md index 2315eecb94a..2fd2f0d70dc 100644 --- a/quest/m3/release/c-retire.md +++ b/quest/m3/release/c-retire.md @@ -10,8 +10,10 @@ served it are deleted once the generated `moq-c` publishes. The generated - #4288 renamed libmoq to `moq-c` (`rs/moq-c`); this quest deletes that hand-written crate. -- Decided 2026-10-25 (maintainer): the hand-written `moq-c` 0.7 ships at the - m1 cut, so it is a published package when the generated 0.8 replaces it. +- Decided 2026-10-25 (maintainer): the hand-written `moq-c` ships at the + next cut, so it is a published package when the generated one replaces it + (0.7.0 and 0.8.0 on `main`; `just version` sets both to the cut's shared + minor). Its `MIGRATE.md` history carries into the generated package. This replaces the 2026-10-20 premise that it never published. - `rs/moq-c/native-libs/` serves the generated package too, and diff --git a/quest/m3/release/cut.md b/quest/m3/release/cut.md index a594442e042..01a875f77f9 100644 --- a/quest/m3/release/cut.md +++ b/quest/m3/release/cut.md @@ -19,6 +19,8 @@ from moq-dev/web-transport, while `js/qmux` is at 0.3.3. Publishing 0.3.3 fails, and `workspace:^` would resolve `@moq/net`'s dependency to web-transport's 0.3.4, a different API. `js/qmux` must publish above 0.3.4; the maintainer picks the version (found 2026-11-07 by the API shape audit). +Recommended 2026-11-12 (/quest-audit): 0.4.0, so a consumer of +web-transport's `^0.3` never resolves to the new API. Before publishing, confirm moq-dev/web-transport's release config no longer publishes `@moq/qmux` or `qmux` (the freeze step of [qmux-js-retire](/quest/m3/release/qmux-js-retire.md) and @@ -101,16 +103,24 @@ From the 2026-10-25 pre-cut run: - Write the moq-cli and moq-gst `MIGRATE.md` entries from the Impact section of [moq-dev/moq#6522](https://github.com/moq-dev/moq/pull/6522); that rename list exists nowhere else. -- Raise each wrapper's moq-ffi version floor (for example Kotlin's - `MOQ_FFI_RANGE`) to the release that ships the `implementation` field from - [moq-dev/moq#6775](https://github.com/moq-dev/moq/pull/6775). From [moq-dev/moq#7702](https://github.com/moq-dev/moq/pull/7702), which merged moq-e2ee and moq-sign into moq-crypto (decided 2026-11-02, no yank): -- Nothing to publish for moq-e2ee: its final code-free release, 0.0.22, - already points at moq-crypto, and `rs/moq-e2ee` is gone. Don't merge a - `release` branch release-plz proposal that bumps it. +- `rs/moq-e2ee` is gone from `main`, but crates.io's newest moq-e2ee is a + full 0.0.25 crate published 2026-11-11 from `release`, not the code-free + 0.0.22 pointer (corrected 2026-11-12, /quest-audit). Don't merge a + `release` branch release-plz proposal that bumps it again; + [retired names](/quest/m3/release/retired-names.md) publishes the pointer + above 0.0.25, with the other names this cut stops publishing. + +Open as of 2026-11-13 (maintainer, /quest-plan): where the next cycle's cuts +live. The steps above place them in m0's release epic and have the wire cut +require m2 and m3, but no quest may require one in a later milestone, and +[m4](/quest/m4/README.md) and [m5](/quest/m5/README.md) now require this cut +until it lands, a gate each cycle should re-create. [Cut +placement](/quest/m0/cut-placement.md) settles both before this cut +re-creates its successor. Cut gate template (moved here 2026-11-09 from the gate's own Plan; see the decision below). The gate is an epic in m3 for the checks the cut needs that @@ -163,9 +173,12 @@ with the suites above. ## Required +- [Cut placement](/quest/m0/cut-placement.md) - where this cut re-creates its successor, and the gate it re-creates - [Breaking release readiness](/quest/m2/README.md) - code, protocol and package proofs - [Migration audit](/quest/m3/release/migrate-audit.md) - every break on the cut PR has its MIGRATE.md entry - [API shape review](/quest/m3/release/api-review.md) - every new or changed public API was reviewed for a better shape, and approved changes landed - [Feature review](/quest/m3/release/feature-review.md) - every major work line since the last release was judged end to end, and its follow-ups landed - [Cargo feature review](/quest/m3/release/feature-flags.md) - every Cargo feature was judged against the feature rules, and approved changes landed +- [Interop and Video pass on main](/quest/m0/interop-green/README.md) - the cut PR's release gate runs `interop.yml`, which failed on `main` every day through 2026-11-12 +- [Nightly passes on main](/quest/m0/nightly/README.md) - the release gate runs `features.yml`, red in the 2026-11-12 Nightly - [Claim new npm names](/quest/m3/release/npm-first-publish.md) - trusted publishing can publish every new `@moq/*` package diff --git a/quest/m3/release/dart-publish.md b/quest/m3/release/dart-publish.md index 4c7ff3f86ee..0dd1ffd8af9 100644 --- a/quest/m3/release/dart-publish.md +++ b/quest/m3/release/dart-publish.md @@ -24,7 +24,7 @@ can proceed independently; cutting the tag cannot. push events), and a rerun of that tag's [release-dart run](https://github.com/moq-dev/moq/actions/runs/43107848443) published 0.1.1. The next tag shows whether it publishes with no rerun. -- Raise `dart/moq/pubspec.yaml`'s `moq_ffi` floor to the release the wrapper now needs (the media namespace and codec types), since `^0.4.3` resolves bindings without them. +- The cut's `just version` raises `dart/moq/pubspec.yaml`'s `moq_ffi` floor from `^0.4.3`, which resolves bindings without the media namespace and codec types (`FFI_REQUIREMENTS` in `sh/version.py`). Check the published `pubspec.yaml` carries the cut's floor. - Cut the next `moq-dart-v*` release with the binding breaks and confirm the workflow publishes it. - Verify the published `moq_ffi` resolves its native asset from a clean machine with no monorepo checkout, since that download path is the one CI @@ -39,4 +39,4 @@ Public API: none. Wire: none. ## Related -- [Dart on iOS](/quest/m3/dart-ios.md) - pub.dev already tags `moq_ffi` for iOS, which nobody has run +- [Dart on iOS](/quest/m4/dart-ios.md) - pub.dev already tags `moq_ffi` for iOS, which nobody has run diff --git a/quest/m3/release/demo-sockets-deploy.md b/quest/m3/release/demo-sockets-deploy.md index ece3bb9f396..5ec3007fe97 100644 --- a/quest/m3/release/demo-sockets-deploy.md +++ b/quest/m3/release/demo-sockets-deploy.md @@ -4,8 +4,8 @@ moq.dev's hosted MoQ Boy and CARLA demos work with the socket versions the cut publishes: the relays accept lite-07, the auth server and the emulator's -and worker's tokens carry the socket grants, and the site's `` drops -its viewer prefix. +and worker's tokens carry the socket grants, and the site builds `` +from `demo/boy` without a viewer prefix. ## Plan @@ -15,11 +15,20 @@ chosen: an m2 cut gate; deploying now in m0. Main's moq-boy ([moq-dev/moq#8008](https://github.com/moq-dev/moq/pull/8008)) and CARLA ([moq-dev/moq#8011](https://github.com/moq-dev/moq/pull/8011)) talk -over sockets, which need lite-07 (or moq-transport) on every hop. Until -moq.dev's relays accept it, the released `` can't connect at all. +over sockets, which need lite-07 (or moq-transport) on every hop, so the +sockets `` can't connect until moq.dev's relays accept it. The grants below are the server and client socket roles that `demo/relay/prod.toml` documents. +The cut publishes no MoQ Boy package +([moq-dev/moq#8446](https://github.com/moq-dev/moq/pull/8446), 2026-11-10): +the `moq-boy` crate is `publish = false` and the element lives in the private +`demo/boy` (`@moq/demo-boy`), which has no `exports`. The site builds the +element from its `moq` submodule, and on 2026-11-12 it still lists the +`moq/js/moq-boy` workspace and imports `@moq/boy/element`, both gone at the +cut. The demo hosts keep building the emulator from the flake's `moq-boy` +package. + ## Waiting - Maintainer: after the cut publishes, deploy moq.dev: @@ -31,7 +40,10 @@ The grants below are the server and client socket roles that - The emulator's token publishes `demo/boy/.server/` and subscribes `demo/boy/.client/**`; the CARLA worker's token publishes its trucks' `.server/` endpoints and subscribes `demo/carla/.client/**`. - - The site's `src/components/boy.tsx` drops `prefix-viewer`. + - The site's `moq` submodule moves to the cut, its `moq/js/moq-boy` + workspace becomes `moq/demo/boy`, and `src/components/boy.tsx` imports the + element from there (`src/element.tsx`) and drops `prefix-viewer`, which + the element no longer reads. Then open both demos in a browser and drive a truck and a game; an agent can drive Chrome for that check. diff --git a/quest/m3/release/feature-flags.md b/quest/m3/release/feature-flags.md index 368ab8c8230..b6762e013e6 100644 --- a/quest/m3/release/feature-flags.md +++ b/quest/m3/release/feature-flags.md @@ -21,8 +21,12 @@ Record what this run learned (inventory method, findings, what to check first) in the cut's PR notes or a follow-up quest, so the next run starts from it. Only `/quest-delete` ends the recurrence. -Starts once m2 lands, so the features it reviews are the ones that ship; the -cut waits for it. Feature changes break downstream builds, which is why this +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ this run +starts now, on the features `main` has today, with a delta pass after m2 +(rejected: one pass after m2 lands, which puts every approved change behind +the routes chain instead of beside it). The first pass's PR rewrites this +quest to the delta pass: features m2 added or changed since, with +[m2](/quest/m2/README.md) under `Required`. The cut waits for the delta pass. Feature changes break downstream builds, which is why this runs before the cut rather than with the post-release audits. Like the [API shape review](/quest/m3/release/api-review.md), this run changes no code (decided 2026-10-26): it proposes each change with a recommendation, asks the @@ -58,7 +62,3 @@ What to check first (seeded 2026-10-26): Related: the [cost audit](/quest/m3/audit-cost.md) owns per-crate build cost and which dependencies exist. - -## Required - -- [Breaking release readiness](/quest/m2/README.md) - the release's features are final before they are reviewed diff --git a/quest/m3/release/feature-review.md b/quest/m3/release/feature-review.md index ec499473508..4d121fae2e5 100644 --- a/quest/m3/release/feature-review.md +++ b/quest/m3/release/feature-review.md @@ -18,7 +18,12 @@ Record what this run learned (inventory method, findings, what to check first) in the cut's PR notes or a follow-up quest, so the next run starts from it. Only `/quest-delete` ends the recurrence. -Starts once m2 lands; the cut waits for it. Inventory the work lines from the +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ this run +starts now, on what has landed, with a delta pass after m2 (rejected: one pass +after m2 lands, which puts every finding behind the routes chain instead of +beside it). The first pass's PR rewrites this quest to the delta pass: what +m2 landed since, with [m2](/quest/m2/README.md) under `Required`. The cut +waits for the delta pass. Inventory the work lines from the epics and quests deleted since the last release and from `git log` by area. Judge each one adversarially (AGENTS.md) and challenge its premise, not only its execution: is the feature worth supporting at all, and would a simpler @@ -33,7 +38,3 @@ separate TS audit): everything since the 2026-10-11 TS audit, judged against the contracts in the [MPEG-TS guide](/doc/bin/ts.md). Question each TS capability's premise the way the move to distinct importer epochs dropped active-active publishers. - -## Required - -- [Breaking release readiness](/quest/m2/README.md) - the release's code is final before it is reviewed diff --git a/quest/m3/release/migrate-audit.md b/quest/m3/release/migrate-audit.md index 80edb54d887..c13503af9a8 100644 --- a/quest/m3/release/migrate-audit.md +++ b/quest/m3/release/migrate-audit.md @@ -23,10 +23,7 @@ m1 quest on 2026-10-22. for its migration note, since squash commits drop the PR body. Write each note into its package's guide. The API-diff report below catches breaks a PR left unmarked; a behavior break with no `!` stays with PR review. -- Correct Unreleased entries that a later PR overturned. Known: `js/net/MIGRATE.md` - still describes `request(path, { announced: true })`, which - [moq-dev/moq#7173](https://github.com/moq-dev/moq/pull/7173) replaced with - `follow(path)` (decided 2026-10-29). Also known, from +- Correct Unreleased entries that a later PR overturned. Known, from [moq-dev/moq#7036](https://github.com/moq-dev/moq/pull/7036)'s rename of `used` to `wanted` (Decided 2026-10-30 (maintainer, /quest-plan of the 2026-10-30 /quest-complete follow-ups)): `rs/moq-json/MIGRATE.md` should say `producer.demand().wanted().is_some()`, not @@ -79,9 +76,6 @@ m1 quest on 2026-10-22. - moq-webrtc: `rs/moq-webrtc/MIGRATE.md` is still titled "Migrating moq-rtc" and documents `wire::Source`, which [moq-dev/moq#6454](https://github.com/moq-dev/moq/pull/6454) removed. - - moq-boy: `rs/moq-boy/MIGRATE.md` documents `.socket/{name}` and - `.socket/viewer-*` paths and grants; the code serves `.server/{name}` - and `.client/{id}`. - moq-cli: `rtmp` is listed under `--max-delay` but its flag is `--delay`; the `--connect-once` removal and the `MOQ_LISTEN_TLS_*` to `MOQ_HLS_TLS_*` rename are missing; the `export archive` entry reads as a diff --git a/quest/m3/release/npm-first-publish.md b/quest/m3/release/npm-first-publish.md index 8e1895c63e4..d95dc2abdcc 100644 --- a/quest/m3/release/npm-first-publish.md +++ b/quest/m3/release/npm-first-publish.md @@ -2,12 +2,12 @@ ## Goal -`@moq/clock`, `@moq/libav`, `@moq/ui`, and `@moq/video` exist on npm, each with trusted +`@moq/clock`, `@moq/libav`, `@moq/stats`, `@moq/ui`, and `@moq/video` exist on npm, each with trusted publishing attached to `moq-dev/moq` and the `npm` Environment. The cut's release workflow can then publish their real versions. A maintainer does this by hand. Delete this quest once the checks pass: -- `npm view @moq/` resolves for all four, and the package settings show +- `npm view @moq/` resolves for all five, and the package settings show the `moq-dev/moq` trusted publisher with Environment `npm`. ## Plan @@ -20,7 +20,11 @@ publish fails on these names and still publishes dependents that can't install. [moq-dev/moq#6243](https://github.com/moq-dev/moq/pull/6243) raised it for `@moq/libav` alone. `@moq/ui`, which `@moq/watch` and `@moq/publish` keep external, joined the list on 2026-11-06, when the final API audit found -it still returned 404. +it still returned 404. `@moq/stats` joined on 2026-11-12 (/quest-audit): +[moq-dev/moq#8502](https://github.com/moq-dev/moq/pull/8502) added `js/stats` +at 0.1.0 with a release script, and the name returns 404. Nothing published +depends on it, so a miss fails only its own publish. All five still returned +404 on 2026-11-12. Decided 2026-10-22 (maintainer): publish a minimal `0.0.0` placeholder (README only) per name to claim it, then attach the trusted publisher. The diff --git a/quest/m3/release/retired-names.md b/quest/m3/release/retired-names.md new file mode 100644 index 00000000000..8d40bd1a832 --- /dev/null +++ b/quest/m3/release/retired-names.md @@ -0,0 +1,40 @@ +# [S] Retired package names point at their replacements + +## Goal + +Every crate and npm package that stops publishing at this cut tells its +users where to go. A renamed one gets a final pointer release; a removed one +is deprecated on its registry with a message naming the replacement. + +## Plan + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ pointer +stubs for renames, deprecation for removals (rejected: deprecation only; +doing nothing). + +- Pointer stub, as `libmoq` has: `moq-rtc` (renamed to `moq-webrtc` in + [moq-dev/moq#8122](https://github.com/moq-dev/moq/pull/8122)) and + `moq-e2ee`. crates.io's newest `moq-e2ee` is a full 0.0.25 crate published + 2026-11-11, not the stub an earlier plan assumed (the stub was 0.0.22), so + the pointer publishes above 0.0.25. +- Deprecate with a message: `moq-room` and `moq-native` + ([moq-dev/moq#8450](https://github.com/moq-dev/moq/pull/8450)), + `moq-binary`, and `moq-boy` with `@moq/boy`, which + [moq-dev/moq#8446](https://github.com/moq-dev/moq/pull/8446) stopped + publishing; that PR's "Release cut" step also removes their trusted + publishers. +- First confirm the list against the registries: compare the crates and + packages `release` publishes with what the cut's `main` publishes, and add + any name this list misses. +- crates.io has no deprecation flag: a final patch release whose README and + description name the replacement is the notice, and `cargo yank` is not + used, since it breaks existing lockfile-free builds. npm uses + `npm deprecate`. + +## Waiting + +- Maintainer: publish the stubs and run the deprecations once the cut's packages are on their registries; an agent has no crates.io or npm publish rights. Done when `cargo info moq-rtc` and `npm view @moq/boy deprecated` each name the replacement. + +## Required + +- [Release cut](/quest/m3/release/cut.md) - the replacements are published before the old names point at them diff --git a/quest/m3/route-keep-next-hop.md b/quest/m3/route-keep-next-hop.md index ed51fc632e1..d95dd202a06 100644 --- a/quest/m3/route-keep-next-hop.md +++ b/quest/m3/route-keep-next-hop.md @@ -22,7 +22,7 @@ What it costs today, from #8487's measurements: a change of hops alone is control messages over real sessions, and on a 32-relay full mesh one link coming up costs 592 route messages against 251 before the hops. -[Cluster routing](/quest/m2/cluster-routing/README.md) already decided that +[Cluster routing](/quest/m3/cluster-routing/README.md) already decided that "a new connection alone does not outrank a different equal-cost route", so this implements a settled rule. Keep the deterministic path hash among equal routes for a request's first hop: that choice is per request and is not this diff --git a/quest/m1/route-trace-validation.md b/quest/m3/route-trace-validation.md similarity index 84% rename from quest/m1/route-trace-validation.md rename to quest/m3/route-trace-validation.md index 3f03f1f46e0..5f2ec74099d 100644 --- a/quest/m1/route-trace-validation.md +++ b/quest/m3/route-trace-validation.md @@ -9,12 +9,17 @@ resync the model. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + Decided 2026-11-10 (maintainer, /quest-plan), as the strongest guard against the route model drifting from the code ([moq-dev/moq#8447](https://github.com/moq-dev/moq/pull/8447)). Decided 2026-11-10 (maintainer, a later /quest-plan the same day): -- ✅ m1 (rejected: staying in m4). The mesh configs beside it, the +- ✅ m1 (rejected: staying in m4; superseded 2026-11-13: m3, above). The mesh configs beside it, the inductive invariant (dropped 2026-11-13: no checker can step the model), and the [cache-aware routing proof](/quest/m2/cache-routing-proof.md) all build on `RouteHandoff.tla`, so drift there costs the most. diff --git a/quest/m1/routing-guide.md b/quest/m3/routing-guide.md similarity index 71% rename from quest/m1/routing-guide.md rename to quest/m3/routing-guide.md index ecadd8aa997..deaabd3b01f 100644 --- a/quest/m1/routing-guide.md +++ b/quest/m3/routing-guide.md @@ -7,7 +7,13 @@ outbound local relays, including priorities, costs, and continuity. ## Plan -The 2026-10-15 release audit keeps federation guidance in m1. Deferred P2P +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + +The 2026-10-15 release audit kept federation guidance in m1, until the move +above. Deferred P2P owns the later application-selected peer setup and extends this concept page. Every feature PR still fixes docs it makes stale. @@ -29,4 +35,4 @@ Public API/wire: documentation only, reflecting completed implementations. ## Required -- [Outbound federation](/quest/m2/cluster-routing/federation.md) - tested operator configuration +- [Outbound federation](/quest/m3/cluster-routing/federation.md) - tested operator configuration diff --git a/quest/m3/session-outcome-records.md b/quest/m3/session-outcome-records.md index a9d8bf429b7..fd9d2637c85 100644 --- a/quest/m3/session-outcome-records.md +++ b/quest/m3/session-outcome-records.md @@ -39,22 +39,24 @@ records): (count per kind over the interval) instead, and the cumulative counters stay exact. Measure the bytes a refusal flood costs with and without the bound. - Refusal records go through `moq_net::stats::Handle::refuse`, so a gateway's - refusals appear the same way as native ones. It takes only `(label, reason)` - today, and the relay's `Refusals::record`/`record_auth` - (`rs/moq-relay/src/refusals.rs`) keep only `(tier, root)`, so the record's - path and transport have no way in. Make the breaking change: `refuse` takes - one struct carrying the label, kind, requested path, and transport, and the - relay passes them from the `moq_auth::Request` it holds at every refusal - site. A field is optional only where a real caller lacks it. + refusals appear the same way as native ones. It takes a `#[non_exhaustive]` + `stats::Refusal`, which carries only the label and kind today, and the + relay's `Refusals::record`/`record_auth` (`rs/moq-relay/src/refusals.rs`) + pass only a kind or the refusal's tier and root, so the record's path and + transport have no way in. Add the requested path and transport as fields of + `Refusal`, and have the relay pass them from the `moq_auth::Request` it + holds at every refusal site. A field is optional only where a real caller + lacks it. - Abnormal-end records are emitted in `moq-net` where `sessions_failed` is bumped today (`SessionInner::drop`, with the kind `Session::end` set), not through `refuse`. The stats session therefore learns its path, transport, - and start time when created, so `Handle::session` takes them too (the same - breaking change; the relay knows `request.transport()` at admission). + and start time when created, so they become fields of the + `#[non_exhaustive]` `stats::SessionInfo` that `Handle::session` takes (the + relay knows `request.transport()` at admission). - Rust and JS readers parse the track. Tests: a refused session and an abnormal end each produce exactly one record under the right label; a flood past the rate produces summaries whose counts match the counters; an admitted session that ends normally produces none. -Changed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): [moq-dev/moq#8266](https://github.com/moq-dev/moq/pull/8266) makes `stats::Handle::{refuse, session}` take `#[non_exhaustive]` structs before the cut, so the records add fields and are not breaking. +Changed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): [moq-dev/moq#8266](https://github.com/moq-dev/moq/pull/8266), since merged, made `stats::Handle::{refuse, session}` take `#[non_exhaustive]` structs, so the records add fields and are not breaking. diff --git a/quest/m3/stats/preflight.md b/quest/m3/stats/preflight.md index 36533dc358c..9bf772b415d 100644 --- a/quest/m3/stats/preflight.md +++ b/quest/m3/stats/preflight.md @@ -54,4 +54,4 @@ Public API: a preflight check and its report type, and a CLI sink. Wire: none. ## Related - [Publisher timeliness](/quest/m3/qos-publisher-timeliness.md) - the relay's view of the same timestamp lateness and monotonicity -- [Intra-refresh](/quest/m3/intra-refresh/README.md) - a stream without IDR keyframes, which the sync point checks must not fail +- [Intra-refresh](/quest/m4/intra-refresh/README.md) - a stream without IDR keyframes, which the sync point checks must not fail diff --git a/quest/m3/stream/js.md b/quest/m3/stream/js.md deleted file mode 100644 index c920db58ed4..00000000000 --- a/quest/m3/stream/js.md +++ /dev/null @@ -1,16 +0,0 @@ -# [S] JS byte streams - -## Goal - -`@moq/net` dials and accepts the same byte streams as Rust, exposed as a -WHATWG `ReadableStream`/`WritableStream` pair, and interoperates with the Rust -side in both directions. - -## Plan - -Mirror the [Rust surface](/quest/m3/stream/rust.md) on `@moq/net`'s socket. Add -the Rust/JS pair to `just test interop`. - -## Required - -- [Rust byte streams](/quest/m3/stream/rust.md) - the reference surface diff --git a/quest/m2/subscribe-ranges/js-fills.md b/quest/m3/subscribe-fills-js.md similarity index 79% rename from quest/m2/subscribe-ranges/js-fills.md rename to quest/m3/subscribe-fills-js.md index 05e49f54cef..7f9e65cf1cb 100644 --- a/quest/m2/subscribe-ranges/js-fills.md +++ b/quest/m3/subscribe-fills-js.md @@ -11,11 +11,12 @@ serves its group fills. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. -- m0 -- ✅ m1, in this epic -- m2 +Moved to m1 2026-11-12 (maintainer, /quest-audit): additive, so the cut does not wait on it; `Fill` only gains fields and the wire is unchanged. It left the [Subscribe ranges](/quest/m2/subscribe-ranges/README.md) epic, where it was planned on 2026-11-03. The maintainer chose to plan these separately from [moq-dev/moq#7696](https://github.com/moq-dev/moq/pull/7696), which landed @@ -61,11 +62,11 @@ through a JS relay between Rust peers (`just test relay-fill`). Public API: `Fill` gains `time` and a frame start, matching Rust's `Fill::time` and `Fill::start`. Wire: none. -Decided 2026-11-05 (maintainer, /quest-plan): this waits for Drop set ([moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476)), since its head-fill -half rewrites the same JS fill and miss code as -[moq-dev/moq#7670](https://github.com/moq-dev/moq/pull/7670) and Drop set, -which also redefine the stale-drop judgment it relies on. Not chosen: -splitting the `Fill.time` lookup off to start now. +Drop set, which this waited on (2026-11-05, /quest-plan) because its +head-fill half rewrites the same JS fill and miss code as +[moq-dev/moq#7670](https://github.com/moq-dev/moq/pull/7670) and relies on +the stale-drop judgment Drop set redefined, finished in +[moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476) (2026-11-12). Decided 2026-11-05 (maintainer, /quest-audit): ✅ this also waits for Lite-07 FETCH is single-group only ([moq-dev/moq#8616](https://github.com/moq-dev/moq/pull/8616)), diff --git a/quest/m2/time-addressing/guide.md b/quest/m3/time-addressing-guide.md similarity index 71% rename from quest/m2/time-addressing/guide.md rename to quest/m3/time-addressing-guide.md index 42a8e074f95..bb625cd1024 100644 --- a/quest/m2/time-addressing/guide.md +++ b/quest/m3/time-addressing-guide.md @@ -11,6 +11,13 @@ authors can follow it without reading the drafts. ## Plan +Moved from m1 to m3 2026-11-13 (maintainer, /quest-plan): it waits on m2 +work, and no quest requires one in a later milestone, since a blocker ranks +by its own milestone. ✅ m3 (rejected: m2, where the cut would wait on it; +pulling the m2 blocker into m1). It ranks right after m2, when it can start. + +Moved to m1 2026-11-12 (maintainer, /quest-audit): a docs page breaks nothing, so the cut does not wait on it. + Decided 2026-10-25 (maintainer): a separate page, not only inline doc updates. Write it against the design [the model](/quest/m2/time-addressing/model.md) and [lite-07](/quest/m2/time-addressing/lite07.md) quests record, including diff --git a/quest/m3/tls-link-alpn.md b/quest/m3/tls-link-alpn.md index 5b17d84ce29..570d7aa3e6d 100644 --- a/quest/m3/tls-link-alpn.md +++ b/quest/m3/tls-link-alpn.md @@ -8,18 +8,16 @@ so admission and logs see the protocol on every transport. ## Plan -Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 +Planned 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7358](https://github.com/moq-dev/moq/pull/7358): m2, as -recommended. [moq-dev/moq#7358](https://github.com/moq-dev/moq/pull/7358) (merged) added client certificates and SNI on `tls://` +[moq-dev/moq#7358](https://github.com/moq-dev/moq/pull/7358) (merged), which added client certificates and SNI on `tls://` and left `link.alpn` `None` there as out of scope; build on it. [moq-dev/moq#7362](https://github.com/moq-dev/moq/pull/7362) (merged) already sets -the ALPN for `tls://` sessions on the relay's shared TCP port -(`rs/moq-tokio/src/tcp.rs:305-319` at be39188000, -`Incoming::tcp(..).with_alpn(alpn)`). A `tls://` listener's own accept loop -(`spawn_tcp_loop`, `rs/moq-tokio/src/server.rs:1542`, building it at -`:1567`) still builds `Incoming::tcp` without it. +the ALPN for `tls://` sessions on the relay's shared TCP port (`accept` in +`rs/moq-tokio/src/tcp.rs`, `Incoming::tcp(..).with_alpn(alpn)`). A `tls://` +listener's own accept loop (`spawn_tcp_loop` in `rs/moq-tokio/src/server.rs`) +still builds `Incoming::tcp` without it. Read the ALPN from the completed rustls handshake there the same way, sharing the read with `tcp.rs` rather than copying it. Plain `tcp://` stays `None`. Test with a `tls://` session asserting the diff --git a/quest/m3/token-auth-followups.md b/quest/m3/token-auth-followups.md deleted file mode 100644 index 03a828afa2b..00000000000 --- a/quest/m3/token-auth-followups.md +++ /dev/null @@ -1,32 +0,0 @@ -# [S] PUBLISH token renewal is consistent across pause and forward, and the multi-token cost is profiled - -## Goal - -On a moq-transport session, a REQUEST_UPDATE carrying an AUTHORIZATION -TOKEN on a PUBLISH gets the same answer whether the PUBLISH is paused or -forwarding: a renewal is accepted in both states, or refused in both. A -serve-level test covers a token on a paused PUBLISH offer and on a -forwarding one. The remaining multi-token gap in -`cargo bench -p moq-net --bench auth -- auth/request_tokens` is profiled, -and its cause either fixed or recorded in this quest's PR with numbers. - -## Plan - -Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from [moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211) (open), -which finishes [request tokens](https://github.com/moq-dev/moq/pull/7211). - -- Renewal: after [moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211) a paused PUBLISH accepts a token, but a forwarding - one refuses it ("AUTHORIZATION TOKEN on a locally initiated PUBLISH", - in [moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211)'s `apply_subscription_update`, which is at - `rs/moq-net/src/ietf/publisher.rs:2266` on main at be39188000), so a - forwarding PUBLISH can't be renewed in place. It fails - closed, so this is a gap, not a hole. Recommended: accept the renewal - while forwarding, held to the same checks [moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211) applies to a paused - PUBLISH's token. Refusing it while - paused too is the fallback if acceptance needs a new grant path. -- Profile: single-token rounds match main; multi-token rounds are 1.5 to - 2.2x slower and grow with tokens times requests. The suspected cost is - each announce polling its guard on the shared verdict. Profile with - `samply` or `perf` on the bench before changing anything. - -Public API: none. Wire: none; behavior only. diff --git a/quest/m2/unbounded-groups/README.md b/quest/m3/unbounded-groups/README.md similarity index 96% rename from quest/m2/unbounded-groups/README.md rename to quest/m3/unbounded-groups/README.md index 19ad9e2c424..624f72ccda7 100644 --- a/quest/m2/unbounded-groups/README.md +++ b/quest/m3/unbounded-groups/README.md @@ -83,9 +83,8 @@ Decided 2026-10-28 (maintainer, after an audit of the plan and HLS and DASH cut a stream into finite segments joiners start at. This epic is `tee()` with a bounded buffer, paced on the fastest branch. -Decided 2026-10-23 (maintainer): moved from m2 to m1, holding the next -breaking cut, so a long-lived log such as a moq-json `Stream` has no record -ceiling in that release. Making `Stream` roll groups instead was rejected: +Moved to m3 2026-11-12 (maintainer, /quest-audit): front eviction and flights already removed the record ceiling for a long-lived log, which is why it held the cut since 2026-10-23, and the children left are additive. +Making `Stream` roll groups instead was rejected (2026-10-23): its readers reject a second group (`Rolled`), and a `Window` that never pops is already the rolling log. Pointing `Stream`'s docs at `Window` in the meantime was declined. @@ -135,7 +134,7 @@ Decided 2026-10-16 (maintainer interview): (d854bf98b1), ahead of Front eviction, including `log::Writer::evict` advancing the base (`rs/kio/src/log.rs`). A ring buffer was rejected: an uncapped group has no size to give it. Numbers are in #6782 and the - [kio contention epic](/quest/m4/kio-channel-contention/README.md). + [kio contention epic](/quest/m9/kio-channel-contention/README.md). Accepted 2026-11-03 (maintainer), from the lock profiles in [moq-dev/moq#7742](https://github.com/moq-dev/moq/pull/7742): no lock-contention regression attributable to #6782. The pre-epoch run on @@ -153,7 +152,7 @@ Decided 2026-10-16 (maintainer interview): - The byte offset goes on the current lite wip version (lite-07-wip today); the lite-07 cut does not wait on it. Decided 2026-10-27 (maintainer): wire quests target whichever wip is current when they land, so - [Byte holes](/quest/m2/unbounded-groups/byte-holes.md) waits only on Front + [Byte holes](/quest/m3/unbounded-groups/byte-holes.md) waits only on Front eviction and Flights. If lite-07 is cut first, lite-07 peers start flights only at frame boundaries, over lite-07's transport frame offsets, which allow a fill to start without the shed prefix (decided 2026-10-25, diff --git a/quest/m2/unbounded-groups/byte-holes.md b/quest/m3/unbounded-groups/byte-holes.md similarity index 90% rename from quest/m2/unbounded-groups/byte-holes.md rename to quest/m3/unbounded-groups/byte-holes.md index aa8a641b840..b9e3e82daaa 100644 --- a/quest/m2/unbounded-groups/byte-holes.md +++ b/quest/m3/unbounded-groups/byte-holes.md @@ -10,7 +10,7 @@ are shed like front frames, and a flight ## Plan -See the [epic](/quest/m2/unbounded-groups/README.md) for the decisions. +See the [epic](/quest/m3/unbounded-groups/README.md) for the decisions. Decided 2026-10-27 (maintainer): the byte offset targets the current lite wip version (lite-07-wip today); the lite-07 cut does not wait on it. This follows [lite-07's freeze policy](/quest/m2/lite07-finalize.md): wire quests target @@ -18,7 +18,10 @@ whichever wip is current when they land, and its Required list is not a closed freeze set. If lite-07 is cut first, the offset goes on the next wip and lite-07 peers start flights only at frame boundaries, like older lite versions. Decided 2026-10-28 (maintainer): shed frames are served again only -by flights, so this waits on Flights and stays in m2 with its epic. +by flights, so this waited on Flights, which landed in +[moq-dev/moq#7944](https://github.com/moq-dev/moq/pull/7944). + +Moved to m3 2026-11-12 (maintainer, /quest-audit), with its epic: the cut does not wait on it, so if lite-07 is cut first the offset goes on the next wip, as above. Decided 2026-11-05 (maintainer, /quest-audit): ✅ the byte offset rides FETCH's frame start plus the group header, not a SUBSCRIBE range end. @@ -60,7 +63,7 @@ This waits for that quest. kept its segments internal. JS already has its chunked read (`nextFrame` and `readChunk`, from [moq-dev/moq#7411](https://github.com/moq-dev/moq/pull/7411)); its wire - half is in [JS byte holes](/quest/m2/unbounded-groups/js-byte-holes.md). + half is in [JS byte holes](/quest/m3/unbounded-groups/js-byte-holes.md). - Test a frame several times the 32 MB tail through a relay chain: a late joiner reads it whole through a flight, and memory stays near 32 MB per copy. diff --git a/quest/m2/unbounded-groups/compressed-stream.md b/quest/m3/unbounded-groups/compressed-stream.md similarity index 92% rename from quest/m2/unbounded-groups/compressed-stream.md rename to quest/m3/unbounded-groups/compressed-stream.md index edc30c9cef8..952de343169 100644 --- a/quest/m2/unbounded-groups/compressed-stream.md +++ b/quest/m3/unbounded-groups/compressed-stream.md @@ -25,7 +25,7 @@ Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 compressed size on the wire, so we don't need to uncompress to learn it." So no wire change: keep the declared compressed size, drop only the guard, and bound decompression while streaming. -- Milestone: ✅ m1. Not chosen: m2. +- Moved to m3 2026-11-12 (maintainer, /quest-audit), with its epic: additive, with no wire change, so the cut does not wait on it. - JS ([moq-dev/moq#7411](https://github.com/moq-dev/moq/pull/7411) merged): drop the guard and the `Encoder.bound` preflight in `js/net/src/group.ts` (`writeFrame`, `createFrame`, binding compression) and JSON's mirrored `MAX_FRAME_BYTES`. Inflate a received slice in chunks @@ -39,7 +39,7 @@ Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 `write_frames` (`group.rs:1719`, `:1798`) check only the plain payload against 32 MB, so an incompressible frame near the cap encodes to a slice every receiver refuses. With no compressed cap, there is nothing to - mismatch; [Byte holes](/quest/m2/unbounded-groups/byte-holes.md) removes + mismatch; [Byte holes](/quest/m3/unbounded-groups/byte-holes.md) removes the plain cap. - The lite draft's FRAME text ("a receiver MUST check it against its limits before inflating") stays; add that a receiver may inflate incrementally and @@ -55,4 +55,4 @@ JSON preflight). Wire: none. ## Required -- [Byte holes](/quest/m2/unbounded-groups/byte-holes.md) - Rust frames past 32 MB at all +- [Byte holes](/quest/m3/unbounded-groups/byte-holes.md) - Rust frames past 32 MB at all diff --git a/quest/m2/unbounded-groups/js-byte-holes.md b/quest/m3/unbounded-groups/js-byte-holes.md similarity index 86% rename from quest/m2/unbounded-groups/js-byte-holes.md rename to quest/m3/unbounded-groups/js-byte-holes.md index e1433da9c9e..377e2dd079d 100644 --- a/quest/m2/unbounded-groups/js-byte-holes.md +++ b/quest/m3/unbounded-groups/js-byte-holes.md @@ -8,9 +8,11 @@ from the byte it reached instead of failing. ## Plan +Moved to m3 2026-11-12 (maintainer, /quest-audit), with its epic: it mirrors Byte holes, which the cut does not wait on. + Decided 2026-10-27 (maintainer): split out of the JS unbounded groups quest. This half mirrors -[Byte holes](/quest/m2/unbounded-groups/byte-holes.md). +[Byte holes](/quest/m3/unbounded-groups/byte-holes.md). Decided 2026-10-31 (maintainer, on [moq-dev/moq#7411](https://github.com/moq-dev/moq/pull/7411)): chunked frame @@ -31,7 +33,7 @@ moq-transport. What is left here is the wire: - Test a frame several times the cache window through a Rust relay, read as chunks, plus Rust/JS interop of a mid-frame refill. Run `just test interop --all`. -- If [JS-only @moq/net from Rust](/quest/m5/rs2ts-js.md) is adopted first, +- If [the Rust model in the browser](/quest/m5/rs2ts-js.md) is adopted first, this shrinks to exposing its byte-offset recovery. Decided 2026-11-05 (maintainer, /quest-audit): ✅ the byte offset rides @@ -48,4 +50,4 @@ Byte holes defines it. ## Required -- [Byte holes](/quest/m2/unbounded-groups/byte-holes.md) - the Rust contract and the lite wire this mirrors +- [Byte holes](/quest/m3/unbounded-groups/byte-holes.md) - the Rust contract and the lite wire this mirrors diff --git a/quest/m3/uring-demux.md b/quest/m3/uring-demux.md index 1a15e72d73f..45c86759606 100644 --- a/quest/m3/uring-demux.md +++ b/quest/m3/uring-demux.md @@ -55,4 +55,4 @@ off on the ring, from shard steering. ## Related -- [Stream transports](/quest/m4/stream-transports/README.md) - the TCP half of moving the relay onto the ring +- [Stream transports](/quest/m9/stream-transports/README.md) - the TCP half of moving the relay onto the ring diff --git a/quest/m4/vaapi-render-tests.md b/quest/m3/vaapi-render-tests.md similarity index 98% rename from quest/m4/vaapi-render-tests.md rename to quest/m3/vaapi-render-tests.md index 91d682178f8..7031370f099 100644 --- a/quest/m4/vaapi-render-tests.md +++ b/quest/m3/vaapi-render-tests.md @@ -70,4 +70,4 @@ which covers only NVIDIA. ## Related - [GPU CI](/quest/m3/gpu-ci.md) - the NVIDIA half of `just rs gpu` -- [VA-API encodes an external Vulkan image](/quest/m3/vaapi-vulkan-import.md) - the encode side of the same VA-API surfaces +- [VA-API encodes an external Vulkan image](/quest/m5/vaapi-vulkan-import.md) - the encode side of the same VA-API surfaces diff --git a/quest/m3/vendor-sync/README.md b/quest/m3/vendor-sync/README.md index 9590a11e9e0..1991b8ba598 100644 --- a/quest/m3/vendor-sync/README.md +++ b/quest/m3/vendor-sync/README.md @@ -39,8 +39,7 @@ This README's own work, once every child has merged: vendored commit. Code whose upstream is gone or frozen (moq-tokio's `quic::web`, from moq-dev/noq) is listed in the README it lives in and gets no child. -2. Re-create the questline at `/quest/m3/vendor-sync/`, linked from - `/quest/m3/README.md`'s `Required`. Every child keeps its Goal and carries +2. Re-create the questline at `/quest/m3/vendor-sync/`. Every child keeps its Goal and carries forward its last checked commit and what this run learned (what to check first, what was ruled out). Each child requires [the release cut](/quest/m3/release/cut.md), which removes itself from @@ -58,6 +57,11 @@ The 2026-11-05 scan found two new upstreams, each now a child: WebRTC commit, so their first runs establish one. moq-congestion's Cubic came from quinn and stays with the quinn child. +The 2026-11-12 audit found moq-webrtc's vendored str0m, dimpl, `is`, and +sctp-proto (vendored 2026-11-07, after that scan) with no child, so they got +[one](/quest/m3/vendor-sync/webrtc-stack.md) starting from the revisions +`rs/moq-webrtc/src/vendor/mod.rs` records. + Out of scope: forks moq no longer consumes (moq-dev/quiche, moq-dev/noq, kixelated/cros-codecs, kixelated/openh264-rs, kixelated/fflate, and kixelated/goog_cc, now in-tree as moq-congestion's `gcc`), the diff --git a/quest/m3/vendor-sync/quinn.md b/quest/m3/vendor-sync/quinn.md index 1dcc8589540..299b2b2c023 100644 --- a/quest/m3/vendor-sync/quinn.md +++ b/quest/m3/vendor-sync/quinn.md @@ -35,7 +35,7 @@ stays in moq-quic's README. Ruled out: quinn's batched `read_datagrams` ([`2697de1e`](https://github.com/quinn-rs/quinn/commit/2697de1e1314cb9f98502b7323acd5bdd836b582)) -waits on the [driver lock profile](/quest/m4/perf/quic-lock-contention.md). +waits on the [driver lock profile](/quest/m9/perf/quic-lock-contention.md). Public API: none expected; a ported fix that changes one says so. Wire: QUIC behavior follows upstream's fixes. diff --git a/quest/m3/vendor-sync/scream.md b/quest/m3/vendor-sync/scream.md index a3f2e46361b..981796bcb39 100644 --- a/quest/m3/vendor-sync/scream.md +++ b/quest/m3/vendor-sync/scream.md @@ -24,7 +24,7 @@ reference's `master` head then was Only the queue-delay target path is ported. The L4S path, competing-flow compensation, and the draft's media-queue terms were left out on purpose (see -the [realtime uplink epic](/quest/m3/quic-realtime/README.md)), so a fix +the [realtime uplink epic](/quest/m4/quic-realtime.md)), so a fix confined to them does not apply. Keep the PR's two recorded deviations (delay backoff counts as congestion; `scl_t` taken after the reduction) unless upstream settles them differently. SCReAMv2 is a bake-off candidate; if the diff --git a/quest/m3/vendor-sync/webrtc-stack.md b/quest/m3/vendor-sync/webrtc-stack.md new file mode 100644 index 00000000000..c46497e08b9 --- /dev/null +++ b/quest/m3/vendor-sync/webrtc-stack.md @@ -0,0 +1,44 @@ +# [S] moq-webrtc's vendored stack has every upstream fix it needs + +## Goal + +moq-webrtc's vendored str0m, `str0m_proto`, `is`, dimpl, and sctp-proto +carry every fix their upstreams +([algesten/str0m](https://github.com/algesten/str0m), +[algesten/dimpl](https://github.com/algesten/dimpl), and +[algesten/sctp-proto](https://github.com/algesten/sctp-proto)) have landed +since the last check, each ported with its test or recorded as not applying, +and the `Upstream` section of `rs/moq-webrtc/src/vendor/mod.rs` names the +newest upstream commit checked for each. + +## Plan + +Recurring child of the [vendor sync](/quest/m3/vendor-sync/README.md), which +says how each run works and recurs. New in the 2026-11-12 audit: the stack was +vendored on 2026-11-07 in +[moq-dev/moq#8163](https://github.com/moq-dev/moq/pull/8163), after the +2026-11-05 scan. + +Start from the revisions the module doc records: str0m `main` at +[`dc317173`](https://github.com/algesten/str0m/commit/dc317173b16fc6ed86633e6b08b229c9195deb87) +(2026-10-26; `str0m_proto`, `is`, and the test-only `netem` come from the same +repository), dimpl 0.9.0 at +[`c17b3b8`](https://github.com/algesten/dimpl/commit/c17b3b8b971b6eb2ecf7284f5dc0ecb86680ac4d), +and sctp-proto 0.10.7 as published on crates.io. Check each repository's +security advisories too: the stack parses untrusted SRTP, RTCP, DTLS, STUN, +and SCTP input from browsers, and cargo-deny cannot see an advisory against +vendored code. + +The vendored code was trimmed to what MoQ uses (the +[WebRTC epic](/quest/m3/webrtc/README.md)'s "What goes" list: GCC bandwidth +estimation, the leaky pacer, H.266, RED, DTMF, G.7xx and comfort noise, VLA +and simulcast layers, RTCP app feedback, stats, and DTLS 1.3), so a fix +confined to removed code does not apply. Deleting an API with no consumer +beats porting a fix to it. + +Out of scope: `str0m_rind` and dimpl's `crypto::rind`, ported from +moq-dev/rind, which moq maintains itself. + +## Required + +- [Cut and publish the prepared breaking release](/quest/m3/release/cut.md) - the next cycle starts after the cut diff --git a/quest/m3/vendor-sync/webrtc.md b/quest/m3/vendor-sync/webrtc.md index 8ae3b6ea1b1..9f3c8c56211 100644 --- a/quest/m3/vendor-sync/webrtc.md +++ b/quest/m3/vendor-sync/webrtc.md @@ -23,7 +23,7 @@ plus the `api/transport`, `api/units`, `modules/pacing`, and apply the changes file by file. Keep the recorded departures (the ALR detector's per-send rounding, the 1 Mbit/s start, restarted probe clusters) unless upstream fixes what they work around. GoogCC is a candidate in the -[realtime uplink bake-off](/quest/m3/quic-realtime/README.md); if the bake-off +[realtime uplink bake-off](/quest/m4/quic-realtime.md); if the bake-off deletes it, drop that half. NetEq: [moq-dev/moq#7856](https://github.com/moq-dev/moq/pull/7856) ported diff --git a/quest/m3/watch-symlink-targets.md b/quest/m3/watch-symlink-targets.md index 26c545913c6..fc87b7230ad 100644 --- a/quest/m3/watch-symlink-targets.md +++ b/quest/m3/watch-symlink-targets.md @@ -9,7 +9,7 @@ covers roots, served chains, and client certificates. ## Plan Decided 2026-11-05 (maintainer, follow-up from -[moq-dev/moq#7906](https://github.com/moq-dev/moq/pull/7906)): ✅ m2. +[moq-dev/moq#7906](https://github.com/moq-dev/moq/pull/7906)): this joined what was then m2, later work, which is m3 since the 2026-11-05 re-cut. `watch::callback` (`rs/moq-tokio/src/watch.rs`) watches each configured file's directory. Its filter already reloads when a path resolves somewhere diff --git a/quest/m3/webcodecs-priming.md b/quest/m3/webcodecs-priming.md index 0480a65e150..1c7a1960789 100644 --- a/quest/m3/webcodecs-priming.md +++ b/quest/m3/webcodecs-priming.md @@ -69,4 +69,4 @@ Decided 2026-11-04 (maintainer, /quest-plan): - [moq-audio in the browser](/quest/m5/audio-web.md) - its WebCodecs backend's `delay()` reuses this measurement - [AudioToolbox encode](/quest/m3/audio-encode-audiotoolbox.md) - the native AAC encoder's delay, folded the same way -- [Chromium's Opus encoder stamps chunks from their input](/quest/m4/chromium-opus-timestamps.md) - another WebCodecs `AudioEncoder` timestamp bug +- [Chromium's Opus encoder stamps chunks from their input](/quest/m5/chromium-opus-timestamps.md) - another WebCodecs `AudioEncoder` timestamp bug diff --git a/quest/m3/webrtc-dial-policy.md b/quest/m3/webrtc-dial-policy.md index be5d093f839..a94db73cef4 100644 --- a/quest/m3/webrtc-dial-policy.md +++ b/quest/m3/webrtc-dial-policy.md @@ -41,7 +41,7 @@ Today: - moq-rtmp has no gap yet, only because it never resolves a name: `Client::connect` takes a `SocketAddr` and `Client::with_stream` a finished stream, so the caller picks the address. The - [rtmps:// quest](/quest/m3/rtmp-rtmps-tcurl.md) makes it dial a URL, and + [rtmps:// quest](/quest/m2/rtmp-rtmps-tcurl.md) makes it dial a URL, and from then on it has the same gap, which is why this waits on it. Guidance: @@ -56,6 +56,10 @@ Guidance: - moq-srt's client takes a `SocketAddr` like moq-rtmp does today. It is out of scope until it dials a URL. +## Waiting + +- Maintainer: where the one dial-policy type lives. (a) A plain value type with no I/O in moq-net, which moq-webrtc and moq-rtmp both already depend on, each crate checking it at connect time; (b) moq-tokio, beside its own dial dispatch, though moq-rtmp uses moq-tokio only as a dev-dependency today; (c) a crate of its own. Recommended: (a), with the moq-webrtc half added through its `#[non_exhaustive]` `client::Config`, and the rtmps quest's new `connect` taking a `#[non_exhaustive]` options value so this lands without a second break. + ## Required -- [moq-rtmp dials rtmps:// and always advertises a tcUrl](/quest/m3/rtmp-rtmps-tcurl.md) - the URL dial in moq-rtmp that the policy covers +- [moq-rtmp dials rtmps:// and always advertises a tcUrl](/quest/m2/rtmp-rtmps-tcurl.md) - the URL dial in moq-rtmp that the policy covers diff --git a/quest/m3/webrtc/README.md b/quest/m3/webrtc/README.md index d1e87be2c15..1f047235dd9 100644 --- a/quest/m3/webrtc/README.md +++ b/quest/m3/webrtc/README.md @@ -15,7 +15,7 @@ Decided 2026-10-30 with the maintainer (planning interview), after the str0m wor - **What goes** (from the 2026-10-30 inventory: we use frame mode, full ICE, and DTLS 1.2, and never enable str0m's BWE, pacer, or stats): GCC bandwidth estimation (8k lines; GCC is ported into `moq-congestion` as `moq_congestion::gcc`), the leaky pacer (replaced by moq-congestion's pacer, made public by [rtc-pacing](/quest/m3/rtc-pacing.md)), H.266, RED, DTMF, G.7xx and comfort noise, VLA and simulcast layers, RTCP app feedback, stats, internal test exports, and dimpl's DTLS 1.3 and version auto-selection. RTCP XR stays: its RRTR/DLRR blocks are the receive side's only RTT measurement, which sets NACK retries and the video recovery deadline. Roughly 17k of str0m's 72k lines and 11k of dimpl's 33k. - **DTLS 1.2 only.** Every browser negotiates 1.2 today, and the vendored rind crypto provider offers 1.2 only to match. - **SCTP stays**, for browser data channels in [P2P](/quest/m4/p2p/README.md), but compiles only with the data-channel feature, so WHIP/WHEP builds carry none of it. Vendored, it can later be simplified or put on moq-congestion. -- **Upstream fixes:** after vendoring, nobody merges upstream automatically. Record the vendored upstream versions in the crate docs; the release-cut audit checks the upstream changelogs for security fixes (malformed SRTP/RTCP input, for example) and ports them. +- **Upstream fixes:** after vendoring, nobody merges upstream automatically. The vendored upstream commits are recorded in `rs/moq-webrtc/src/vendor/mod.rs`'s `Upstream` section; the [vendor sync](/quest/m3/vendor-sync/webrtc-stack.md) checks the upstream changes once per release cycle for security fixes (malformed SRTP/RTCP input, for example) and ports them. - Docs a change makes stale (crate names, `doc/lib/rs`, `doc/setup/upgrade.md`, other quests naming a crate or module) change in the PR that makes them stale. ## Related diff --git a/quest/m4/whip-whep-browser.md b/quest/m3/whip-whep-browser.md similarity index 88% rename from quest/m4/whip-whep-browser.md rename to quest/m3/whip-whep-browser.md index 2494b465152..334469a2639 100644 --- a/quest/m4/whip-whep-browser.md +++ b/quest/m3/whip-whep-browser.md @@ -14,7 +14,9 @@ Planned 2026-11-08 (maintainer, approved follow-up of [moq-dev/moq#8229](https://github.com/moq-dev/moq/pull/8229), which moved the stack onto rind assuming browser interop tests existed). None drives a browser against moq-webrtc today: no recipe in `test/` or workflow mentions -WHIP or WHEP. Milestone: m4, a test nobody waits on. +WHIP or WHEP. + +Moved to m3 2026-11-12 (maintainer, /quest-audit): a downstream deployment serves WHIP and WHEP from this stack. - Reuse the Playwright setup the interop and audio-quality harnesses use (`test/interop/clients/js/`), with the browsers' fake capture devices. diff --git a/quest/m4/2819-moq-video-carry-pipewire-dma-bufs-safely-into-the-vulkan.md b/quest/m4/2819-moq-video-carry-pipewire-dma-bufs-safely-into-the-vulkan.md index dced1559fbc..63e88f07c7a 100644 --- a/quest/m4/2819-moq-video-carry-pipewire-dma-bufs-safely-into-the-vulkan.md +++ b/quest/m4/2819-moq-video-carry-pipewire-dma-bufs-safely-into-the-vulkan.md @@ -1,31 +1,30 @@ -# [M] moq-video: validate PipeWire DMA-BUFs into the Vulkan renderer +# [S] moq-video: run the PipeWire DMA-BUF render path on a desktop GPU ## Goal The Linux zero-copy spine from [#2819](https://github.com/moq-dev/moq/issues/2819), -`PipeWire DMA-BUF -> Surface::DmaBuf -> Vulkan import -> render shader`, is -proven on real hardware. +`PipeWire DMA-BUF -> Surface::DmaBuf -> Vulkan import -> render shader`, has +been run on an Intel or AMD desktop, and the DRM modifiers its captures +arrive with are recorded. ## Plan -Built already: the `dmabuf` feature and `Surface::DmaBuf`, PipeWire DMA-BUF -negotiation with the shared-memory fallback, the dequeued buffer retained -until the last clone drops (#2839), packed RGB import in -`rs/moq-video/src/render/dmabuf.rs`, and NV12 import (#3331), which aliases -the buffer as an `R8` luma and an `RG8` chroma texture with no copy. VA-API -encode takes a `Surface::DmaBuf` directly. +The code and its ignored hardware tests exist: packed capture through Vulkan +(`packed_dmabuf_renders_through_vulkan`, `rs/moq-video/src/render/renderer.rs`), +and NV12 and I420 import from VA-API surfaces +(`the_nv12_dmabuf_import_matches_the_cpu_path` and its siblings). A Hyprland +capture now delivers DMA-BUF frames +([moq-dev/moq#8579](https://github.com/moq-dev/moq/pull/8579)). What remains: -- Hardware gates, as ignored tests with a reason where CI lacks the device: - native Linux `--features pipewire,render` tests; the shared-memory fallback - still captures; packed and NV12 DMA-BUF capture renders with zero CPU - download on an Intel or AMD desktop; holding several frames never shows - reused content or exhausts the PipeWire pool for good. -- Modifier mismatch: `render/dmabuf.rs` downloads a buffer whose modifier the - driver will not import, and re-tiles nothing. A VA-API VPP re-tile is only - worth adding if a measured capture source lands on such a modifier; record - the modifiers seen and decide. +- Run the ignored packed test on an Intel or AMD desktop and record the + result in the PR, including that holding several frames never shows reused + content or exhausts the PipeWire pool for good. +- Modifiers: `render/dmabuf.rs` refuses a buffer whose modifier the driver + lists for neither format and re-tiles nothing. Log the modifiers each + capture source arrives with. A VA-API VPP re-tile is only worth adding if a + measured source lands on an unimportable one. Decided in the 2026-09-30 audit: V4L2 `VIDIOC_EXPBUF` export is dropped. V4L2 capture only takes YUYV and MJPEG (`rs/moq-video/src/capture/v4l2.rs`), @@ -33,3 +32,7 @@ both of which need a CPU conversion or decode anyway, so exporting the buffer saves no copy. Revisit only if NV12 camera capture lands. Refs #2481, #1837. + +## Related + +- [Video hardware validation](/quest/m5/video-hardware.md) - owns the KDE capture stall and the shared-memory fallback check diff --git a/quest/m4/README.md b/quest/m4/README.md index 5671e191e2a..8752c724b16 100644 --- a/quest/m4/README.md +++ b/quest/m4/README.md @@ -9,8 +9,19 @@ plans need. ## Plan +Decided 2026-11-13 (maintainer, /quest-plan): this milestone requires the +[release cut](/quest/m3/release/cut.md), so until it lands spare capacity goes +to m0 to m3 and then to [m9](/quest/m9/README.md)'s lab work (rejected: +requiring m0, m1, and m2, which lifts only when they empty, and m0 rarely +does; also gating m3, which holds the cut). The cut's PR deletes the cut +quest, which lifts this gate. + Decided 2026-11-05 (maintainer, milestone re-cut): the lower half of the old m2 ("later work"), as [m3](/quest/m3/README.md) describes. After each cut m4's quests move to m3 and m4 starts empty. A quest here that gains a consumer moves to m3; one that loses its last reason moves to m5 or is deleted. + +## Required + +- [Release cut](/quest/m3/release/cut.md) - the cut lands before this milestone's work starts diff --git a/quest/m4/aec-tests.md b/quest/m4/aec-tests.md index 08cce01b113..4514ade24f0 100644 --- a/quest/m4/aec-tests.md +++ b/quest/m4/aec-tests.md @@ -12,8 +12,7 @@ partial frame into the next output. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7369](https://github.com/moq-dev/moq/pull/7369): m2, as -recommended. [moq-dev/moq#7369](https://github.com/moq-dev/moq/pull/7369) (merged) exposes `aec::Processor`, +[moq-dev/moq#7369](https://github.com/moq-dev/moq/pull/7369). [moq-dev/moq#7369](https://github.com/moq-dev/moq/pull/7369) (merged) exposes `aec::Processor`, `aec::Stream`, and `aec::Stats` for caller-supplied PCM. Add the tests beside [moq-dev/moq#7369](https://github.com/moq-dev/moq/pull/7369)'s unit tests (`mod tests`, diff --git a/quest/m4/announce-interest-memory.md b/quest/m4/announce-interest-memory.md index c361b848092..be3cc0dc964 100644 --- a/quest/m4/announce-interest-memory.md +++ b/quest/m4/announce-interest-memory.md @@ -15,16 +15,21 @@ From the 2026-10-24 pre-cut audit at 151a529670. Source-inferred, not measured; inherited (release's `lite/publisher.rs:659` has the same per-stream `origin.announced()`). -Each stream runs `origin.announced()` (`rs/moq-net/src/lite/publisher.rs` -about line 968, with no `Slots` acquire), materializing its own -`TableCursor.current`, `OriginConsumerState` pending/due/delivered entries, -and the `AnnounceRun` live map (`model/origin.rs` `register_cursor` about -line 5075, `Consumer::announced` about 6116). On lite-05/06, `init()` also -encodes the whole initial set before the first flush. Only QUIC stream -credit bounds the count (`DEFAULT_MAX_STREAMS = 10_000`), so the worst case +Each stream scopes the origin and opens its own `origin.routes()` cursor +(`rs/moq-net/src/lite/publisher.rs`, the announce stream's `start`, about +line 1058, with no `Slots` acquire), materializing its own `TableCursor` +with `OriginConsumerState` pending/due/delivered entries, and the +`AnnounceRun` live map (`model/origin.rs` `register_cursor` about line 6481, +`Consumer::routes` about 7698). On lite-05/06, `init()` also encodes the +whole initial set before the first flush. Only QUIC stream credit bounds the +count (`moq_tokio::quic::DEFAULT_MAX_STREAMS = 10_000`), so the worst case is about 10k times N times a few hundred bytes. Incoming announces are bounded by `session::Limits`; outgoing interest streams are not. +Once [one route stream](/quest/m2/cluster-routes/route-stream.md) +lands, lite-07 holds one interest stream per side, so what remains exposed is +lite-05/06 and moq-transport. + Bound it by design, not with a cap: cursors walk a shared, refcounted table snapshot lazily, materializing only what is in flight and gated on stream write credit; lite-05/06 encodes its initial set incrementally under @@ -33,6 +38,15 @@ the same pattern and gets the same fix. Public API: none. Wire: none. +## Waiting + +- Maintainer: how to bound the legacy versions' interest streams: the shared + snapshot redesign above, a `session::Limits` cap on outgoing interest + streams (which the Plan rules out today), or deleting this once lite-07 is + the oldest version served. Recommended: the cap for lite-05/06 and + moq-transport now, then delete, since redesigning `model/origin.rs` for + versions being retired costs more than it saves. + ## Related -- [Cluster routing](/quest/m2/cluster-routing/README.md) - reshapes the route table this walks +- [Cluster routing](/quest/m3/cluster-routing/README.md) - reshapes the route table this walks diff --git a/quest/m3/archive-dvr.md b/quest/m4/archive-dvr.md similarity index 100% rename from quest/m3/archive-dvr.md rename to quest/m4/archive-dvr.md diff --git a/quest/m4/archive-store-bench.md b/quest/m4/archive-store-bench.md deleted file mode 100644 index 66fe1b73f57..00000000000 --- a/quest/m4/archive-store-bench.md +++ /dev/null @@ -1,16 +0,0 @@ -# [M] Archive reads against a real object store - -## Goal - -A benchmark or nightly probe measures `moq_archive::Consumer` against a real -object store (S3, R2, or MinIO), counting LIST and GET calls and latency as -the track count grows, so a regression in archive reads shows up. - -## Plan - -- The cost is store round trips (per refresh, one LIST of the archive prefix - plus one per track; one GET per object), which an `InMemory` Criterion - bench would hide ([moq-dev/moq#6262](https://github.com/moq-dev/moq/pull/6262)). - `object_store`'s `aws` feature is already a dev feature. -- Sweep both axes, track count and objects per track, per the fan-out - benchmark rule. diff --git a/quest/m4/audio-bufferbloat-av.md b/quest/m4/audio-bufferbloat-av.md index 7fdb85e42e4..805d3797456 100644 --- a/quest/m4/audio-bufferbloat-av.md +++ b/quest/m4/audio-bufferbloat-av.md @@ -58,5 +58,5 @@ Public API: none. Wire: none. ## Related +- [ViSQOL grading](/quest/m5/audio-visqol.md) - adds a quality score to the same harness - [Audio playout](/quest/m2/audio-playout/README.md) - the playout rule and estimator whose budgets these rows extend -- [ViSQOL grading](/quest/m4/audio-visqol.md) - adds a quality score to the same harness diff --git a/quest/m3/audio-flush-group.md b/quest/m4/audio-flush-group.md similarity index 69% rename from quest/m3/audio-flush-group.md rename to quest/m4/audio-flush-group.md index 57ce2ffadb1..2174c4b745b 100644 --- a/quest/m3/audio-flush-group.md +++ b/quest/m4/audio-flush-group.md @@ -20,3 +20,12 @@ Players set their own latency budget, so this isn't urgent. Check against `rs/moq-audio/src/encode/producer.rs` on `main` first: the group rules there (a discontinuity or reanchor opens a marker group) may already cover the flush. + +Checked 2026-11-12 (/quest-audit): not done. When the codec's terminal drain +carries discard padding, `publish_terminal` +(`rs/moq-audio/src/encode/producer.rs:610`) writes an empty `keyframe: true` +frame, which opens a new group for the tail. The same race can hit any group +boundary under a zero `delay.max`; the final flush is only where it costs the +most. + +Moved to m4 2026-11-12 (maintainer, /quest-audit): no consumer waits on it, since the downstream deployment worked around it. diff --git a/quest/m4/auth-session-index.md b/quest/m4/auth-session-index.md index be3b54eb334..95294d2188e 100644 --- a/quest/m4/auth-session-index.md +++ b/quest/m4/auth-session-index.md @@ -8,12 +8,12 @@ With `--limit-token` or `--limit-remote` set, the auth server's `connect`, `end` Found by [moq-dev/moq#7512](https://github.com/moq-dev/moq/pull/7512), which attributed JWT verification cost and runs it nightly: one `connect` plus `end` costs 66.5k instructions (9.1 µs) at 0 live sessions, 147k (18.6 µs) at 1024, and 1.27M (158 µs) at 16,384, about 73 instructions per live session. The cost is `Sessions::sweep` plus the per-token and per-remote count scans in `Sessions::connect` (`rs/moq-auth/src/serve.rs`, about line 375); `Revalidate` pays the sweep too. Its `verify/sessions/{0,1024,16384}` bench is the baseline. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2. This waits on [moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211) landing, which rewrites `Sessions`' token bookkeeping (shared request-token grants and attachment leases); indexing the table first would conflict with it, as #7512 noted. +[moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211), which rewrote `Sessions`' token bookkeeping and which this waited on, has merged, so nothing blocks it. The count scans are now around `rs/moq-auth/src/serve.rs` lines 587 to 660. - Keep `HashMap` and `HashMap` beside the slots, updated on connect, end, a new token attaching to a known session, and expiry. `Sessions::revalidate` updates them too, since it re-inserts an aged-out or restart-lost slot and appends a token the slot hadn't seen, and it keeps today's rule of enforcing no cap. - Age slots from a queue ordered by `seen`, refreshed on each sighting (a re-push with a generation, or an ordered map keyed by `(seen, id)`), so a sweep pops only expired entries. Keep today's rule: a session that missed two cadences is dropped. - Keep the semantics the doc comment on the limits describes, including a known id attaching another credential counting only against the new JWT's cap. -- Key the per-token count on the token's signing input (header and payload) or its decoded claims, not the whole compact JWS: an ECDSA signature can be malleated to a second valid string, which today doubles `limits.token` for an ES256 or ES384 holder (found by the 2026-11-07 invariant audit, auth-6). +- The per-token count's key is [its own fix](/quest/m1/token-count-signing-input.md), split out 2026-11-12 (maintainer, /quest-audit) because it corrects shipped behavior; index whatever key that leaves. - Measure with #7512's `verify/sessions` bench before and after; the slope should be flat. Public API: none. Wire: none. diff --git a/quest/m4/bench-hygiene.md b/quest/m4/bench-hygiene.md index 211d2e0af7c..c3dd59972f5 100644 --- a/quest/m4/bench-hygiene.md +++ b/quest/m4/bench-hygiene.md @@ -3,26 +3,24 @@ ## Goal Three gaps in the nightly benchmarks are closed: hang's buffered bench has a -`burst` variant with a nonzero warmup, the benchmark report prunes series +`burst` variant with a nonzero warmup, the benchmark report drops sources that haven't been measured in N days, and the JS slow-group bench's `bytes` column stays out of series names. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: - -- m0 -- m1 -- ✅ m2 +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs). - **Warm burst.** `js/hang/bench/buffered.ts`'s `burst` scenario runs with zero warmup, so it never reaches the warming path [moq-dev/moq#7523](https://github.com/moq-dev/moq/pull/7523) kept linear. Add a variant with a nonzero warmup (and min delay) that would go quadratic without #7523's fix. -- **Prune stale series.** Series nothing measures any more, such as the - retired `bench-ffi` ones, linger in the report (`sh/gh/bench-report.py`). - Drop series not measured in N days, choosing N and recording why. +- **Prune stale sources.** The report already hides a series its source's + latest run did not measure, and labels a source that stopped running as + stale with its age (`sh/gh/bench-report.py`, about lines 531 and 545). A + retired source, such as `bench-ffi`, still keeps its stale section. Drop a + source not measured in N days, choosing N and recording why. - **Slow-group names.** `js/hang/bench/slow-group.ts`'s `bytes` column ends up in its series names; report it as a value instead, so a change in bytes doesn't start a new series. diff --git a/quest/m4/bench-sources.md b/quest/m4/bench-sources.md index ebb29b517b9..02a4ad23e1e 100644 --- a/quest/m4/bench-sources.md +++ b/quest/m4/bench-sources.md @@ -19,8 +19,8 @@ series. The relay load benches left Nightly in through the same input format; the sources decided with the maintainer on 2026-10-15 left it for later: -- The audio-quality job's benchmarks: `just js qmux-bench`, `serve-bench` and - `ring-bench`. Check each prints a CSV table the `js` converter reads before +- The audio-quality job's benchmarks: `just js qmux-bench`, `serve-bench`, + `raw-audio-bench`, and `ring-bench`. Check each prints a CSV table the `js` converter reads before recording it. A source joins as a `bench-*` artifact with a recorded outcome, which the diff --git a/quest/m4/binding-export-coverage.md b/quest/m4/binding-export-coverage.md index 19445a7a311..7191cf9a4a5 100644 --- a/quest/m4/binding-export-coverage.md +++ b/quest/m4/binding-export-coverage.md @@ -16,3 +16,11 @@ it against the wrapper's public types, with an explicit allowlist for items a wrapper deliberately hides. Kotlin's fixed list moves to the shared one. When [the exporter](/quest/m3/ffi-exporter.md) replaces UniFFI, the list comes from it. + +## Waiting + +- Maintainer: when to build the FFI item list: from UniFFI's metadata now, + rebuilt once [the exporter](/quest/m3/ffi-exporter.md) lands, or only from + the exporter, by making that quest Required here. Recommended: only from + the exporter, since its switch keeps every wrapper's API and the UniFFI + version would be written twice. diff --git a/quest/m3/bun-check.md b/quest/m4/bun-check.md similarity index 100% rename from quest/m3/bun-check.md rename to quest/m4/bun-check.md diff --git a/quest/m4/cache-fairness.md b/quest/m4/cache-fairness.md index ada1561601a..c7b36a32faf 100644 --- a/quest/m4/cache-fairness.md +++ b/quest/m4/cache-fairness.md @@ -32,7 +32,7 @@ limits (`rs/moq-relay/src/limits.rs`) bound each session, not a tenant's total across sessions. Decided 2026-10-28 (maintainer): measure first, then propose. Waits for -[unbounded groups](/quest/m2/unbounded-groups/README.md), which bounds each +[unbounded groups](/quest/m3/unbounded-groups/README.md), which bounds each group copy at a 32 MB tail and lets front eviction shed a live latest group, so the measurement runs against the final eviction rules. Show with a test whether a heavy tenant (many sessions and tracks, or a stopped publisher) @@ -50,7 +50,7 @@ Wire: none. ## Required -- [Unbounded groups](/quest/m2/unbounded-groups/README.md) - each group copy holds a 32 MB tail and front eviction can shed a live latest group +- [Unbounded groups](/quest/m3/unbounded-groups/README.md) - each group copy holds a 32 MB tail and front eviction can shed a live latest group ## Related diff --git a/quest/m3/capture-app-audio-macos.md b/quest/m4/capture-app-audio-macos.md similarity index 100% rename from quest/m3/capture-app-audio-macos.md rename to quest/m4/capture-app-audio-macos.md diff --git a/quest/m4/capture-exclude.md b/quest/m4/capture-exclude.md index dd0aa2b18a8..e1d4d06d9ef 100644 --- a/quest/m4/capture-exclude.md +++ b/quest/m4/capture-exclude.md @@ -91,11 +91,18 @@ Tests: Public API: breaking if the exclusion joins `Source::Display`, additive otherwise. Wire: none. +Decided 2026-11-12 (maintainer, /quest-audit before the cut): ✅ this quest +stays after the cut, and +[capture listings and selections can grow](/quest/m2/api-capture-fences.md) +fences the display selection the exclusion joins in m2 (rejected: pulling this quest into m2; a second moq-video +break). What this adds is then additive. + ## Required +- [Capture listings and selections can grow](/quest/m2/api-capture-fences.md) - the fenced structs this adds fields to - [A capture listing a picker can draw](/quest/m3/capture-list.md) - the app identity an exclusion names ## Related - [The foreground app](/quest/m4/capture-foreground.md) - names apps by the same identity -- [Per-app audio on macOS](/quest/m3/capture-app-audio-macos.md) - the same content filter, limiting audio to one app +- [Per-app audio on macOS](/quest/m4/capture-app-audio-macos.md) - the same content filter, limiting audio to one app diff --git a/quest/m3/cmsg-len-send.md b/quest/m4/cmsg-len-send.md similarity index 100% rename from quest/m3/cmsg-len-send.md rename to quest/m4/cmsg-len-send.md diff --git a/quest/m3/crypto-key-load.md b/quest/m4/crypto-key-load.md similarity index 68% rename from quest/m3/crypto-key-load.md rename to quest/m4/crypto-key-load.md index 2010f01f0f6..7237610d6f9 100644 --- a/quest/m3/crypto-key-load.md +++ b/quest/m4/crypto-key-load.md @@ -15,4 +15,6 @@ Planned 2026-11-03 (maintainer) from Kixel's `@kixel/mesh` had to carry its own Ed25519 and z-base-32 code because `SecretKey` cannot load a stored key and its `sign` is internal. Keep generated keys working, keep the browser path non-extractable, and test that a key loaded from bytes -signs what Rust verifies. Ranked at the top of m3 beside the mesh libraries. +signs what Rust verifies. A raw 32-byte seed already loads through +`SecretKey.fromSeed` (`js/crypto/src/sign/key.ts`), but it and `sign` are +marked `@internal`, and no `CryptoKey` path exists. Moved to m4 2026-11-12 (maintainer, /quest-audit): the downstream application it was ranked for made moq-crypto opt-in and plans nothing on it. diff --git a/quest/m3/dart-ios.md b/quest/m4/dart-ios.md similarity index 100% rename from quest/m3/dart-ios.md rename to quest/m4/dart-ios.md diff --git a/quest/m4/env-knobs.md b/quest/m4/env-knobs.md index 371c970c3a8..c331abf0f29 100644 --- a/quest/m4/env-knobs.md +++ b/quest/m4/env-knobs.md @@ -8,9 +8,12 @@ from [moq-dev/moq#6633](https://github.com/moq-dev/moq/pull/6633). ## Plan -- Rename test and bench knobs out of `MOQ_*` (for example `MOQ_SHAPER_SEED`, - `MOQ_QUIC_SEED`, `MOQ_JSON_SEED`, `MOQ_AUDIO_DEVICE`, `MOQ_BENCH_*`, - `MOQ_X11_BENCH_LOG`), updating CI and docs that set them. +- Rename test and bench knobs out of `MOQ_*`: `MOQ_SHAPER_SEED`, + `MOQ_QUIC_SEED`, `MOQ_JSON_SEED`, `MOQ_AUDIO_DEVICE`, `MOQ_X11_BENCH_LOG`, + and the binding bench runners' `MOQ_BENCH_SECONDS` and `MOQ_BENCH_DIR` + (`rs/moq-ffi/src/bench.rs`, `sh/*/bench.sh`), updating CI and docs that + set them. moq-bench's own `MOQ_BENCH_*` settings stay: that binary declares + them (`rs/moq-bench/src/config.rs`). - Decided 2026-10-25 (recommendation, maintainer delegated): a library reads no environment; the binaries that link it declare their settings and pass them in. moq-video's half (`MOQ_VAAPI_DEVICE`, `MOQ_V4L2_*`, and the public diff --git a/quest/m3/ffi-future-size.md b/quest/m4/ffi-future-size.md similarity index 100% rename from quest/m3/ffi-future-size.md rename to quest/m4/ffi-future-size.md diff --git a/quest/m4/format-gate-followups.md b/quest/m4/format-gate-followups.md index 41cddd6a133..0f2285a417f 100644 --- a/quest/m4/format-gate-followups.md +++ b/quest/m4/format-gate-followups.md @@ -1,27 +1,23 @@ -# [XS] One Ruff pin and Swift formatting on aarch64-linux +# [XS] One Ruff pin ## Goal Only the dev shell pins Ruff: `ruff` leaves the uv dev group in -`pyproject.toml` (and `uv.lock`). The Linux gate checks Swift formatting on -aarch64-linux too, once nixpkgs builds `swiftc` there. +`pyproject.toml` (and `uv.lock`). ## Plan -Decided 2026-10-31 (maintainer, the /quest-complete requeue session), the -follow-ups of [moq-dev/moq#7319](https://github.com/moq-dev/moq/pull/7319) +Decided 2026-10-31 (maintainer, the /quest-complete requeue session), a +follow-up of [moq-dev/moq#7319](https://github.com/moq-dev/moq/pull/7319) (format gates), which has landed. -- **Ruff.** #7319 makes the repo-wide `python` check module run the dev - shell's Ruff and keeps `ruff` in the uv dev group (`pyproject.toml:61`) - only for editors, so two Ruff versions are pinned. Drop it from the dev - group and `uv.lock`, and point editor setup at the dev shell's Ruff. -- **Swift on aarch64-linux.** #7319 adds `swift-format` to the dev shell - only off aarch64-linux (`flake.nix`, around line 306), because nixpkgs - marks its `swiftc` broken there, so `sh/swift/format.sh` skips and - release-PR Check (`ubuntu-24.04-arm`) checks no Swift formatting. When - nixpkgs builds it, add `swift-format` there and drop the skip. Until then - this half waits; check nixpkgs' `swift` package `meta.broken` for - aarch64-linux. +#7319 makes the repo-wide `python` check module run the dev shell's Ruff and +keeps `ruff` in the uv dev group (`pyproject.toml:61`) only for editors, so +two Ruff versions are pinned. Drop it from the dev group and `uv.lock`, and +point editor setup at the dev shell's Ruff. + +Dropped 2026-11-12 (/quest-audit): checking Swift formatting on aarch64-linux +once nixpkgs builds `swiftc` there. The x86 gates already check it, and only +the release-PR Check on `ubuntu-24.04-arm` misses it. Public API: none. Wire: none. diff --git a/quest/m4/fuzz-corpus-js.md b/quest/m4/fuzz-corpus-js.md index 5ee2a68ac13..83d11aa3993 100644 --- a/quest/m4/fuzz-corpus-js.md +++ b/quest/m4/fuzz-corpus-js.md @@ -36,4 +36,4 @@ Public API: none. Wire: none. ## Related -- [Shared corpora live in each crate's fixtures/](/quest/m4/corpus-fixtures.md) - where the committed inputs live +- [Shared corpora live in each crate's fixtures/](/quest/m5/corpus-fixtures.md) - where the committed inputs live diff --git a/quest/m3/ietf-fill-refill.md b/quest/m4/ietf-fill-refill.md similarity index 76% rename from quest/m3/ietf-fill-refill.md rename to quest/m4/ietf-fill-refill.md index a9bc94c21f2..49d30043c12 100644 --- a/quest/m3/ietf-fill-refill.md +++ b/quest/m4/ietf-fill-refill.md @@ -24,15 +24,15 @@ Group stitching across two streams drew repeated fixes: [#7322](https://github.com/moq-dev/moq/pull/7322) (headless groups) and join after finish ([moq-dev/moq#7306](https://github.com/moq-dev/moq/pull/7306)). The `Fill` enum and its handoff exist only because the model has one producer per group. In -`rs/moq-net/src/ietf/subscriber.rs` on be39188000 (refreshed 2026-11-02): -`Fill`, `FillRequest`, and `Filling` at lines 261-441, `claim_fill` at -5070-5110, and `recv_fill`, `run_fill`, and `run_fill_objects` from 5177. -Search by those names if the lines have moved again. +`rs/moq-net/src/ietf/subscriber.rs` on a86433043 (refreshed 2026-11-12): +`Fill`, `FillRequest`, and `Filling` at lines 308-498, `poll_claim_fill` at +6655, and `recv_fill` and `run_fill` from 6753. Search by those names if the +lines have moved again. Decided 2026-10-28 (maintainer): nothing is written into a group copy below -its live log, so the head no longer fills the tail's group in place. Once -[Flights](https://github.com/moq-dev/moq/pull/7944) lands, the head is a flight -from object 0 that hands over to the tail copy, which also keeps a tail +its live log, so the head no longer fills the tail's group in place. +[Flights](https://github.com/moq-dev/moq/pull/7944) have landed (8cd44b57f), +so the head becomes a flight from object 0 that hands over to the tail copy, which also keeps a tail whose head never arrives, as the headless group quest decided. Joining at the next group instead was considered and rejected: it costs a GOP of startup and drops objects already received. @@ -40,4 +40,4 @@ Public API: none. Wire: none. ## Required -- [moq-transport requests end through one lease](/quest/m3/ietf-request-lease.md) - restructures the same files first +- [moq-transport requests end through one lease](/quest/m4/ietf-request-lease.md) - restructures the same files first diff --git a/quest/m3/ietf-request-lease.md b/quest/m4/ietf-request-lease.md similarity index 100% rename from quest/m3/ietf-request-lease.md rename to quest/m4/ietf-request-lease.md diff --git a/quest/m3/import-metadata-policy.md b/quest/m4/import-metadata-policy.md similarity index 86% rename from quest/m3/import-metadata-policy.md rename to quest/m4/import-metadata-policy.md index a941b64383d..857367b663f 100644 --- a/quest/m3/import-metadata-policy.md +++ b/quest/m4/import-metadata-policy.md @@ -11,7 +11,9 @@ The two importers settled this separately during the 2026-10-27 security audit: - fMP4 ([moq-dev/moq#7232](https://github.com/moq-dev/moq/pull/7232), merged): what must be held whole is refused at its header, which reversed the 2026-10-28 incremental plan, because incremental parsing could not bound what it kept. [moq-dev/moq#7687](https://github.com/moq-dev/moq/pull/7687) then measured that mp4-atom expands a parsed body up to tenfold, so a `moov` or `moof` body is bounded by `fmp4::MAX_METADATA_BYTES` (2 MiB) and a text sample by `fmp4::MAX_HELD_BYTES` (32 MiB, the transport's per-group `MAX_CACHE_BYTES`). A `moof` declaring more than `fmp4::MAX_FRAGMENT_SAMPLES` (2^17) is refused when it parses, and a `moov` declaring more than `fmp4::MAX_TRACKS` (256) before any track is created. - MKV ([moq-dev/moq#7234](https://github.com/moq-dev/moq/pull/7234), merged): metadata parses incrementally with no blanket size cap (maintainer, 2026-10-28). Each collected master keeps one value per field, so repeats replace rather than accumulate. [moq-dev/moq#7629](https://github.com/moq-dev/moq/pull/7629) bounded the fields read whole: `DocType` and `CodecID` by the longest supported value, and `CodecPrivate` by each codec's supported configuration, refused at its header (FLAC STREAMINFO alone, the RFC 7845 OpusHead, av1C with a sequence header and HDR metadata, avcC by its count fields at about 34 MiB, and an AudioSpecificConfig by its syntax maximum). An hvcC is canonicalized as it arrives instead: its parameter sets are kept, at most one per id, and SEI is discarded unread. These are private constants in `rs/moq-mux/src/container/mkv/import.rs`, `codec::h264::AVCC_MAX_LEN`, and `codec::aac::CONFIG_MAX_LEN`. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2. Pick one policy for both importers, rather than leaving each at the decision its own audit PR made. Starting point: a byte bound at the header for whatever must be held whole, since #7232 found that is what actually bounds memory, while MKV keeps streaming what it can consume incrementally. Bring the choice to the maintainer in the PR if the evidence points elsewhere. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): pick one policy for both importers, rather than leaving each at the decision its own audit PR made. Starting point: a byte bound at the header for whatever must be held whole, since #7232 found that is what actually bounds memory, while MKV keeps streaming what it can consume incrementally. Bring the choice to the maintainer in the PR if the evidence points elsewhere. + +Moved to m4 2026-11-12 (maintainer, /quest-audit): no consumer, and both importers are already bounded. - The fMP4 side has landed (`MAX_HELD_BYTES`, `MAX_METADATA_BYTES`, `MAX_FRAGMENT_SAMPLES`, and `MAX_TRACKS` in `rs/moq-mux/src/container/fmp4/mod.rs`). Note the MKV avcC bound (about 34 MiB) sits far above fMP4's 2 MiB `MAX_METADATA_BYTES`. - Inventory what each importer still holds whole, and how much parsed state it builds from it (fMP4's parse expands up to tenfold, which is what sets `MAX_METADATA_BYTES`). diff --git a/quest/m3/intra-refresh/README.md b/quest/m4/intra-refresh/README.md similarity index 100% rename from quest/m3/intra-refresh/README.md rename to quest/m4/intra-refresh/README.md diff --git a/quest/m3/intra-refresh/consumer-warmup.md b/quest/m4/intra-refresh/consumer-warmup.md similarity index 97% rename from quest/m3/intra-refresh/consumer-warmup.md rename to quest/m4/intra-refresh/consumer-warmup.md index 6acab564d5f..37797e8ba06 100644 --- a/quest/m3/intra-refresh/consumer-warmup.md +++ b/quest/m4/intra-refresh/consumer-warmup.md @@ -54,4 +54,4 @@ The fix is history rather than a withhold rule: ## Required -- [Plan: recovery boundaries](/quest/m3/intra-refresh/recovery-boundary.md) - K derives from `warmup`, which must be a safe upper bound +- [Plan: recovery boundaries](/quest/m4/intra-refresh/recovery-boundary.md) - K derives from `warmup`, which must be a safe upper bound diff --git a/quest/m3/intra-refresh/dual-decoder.md b/quest/m4/intra-refresh/dual-decoder.md similarity index 93% rename from quest/m3/intra-refresh/dual-decoder.md rename to quest/m4/intra-refresh/dual-decoder.md index c0f5a1dac58..cc3f53530a6 100644 --- a/quest/m3/intra-refresh/dual-decoder.md +++ b/quest/m4/intra-refresh/dual-decoder.md @@ -34,5 +34,5 @@ What remains is checking the other engines. ## Related -- [H.264 import](/quest/m3/intra-refresh/h264-import.md) - publishes the real `warmup` for an end-to-end x264 check -- [Consumer warmup](/quest/m3/intra-refresh/consumer-warmup.md) - the transcoder decodes history so it never encodes stripes +- [H.264 import](/quest/m4/intra-refresh/h264-import.md) - publishes the real `warmup` for an end-to-end x264 check +- [Consumer warmup](/quest/m4/intra-refresh/consumer-warmup.md) - the transcoder decodes history so it never encodes stripes diff --git a/quest/m3/intra-refresh/export-sync-flags.md b/quest/m4/intra-refresh/export-sync-flags.md similarity index 95% rename from quest/m3/intra-refresh/export-sync-flags.md rename to quest/m4/intra-refresh/export-sync-flags.md index 1561fd9c2a1..0c2c202c741 100644 --- a/quest/m3/intra-refresh/export-sync-flags.md +++ b/quest/m4/intra-refresh/export-sync-flags.md @@ -27,4 +27,4 @@ Audio `warmup` is Opus pre-roll and changes no export flag. ## Required -- [Consumer warmup](/quest/m3/intra-refresh/consumer-warmup.md) - the IDR check this reuses +- [Consumer warmup](/quest/m4/intra-refresh/consumer-warmup.md) - the IDR check this reuses diff --git a/quest/m3/intra-refresh/h264-import.md b/quest/m4/intra-refresh/h264-import.md similarity index 97% rename from quest/m3/intra-refresh/h264-import.md rename to quest/m4/intra-refresh/h264-import.md index 2f6707aaff1..9895ff8acc3 100644 --- a/quest/m3/intra-refresh/h264-import.md +++ b/quest/m4/intra-refresh/h264-import.md @@ -48,4 +48,4 @@ follows the hang draft, which rounds up ## Required -- [Plan: recovery boundaries](/quest/m3/intra-refresh/recovery-boundary.md) - resolve the later-mutated warmup contract before implementing it +- [Plan: recovery boundaries](/quest/m4/intra-refresh/recovery-boundary.md) - resolve the later-mutated warmup contract before implementing it diff --git a/quest/m3/intra-refresh/h265-import.md b/quest/m4/intra-refresh/h265-import.md similarity index 97% rename from quest/m3/intra-refresh/h265-import.md rename to quest/m4/intra-refresh/h265-import.md index 93cb14a1a70..8a0253e2025 100644 --- a/quest/m3/intra-refresh/h265-import.md +++ b/quest/m4/intra-refresh/h265-import.md @@ -39,4 +39,4 @@ that machinery in parallel. ## Required -- [H.264 import](/quest/m3/intra-refresh/h264-import.md) - the shared measurement and catalog mutation, built on catalog warmup +- [H.264 import](/quest/m4/intra-refresh/h264-import.md) - the shared measurement and catalog mutation, built on catalog warmup diff --git a/quest/m3/intra-refresh/recovery-boundary.md b/quest/m4/intra-refresh/recovery-boundary.md similarity index 74% rename from quest/m3/intra-refresh/recovery-boundary.md rename to quest/m4/intra-refresh/recovery-boundary.md index 5c651e13746..764fa7dd239 100644 --- a/quest/m3/intra-refresh/recovery-boundary.md +++ b/quest/m4/intra-refresh/recovery-boundary.md @@ -8,9 +8,12 @@ earlier ones. Rewrite the import and consumer quests around that contract. ## Plan -Decided 2026-11-08 (maintainer, /quest-plan): moved back from m2. Intra -refresh is not part of this cut, so `warmup` ships with its current -duration contract and this plan settles only what the m3 importer needs. +Decided 2026-11-08 (maintainer, /quest-plan): intra refresh is not part of +this cut, so `warmup` ships with its current duration contract and this plan +settles only what the importer needs. + +Moved to m4 2026-11-12 (maintainer, /quest-audit): no downstream application +waits on intra-refresh import, and no native publisher sets video `warmup`. The boundary stays a catalog field, never a LOC frame property or other per-frame container field, and it is a duration, not a frame count. A per-group exact boundary carried in the media was offered and declined. @@ -53,7 +56,11 @@ Produce implementation and automated software-decode tests in the existing quests. Use generated fixtures, run short cases in CI, and retain full reference-decode comparisons for continuous playback and each recovery case. +## Waiting + +- Maintainer: how a catalog `warmup` stays safe when a later sweep is longer than the published one. Options: a bound published before the content it covers (costs one sweep of latency up front); a fixed upper bound from importer config; refusing the stream once a sweep exceeds the published `warmup`. Recommended: refuse, matching h264-import's rule that a stream is refused rather than published with a guess, and let a configured bound raise the limit where an operator knows its encoder. + ## Related -- [H.264 import](/quest/m3/intra-refresh/h264-import.md) - publishes recovery information -- [Consumer warmup](/quest/m3/intra-refresh/consumer-warmup.md) - uses it to size the transcoder's decode history +- [H.264 import](/quest/m4/intra-refresh/h264-import.md) - publishes recovery information +- [Consumer warmup](/quest/m4/intra-refresh/consumer-warmup.md) - uses it to size the transcoder's decode history diff --git a/quest/m4/lint-examples.md b/quest/m4/lint-examples.md index 7f8b02bb825..95e82923aa3 100644 --- a/quest/m4/lint-examples.md +++ b/quest/m4/lint-examples.md @@ -1,29 +1,29 @@ -# [S] Branch checks lint examples +# [XS] PR CI builds the C++, Go, and Python interop clients ## Goal -On Linux, `just check` runs clippy on the example targets of each crate it -selects, so a broken or lint-failing example fails the branch rather than the -next Nightly. PR CI also builds the C++, Go, and Python interop clients. +PR CI builds the C++, Go, and Python interop clients +(`test/interop/clients/{cpp,go,python}`) when their bindings or the client +change, so a binding API change that breaks one fails the branch rather than +the next interop run. ## Plan -Found 2026-10-28 by [moq-dev/moq#7142](https://github.com/moq-dev/moq/pull/7142): -`just check`'s Rust steps all use `--tests`, so the workspace's 23 examples, -including the interop clients that PR brought into scope, get clippy only from -`just fix`, the Nightly features run, and the compile-only Windows/macOS -check. - -Decided 2026-10-28 (maintainer): examples only. The 68 benches stay with -Nightly, since they cost more to build. Measure the added build time on a -moq-cli change and record it in the PR. - -Decided 2026-10-31 (maintainer, the /quest-complete requeue session): PR CI -also builds the C++, Go, and Python interop clients -(`test/interop/clients/{cpp,go,python}`) when their bindings or the client -change. They went stale unnoticed while +Decided 2026-10-31 (maintainer, the /quest-complete requeue session): the +clients went stale unnoticed while [moq-dev/moq#7311](https://github.com/moq-dev/moq/pull/7311) changed the decoder API, since only the interop runs build them. A build (and Python import or type check) is enough; running them stays with interop. +Today no module in `sh/dispatch.sh` matches those directories (the `go`, +`cpp`, and `py` modules cover only `go/`, `cpp/moq/`, and `py/`), so a change +there or a bindings change selects nothing that builds them. Widen the +matching modules, or add a step to them, so each client builds when its +binding or its own directory changes. + +The Rust half of the original quest landed in +[moq-dev/moq#7786](https://github.com/moq-dev/moq/pull/7786): the gate's +`just rs test` adds `--benches --examples` (`sh/rs/select.sh`), so the Rust +interop clients and every example are linted on the branch. + Public API: none. Wire: none. diff --git a/quest/m4/miri-seeds-tsan.md b/quest/m4/miri-seeds-tsan.md index a0963e8a0ed..cd03fd4293f 100644 --- a/quest/m4/miri-seeds-tsan.md +++ b/quest/m4/miri-seeds-tsan.md @@ -36,5 +36,5 @@ Public API: none. Wire: none. ## Related -- [Wire fuzz coverage](/quest/m1/wire-fuzz-coverage.md) - adds moq-c, moq-uring, and moq-sock to the same Miri recipe -- [Drop kio's epoch fence](/quest/m4/crossbeam-pin-fence.md) - the ordering in kio's log these runs watch +- [Wire fuzz coverage](/quest/m1/wire-fuzz-coverage.md) - adds moq-uring and moq-sock to the same Miri recipe +- [Drop kio's epoch fence](/quest/m5/crossbeam-pin-fence.md) - the ordering in kio's log these runs watch diff --git a/quest/m3/mkv-flac-channel-mask.md b/quest/m4/mkv-flac-channel-mask.md similarity index 100% rename from quest/m3/mkv-flac-channel-mask.md rename to quest/m4/mkv-flac-channel-mask.md diff --git a/quest/m3/quic-realtime/moqsink-feedback.md b/quest/m4/moqsink-feedback.md similarity index 86% rename from quest/m3/quic-realtime/moqsink-feedback.md rename to quest/m4/moqsink-feedback.md index 5c22962ff38..092f52aa8d6 100644 --- a/quest/m3/quic-realtime/moqsink-feedback.md +++ b/quest/m4/moqsink-feedback.md @@ -8,7 +8,7 @@ capacity drops instead of queueing media behind the pacer. ## Plan -Decided 2026-10-17: its own quest in the realtime epic, independent of the +Decided 2026-10-17: its own quest beside the [realtime uplink](/quest/m4/quic-realtime.md), independent of the controller bake-off. It works with today's BBR3 estimate and gains the full latency win once `Realtime` lands. @@ -29,4 +29,4 @@ Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and r ## Related -- [Realtime uplink congestion control](/quest/m3/quic-realtime/README.md) - the controller whose rate this follows +- [Realtime uplink congestion control](/quest/m4/quic-realtime.md) - the controller whose rate this follows diff --git a/quest/m4/moqsrc-reconnect-order.md b/quest/m4/moqsrc-reconnect-order.md index 30fe96c6d9f..619a553352d 100644 --- a/quest/m4/moqsrc-reconnect-order.md +++ b/quest/m4/moqsrc-reconnect-order.md @@ -11,8 +11,7 @@ live pad per rendition, no duplicate or leaked pads, and no deadlock. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7368](https://github.com/moq-dev/moq/pull/7368): m2, as -recommended. [moq-dev/moq#7368](https://github.com/moq-dev/moq/pull/7368) (merged) moved each rendition pad +[moq-dev/moq#7368](https://github.com/moq-dev/moq/pull/7368). [moq-dev/moq#7368](https://github.com/moq-dev/moq/pull/7368) (merged) moved each rendition pad into one actor (`rs/moq-gst/src/source/`). Reuse [moq-dev/moq#7368](https://github.com/moq-dev/moq/pull/7368)'s restart/relist/retire/finish tests and their blocked-pad diff --git a/quest/m3/net-broadcast-requests.md b/quest/m4/net-broadcast-requests.md similarity index 87% rename from quest/m3/net-broadcast-requests.md rename to quest/m4/net-broadcast-requests.md index 7af45f167b7..a53bc9a0910 100644 --- a/quest/m3/net-broadcast-requests.md +++ b/quest/m4/net-broadcast-requests.md @@ -17,5 +17,6 @@ track per thread created up front, so T3 claims a broadcast per entity instead and can collapse to one once this lands. Today the request pull (`requested()` in `js/net/src/broadcast.ts`) is reachable only by wire layers, and only the origin has a public `requests()`; keep the internal -wire path on the new handler. Ranked at the top of m3 beside the mesh -libraries, which Kixel builds on from `main`. +wire path on the new handler. + +Moved to m4 2026-11-12 (maintainer, /quest-audit): the downstream host it was ranked for is parked. diff --git a/quest/m4/net-test-subscriber.md b/quest/m4/net-test-subscriber.md deleted file mode 100644 index cb99ac19a5f..00000000000 --- a/quest/m4/net-test-subscriber.md +++ /dev/null @@ -1,31 +0,0 @@ -# [XS] moq-net's library tests pass under cargo test - -## Goal - -`cargo test -p moq-net --lib` passes, not only under nextest: -`setup::tests::credentials_stay_out_of_traces` and the model tests using -`model/test_tracing.rs` no longer each install a process-global tracing -subscriber, so whichever runs second in the shared process still captures -its events. - -## Plan - -Decided 2026-10-31 (maintainer, the /quest-complete requeue session). -Nextest runs each test in its own process, which hides the clash. - -`credentials_stay_out_of_traces` (`rs/moq-net/src/setup.rs:306`) uses -`#[tracing_test::traced_test]`, and `count_drop_warnings` -(`rs/moq-net/src/model/test_tracing.rs:39`) calls -`tracing::subscriber::set_global_default` and expects no other global -subscriber. Its doc explains why it is global: tracing caches callsite -interest process-wide, so a scoped dispatcher misses events another thread -cached as disabled. Keep that property. Likely one shared global test -subscriber that both use (`count_drop_warnings`' thread-local capture slot -extended to the credential check), or the credential test moved onto -`count_drop_warnings`' capture. Confirm with `cargo test -p moq-net --lib`. - -Public API: none. Wire: none. - -## Related - -- [moq-dev/moq#7243](https://github.com/moq-dev/moq/pull/7243) - fixed the same nextest-only pass in moq-clock (now `kio::clock`) by isolating its process state diff --git a/quest/m4/p2p/README.md b/quest/m4/p2p/README.md index 3709b2a7236..4d4b22d9539 100644 --- a/quest/m4/p2p/README.md +++ b/quest/m4/p2p/README.md @@ -118,7 +118,7 @@ channel, and has no reset code without a side channel. ### Routes and migration -Routing is the [cluster routing line](/quest/m2/cluster-routing/README.md)'s: +Routing is the [cluster routing line](/quest/m3/cluster-routing/README.md)'s: every session speaks per-node ROUTEs and path-less ANNOUNCEs, and the application priority followed by total cost picks the direct link or relay. diff --git a/quest/m4/pcm-bulk-encode.md b/quest/m4/pcm-bulk-encode.md index d1d0bdd65b6..9a1a580085e 100644 --- a/quest/m4/pcm-bulk-encode.md +++ b/quest/m4/pcm-bulk-encode.md @@ -8,6 +8,6 @@ moq-audio's PCM encoder (`rs/moq-audio/src/encode/backend/pcm.rs`) writes a fram Found by [moq-dev/moq#7520](https://github.com/moq-dev/moq/pull/7520), whose `batches` bench uses PCM as the codec so buffering cost isn't hidden by encoder work. The per-sample `extend_from_slice` loop's instruction count moves with LTO code generation, which makes the PCM control noisier than it needs to be. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2. On little-endian targets the `f32` slice's bytes are already the payload, so copy them at once; keep a per-sample path only for big-endian, behind `cfg(target_endian = "big")`. Check the PCM decoder's mirror loop (`rs/moq-audio/src/decode/backend/pcm.rs`) and give it the same treatment if it has one. Measure with #7520's `rs/moq-audio/benches/batches.rs` count table (`--bench __none__`) once #7520 lands, and keep its `batch_size_does_not_change_the_published_track` test passing. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): on little-endian targets the `f32` slice's bytes are already the payload, so copy them at once; keep a per-sample path only for big-endian, behind `cfg(target_endian = "big")`. Check the PCM decoder's mirror loop (`rs/moq-audio/src/decode/backend/pcm.rs`) and give it the same treatment if it has one. Measure with #7520's `rs/moq-audio/benches/batches.rs` count table (`--bench __none__`), and keep its `batch_size_does_not_change_the_published_track` test passing. Public API: none. Wire: none. diff --git a/quest/m4/perf/cursor-version-skip.md b/quest/m4/perf/cursor-version-skip.md deleted file mode 100644 index 34fcc7ec750..00000000000 --- a/quest/m4/perf/cursor-version-skip.md +++ /dev/null @@ -1,21 +0,0 @@ -# [M] Track-state reads skip when nothing changed - -## Goal - -`Cursor::sync`, `Cursor::probe` and `GroupExpiry::is_expired` skip their -track-state reads when the track's version is unchanged, and fanout's mutex wait -falls with them. `Subscriber::poll_recv_group` and `poll_recv_datagram` then -move onto a `kio::Watch`. The watch resets when the tail-grace deadline arms -or the floor moves. - -## Plan - -Planned 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-complete run). Found by [moq-dev/moq#7532](https://github.com/moq-dev/moq/pull/7532). With -`poll_finished` on the watch, `Cursor::sync`'s track-state read is the largest -fanout wait (52% at 16 workers), followed by `probe`. - -- First make sure each read depends only on state that bumps the version. - `poll_recv_group` reads the wall clock (a copy's tail grace), and its floor - changes without a `Ready`, so it moves only with explicit resets. -- Measure fanout and chat with `just bench-profile lock` against main, at 4 - and 16 workers, on a quiet host. #7532 couldn't capture 16-worker chat. diff --git a/quest/m4/perf/kio-waiter-costs.md b/quest/m4/perf/kio-waiter-costs.md deleted file mode 100644 index c734c52bc32..00000000000 --- a/quest/m4/perf/kio-waiter-costs.md +++ /dev/null @@ -1,27 +0,0 @@ -# [S] kio waiters stop churning on long-lived drivers and per-chunk watches - -## Goal - -Three kio waiter costs are measured and fixed where they are real, each with before/after instruction counts: - -- A long-lived waiter parked on more than `PARKED` (16) lists is no longer discarded and rebuilt on every poll, measured on a relay session driver. -- IETF fetch serving keeps one close watch per fetch stream instead of building one per chunk. -- A short-lived `kio::log` costs measurably less than about 1.3k instructions to set up and clean up, or the measurement shows why it can't. - -## Plan - -From [moq-dev/moq#7556](https://github.com/moq-dev/moq/pull/7556) (a measured no-win on adapter waker allocation) and [moq-dev/moq#7531](https://github.com/moq-dev/moq/pull/7531) (borrow-free QUIC state prototypes). - -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, in the perf epic, one quest for the three, since they share kio's waiter and log internals and one measurement setup. - -- **`Park::hold`'s retire rule** (`rs/kio/src/waiter.rs`): `hold` replaces a waiter that is not `reusable()`, and a waiter parked on more than `PARKED = 16` lists is not ([moq-dev/moq#7789](https://github.com/moq-dev/moq/pull/7789) raised it from 8 to fix subscription ordering bugs). #7556 traced about 4 `Arc` per group on a session driver (`Handle::run` / `time::run`) to this: the new waiter's first registration lands on the session-close `Bridge` (moq-uring `adapter.rs` and moq-tokio `owned.rs` `poll_closed`). A relay session driver parks on many more lists, so measure there first (allocations and instructions per delivered group over `just bench-runtime` or a session bench). Then choose between raising `PARKED`, changing the retire rule, or leaving it, on the numbers. -- **IETF fetch close watch** (`rs/moq-net/src/ietf/publisher.rs`): the fetch chunk loop pins a fresh `stream.closed()` per chunk (about lines 2387, 2450, and 2488), so each chunk builds a `kio::wait` whose waiter registers on the send stream's close `Bridge`. Hoist it to one per fetch stream. #7556 did not measure this path; a FETCH delivery bench with many chunks per stream shows the before/after. -- **`kio::log` lifetime cost**: #7531 attributed about 1.3k instructions per short-lived log to its epoch pin, flush, and collect, plus segment retirement on drop. That matters to any design holding one log per stream or per group. Profile a create/append/drop microbench, and cut what doesn't need to run for a log that never grew past its first segment. - -A measured no-win for any of the three is a valid outcome; record it here and drop that part. - -Public API: none expected. Wire: none. - -## Related - -- [kio channel contention](/quest/m4/kio-channel-contention/README.md) - the append-only log design that would pay `kio::log`'s cost per stream diff --git a/quest/m3/plan-watch-worker.md b/quest/m4/plan-watch-worker.md similarity index 88% rename from quest/m3/plan-watch-worker.md rename to quest/m4/plan-watch-worker.md index 8b096afee96..a899031869d 100644 --- a/quest/m3/plan-watch-worker.md +++ b/quest/m4/plan-watch-worker.md @@ -3,7 +3,7 @@ ## Goal A measured decision on how `@moq/watch` moves its heavy lifting off the main -thread, recorded by rewriting [Watch worker](/quest/m3/watch-worker.md). The +thread, recorded by rewriting [Watch worker](/quest/m4/watch-worker.md). The jank harness and the N-player sweep that decided it land and run nightly; the prototypes themselves are discarded. @@ -16,6 +16,12 @@ keeps one session for publishing and watching. ## Plan +Moved from the lab (m9) to m4 2026-11-13 (maintainer, /quest-plan): +[watch-worker](/quest/m4/watch-worker.md) in m4 requires it, and no quest +requires one in a later milestone. ✅ m4, beside its only consumer, which +waits for the cut too (rejected: m3, planning ahead of a consumer that cannot +start; moving watch-worker and watch-decode-gate to the lab). + Prototype both models on the same pipeline and measure them against each other: - **Invisible worker, backend handles.** One lazily created worker per page @@ -71,4 +77,4 @@ follow-up wires the capture worklet to the encoder worker with a ## Related -- [Watch worker](/quest/m3/watch-worker.md) - the implementation this rewrites +- [Watch worker](/quest/m4/watch-worker.md) - the implementation this rewrites diff --git a/quest/m4/qmux-02-subprotocols.md b/quest/m4/qmux-02-subprotocols.md index 5822aac740d..3485247c885 100644 --- a/quest/m4/qmux-02-subprotocols.md +++ b/quest/m4/qmux-02-subprotocols.md @@ -11,23 +11,23 @@ fall back to 01 or 00 as today. ## Plan Decided 2026-11-03 (maintainer, /quest-plan of the follow-ups from -[moq-dev/moq#7664](https://github.com/moq-dev/moq/pull/7664)): milestone m2. +[moq-dev/moq#7664](https://github.com/moq-dev/moq/pull/7664)). - Decision prompt (maintainer, 2026-11-03): what this quest does about qmux-02: - ✅ Advertise 02 first, in the relay and moq-tokio, from one shared list. - Advertise 02 after 01. - Leave the choice to this quest. -- Today `rs/moq-relay/src/websocket.rs:230` lists +- Today `rs/moq-relay/src/websocket.rs:219` (at 4bf501b535) lists `QMUX_VERSIONS = [QMux01, QMux00]`, and qmux-02 is accepted only as a bare - fallback. moq-tokio pins `QMux01` in `websocket.rs:548`, `tcp.rs`, - `unix.rs:16`, `server.rs`, `connection.rs`, and `webrtc.rs`. + fallback. moq-tokio pins `QMux01` in `websocket.rs` (`qmux_versions_for`, + `:594`), `tcp.rs:19`, `unix.rs:16`, `server.rs`, and `connection.rs`. `js/qmux/src/session.ts` already offers `qmux-02.` pairs first. - `Version::QMux02` exists in Rust (`rs/moq-quic/src/connection/mux/tests.rs`). + `Version::QMux02` lives in the `rs/qmux` crate (`config.rs` already + defaults to it). - Keep the per-ALPN floor (`QMUX01_ONLY_ALPNS`: moq-transport-18 and newer need qmux-01 or later), and extend it to allow 02. -- Move the version list into one place that the relay and moq-tokio share, - as the comment beside `QMUX01_ONLY_ALPNS` asks. +- Move the version list into one place that the relay and moq-tokio share. - Added 2026-11-08 (maintainer, approved follow-up of [moq-dev/moq#8224](https://github.com/moq-dev/moq/pull/8224)): the relay now derives its subprotocols from `moq_net::Versions`, keeping every diff --git a/quest/m4/quic-deadline.md b/quest/m4/quic-deadline.md index 65da67307ca..df3a3eb7134 100644 --- a/quest/m4/quic-deadline.md +++ b/quest/m4/quic-deadline.md @@ -24,7 +24,7 @@ Implement in `moq-quic`. deadline, estimate the arrival instant as now plus the forward one-way delay. Start with `min_rtt / 2`, corrected by the peer's reported ACK delay. Measure whether receive-time differences (`moq_congestion::Acked::received`, - from the [realtime epic](/quest/m3/quic-realtime/README.md)) improve that + from the [realtime epic](/quest/m4/quic-realtime.md)) improve that estimate, accounting for clock offset, drift, and uncertainty. They do not alone measure absolute one-way delay. If the estimate is past the deadline, reset the stream with a dedicated error code and drop its retransmit ranges, @@ -91,7 +91,7 @@ Restored 2026-11-12 in m4 (maintainer, `/quest-import` of m5 for want of a consumer. The consumer: MPEG-TS trunks under a fixed `moq export ts --delay`, which compete with SRT on delivering whatever can still arrive within the delay. The outage loss that issue measured belongs to -[Buffered exports pick their group order by measurement](/quest/m1/export-order.md) +[Buffered exports pick their group order by measurement](/quest/m3/export-order.md) and [A path that's back is used within an RTT](/quest/m3/quic-path-return.md), not this quest: the early probe never exceeds the congestion window, which is full of unacknowledged packets during an outage. This quest is the diff --git a/quest/m4/quic-dev-dep-path.md b/quest/m4/quic-dev-dep-path.md deleted file mode 100644 index e650d318004..00000000000 --- a/quest/m4/quic-dev-dep-path.md +++ /dev/null @@ -1,13 +0,0 @@ -# [XS] moq-quic's internal dev-dependencies are path-only - -## Goal - -moq-quic's dev-dependency on moq-congestion uses a path, as every internal dev-dependency should, so publishing moq-quic never needs a matching moq-congestion release. External dev-dependencies take the workspace version, so moq-tokio no longer pins its own `tempfile`. - -## Plan - -Found by the agent on [moq-dev/moq#7377](https://github.com/moq-dev/moq/pull/7377): moq-quic's `[dev-dependencies]` take moq-congestion with `workspace = true`. Decided 2026-10-30 (maintainer, follow-up planning after the /quest-afk run). Switch it to a path and check no other crate has the same pattern (`cargo metadata`, or grep the `[dev-dependencies]` tables). - -Widened 2026-10-31 from [moq-dev/moq#7260](https://github.com/moq-dev/moq/pull/7260): `rs/moq-tokio/Cargo.toml` pins its dev-dependency `tempfile = "3"` directly, while every other crate takes `tempfile = { workspace = true }`. Switch it to the workspace entry in the same PR, and check the dev-dependency tables for other external crates pinned outside the workspace. - -Public API: none. Wire: none. diff --git a/quest/m3/quic-long-type.md b/quest/m4/quic-long-type.md similarity index 100% rename from quest/m3/quic-long-type.md rename to quest/m4/quic-long-type.md diff --git a/quest/m3/quic-realtime/README.md b/quest/m4/quic-realtime.md similarity index 95% rename from quest/m3/quic-realtime/README.md rename to quest/m4/quic-realtime.md index 087c51f9151..8ee7caa14ed 100644 --- a/quest/m3/quic-realtime/README.md +++ b/quest/m4/quic-realtime.md @@ -1,4 +1,4 @@ -# Realtime uplink congestion control +# [L] Realtime uplink congestion control ## Goal @@ -112,7 +112,7 @@ delay bound on `CongestionControl::Delay` before the bake-off: BBR3, versus 8.3 s with the delay bound. The sim step was lossy; it explains the remaining model hold, not the lossless socket result. -Remaining work once the candidates finish (this README's own job): +Remaining work once the candidates finish: - Run the shared comparison on the uplink presets: queuing delay p50/p95/p99, throughput, reaction time to a capacity step down, and share against one @@ -162,19 +162,16 @@ Remaining work once the candidates finish (this README's own job): on `cellular-step` and `cellular-deep` where BBR3 gets 98% ([moq-dev/moq#7544](https://github.com/moq-dev/moq/pull/7544)). -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs): both items above join this README's remaining work. Options -shown: +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs): the last two items above join this quest's remaining work. -- m0 -- m1 -- ✅ m2, in this epic +Moved to m4 2026-11-12 (maintainer, /quest-audit): its consumer, teleoperation uplinks, is deferred work, so nobody waits on it yet. The winner's target rate already reaches moq-cli and moq-ffi encoders through `bandwidth_estimate` → `estimated_send_rate` → `set_bitrate`. `moqsink` only exposes a read-only `estimated-send-rate` and never retunes the upstream encoder, so a GStreamer pipeline at a fixed bitrate still queues -when capacity drops; [its own quest](/quest/m3/quic-realtime/moqsink-feedback.md) -owns that feedback. +when capacity drops; [its own quest](/quest/m4/moqsink-feedback.md) +owns that feedback. The bake-off and shipping `Realtime` do not wait on it. ## Related diff --git a/quest/m3/quic-sansio-driver.md b/quest/m4/quic-sansio-driver.md similarity index 93% rename from quest/m3/quic-sansio-driver.md rename to quest/m4/quic-sansio-driver.md index 58102c8bec0..7ac6aea43ca 100644 --- a/quest/m3/quic-sansio-driver.md +++ b/quest/m4/quic-sansio-driver.md @@ -11,10 +11,13 @@ No tokio or io_uring is in its dependency tree. The first consumer is [moq-dev/rust2](https://github.com/moq-dev/rust2)'s exporter, whose C, C++, and Kotlin bindings drive it from their own loops -([sessions on sans-IO QUIC](https://github.com/moq-dev/rust2/blob/main/quest/m1/ffi-export/session-quic.md)). +([sessions on sans-IO QUIC](https://github.com/moq-dev/rust2/blob/main/quest/m3/session-quic.md), +which rust2 keeps in its deferred milestone). ## Plan +Moved to m4 2026-11-12 (maintainer, /quest-audit): its only consumer is deferred in rust2. + Decided with the maintainer on 2026-10-28, from [moq-dev/rust2#215](https://github.com/moq-dev/rust2/pull/215): diff --git a/quest/m3/readers-range-live.md b/quest/m4/readers-range-live.md similarity index 100% rename from quest/m3/readers-range-live.md rename to quest/m4/readers-range-live.md diff --git a/quest/m4/reconnect-trace-validation.md b/quest/m4/reconnect-trace-validation.md index c7e13d27115..787bdc7596c 100644 --- a/quest/m4/reconnect-trace-validation.md +++ b/quest/m4/reconnect-trace-validation.md @@ -15,7 +15,7 @@ moq-net): every TLA+ model gets trace validation, not route alone. Milestone m4, behind route's. - Reuse the trace format and the TLC recipe - [route trace validation](/quest/m1/route-trace-validation.md) builds, with + [route trace validation](/quest/m3/route-trace-validation.md) builds, with the `verify-reconnect` skill's Coverage table as the mapping from code to model actions. - `reconnect/tests.rs` holds fixtures, not a sweep. Add a seeded one on the @@ -32,4 +32,4 @@ Public API: none. Wire: none. ## Required -- [Route sweeps are checked against the TLA+ model](/quest/m1/route-trace-validation.md) - the trace format and TLC recipe this reuses +- [Route sweeps are checked against the TLA+ model](/quest/m3/route-trace-validation.md) - the trace format and TLC recipe this reuses diff --git a/quest/m4/regrant-route-recheck.md b/quest/m4/regrant-route-recheck.md index a9b64b4e29f..e6bca73f85f 100644 --- a/quest/m4/regrant-route-recheck.md +++ b/quest/m4/regrant-route-recheck.md @@ -6,14 +6,16 @@ When a peer's grant changes (a new `AUTH_OK`), a moq-lite announce stream's cost ## Plan -Found by [moq-dev/moq#7536](https://github.com/moq-dev/moq/pull/7536), which swept the worst known peer grant shapes over `AUTH_OK` size and left this scope question to a follow-up (its decision 2, recommended there because it is route fan-out rather than grant algebra). On a regrant, each announce stream's `regrant` (`AnnounceRun` in `rs/moq-net/src/lite/publisher.rs`, about line 1045): +Found by [moq-dev/moq#7536](https://github.com/moq-dev/moq/pull/7536), which swept the worst known peer grant shapes over `AUTH_OK` size and left this scope question to a follow-up (its decision 2, recommended there because it is route fan-out rather than grant algebra). On a regrant, each announce stream's `regrant` (`AnnounceRun::regrant` in `rs/moq-net/src/lite/publisher.rs`, line 1207 at 4bf501b535): - filters every live route through `permitted` (a match against every member of the new permit) to find what the new grant revokes, and - when the grant grew, takes a fresh `origin.announced()` snapshot of every route under its prefix and matches each one it doesn't hold. -So one `AUTH_OK` costs routes × members per stream, times K announce streams, a product none of [peer grant bound](/quest/m4/peer-grant-bound.md)'s options remove. +So one `AUTH_OK` costs routes × members per stream, times K announce streams, a product none of [peer grant bound](/quest/m5/peer-grant-bound.md)'s options remove. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, its own quest beside the grant-algebra bound. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): its own quest beside the grant-algebra bound. + +Re-scoped 2026-11-12 (/quest-audit): [Each side says everything about routing on one stream](/quest/m2/cluster-routes/route-stream.md) gives a lite-07 side one Announce Stream and turns a token-scope change into an edit of its interest set, so on lite-07 the factor K becomes one and the per-stream `regrant` this plan targets remains only for lite-06 and older. Start after it lands: re-measure the regrant (or interest-set edit) cost on lite-07 with the benches below, and apply the steps that follow to whichever path still pays routes × members. If only the older versions still pay it, say so in the PR and leave them. - Derive the change once per `AUTH_OK`, not per stream: the members added and removed between the old and new grant. Each stream then re-checks only the routes under a changed member's prefix, using the origin's prefix structure to find them instead of walking every route. A member's prefix here is its literal segments before its first wildcard (`*` or `**`), since grant members are `Patterns` (`Permit` in `rs/moq-net/src/auth.rs`); a member that starts with a wildcard has no literal prefix and re-checks every route under the stream's prefix, as today. Bench that shape on its own row, and test wildcard members, removed members, and old and new grants that overlap. - A grown grant needs only routes under the added members, not a whole-origin snapshot per stream. @@ -22,7 +24,11 @@ Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-a Public API: none. Wire: none. +## Required + +- [Each side says everything about routing on one stream](/quest/m2/cluster-routes/route-stream.md) - replaces lite-07's per-stream regrant with one interest-set edit per side, which this re-scopes to + ## Related -- [Bound peer grant processing by design](/quest/m4/peer-grant-bound.md) - may change the received-grant type this compares against; whichever lands second adapts +- [Bound peer grant processing by design](/quest/m5/peer-grant-bound.md) - may change the received-grant type this compares against; whichever lands second adapts - [Announce interest streams share one route table](/quest/m4/announce-interest-memory.md) - the per-stream route state that this walk reads diff --git a/quest/m4/relay-capped-credit-test.md b/quest/m4/relay-capped-credit-test.md index a8ec523e207..76f7f4cfb93 100644 --- a/quest/m4/relay-capped-credit-test.md +++ b/quest/m4/relay-capped-credit-test.md @@ -11,8 +11,7 @@ once it resumes. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7342](https://github.com/moq-dev/moq/pull/7342): m2, as -recommended. +[moq-dev/moq#7342](https://github.com/moq-dev/moq/pull/7342). [moq-dev/moq#6990](https://github.com/moq-dev/moq/pull/6990) fixed a `Start::Latest` join that subscribed with only the warmup as its budget, so @@ -22,7 +21,7 @@ relay's own resets while the reader withholds stream credit. - Put it in `rs/moq-relay/tests/integration/` with a relay, a publisher bursting a few groups, and a subscriber with a small `max_streams` (see - `peer_limits.rs:39`) and a `delay.max` larger than the burst, that stalls + `peer_limits.rs:48` at 4bf501b535) and a `delay.max` larger than the burst, that stalls before reading. Assert every stream it receives completes with no `StreamError::Old` and that a group past the budget is reset `Old`. - Drive the stall with the simulated clock, not wall-clock sleeps. diff --git a/quest/m3/relay-io-uring-package.md b/quest/m4/relay-io-uring-package.md similarity index 97% rename from quest/m3/relay-io-uring-package.md rename to quest/m4/relay-io-uring-package.md index 69ec9821dae..11b966bfea7 100644 --- a/quest/m3/relay-io-uring-package.md +++ b/quest/m4/relay-io-uring-package.md @@ -35,7 +35,7 @@ ships, not the default. and handed to their tokio stacks), what a ring relay still serves from the shared tokio runtime (TCP, WebSocket, Unix, HTTP, outbound cluster dials), and what it refuses loudly - (`listen.lb_id` QUIC-LB server ids). That refusal is a recorded exception + (`listen.lb.id` QUIC-LB server ids). That refusal is a recorded exception to parity, not a prerequisite, until a consumer asks. Decided 2026-11-05 (maintainer, /quest-audit): ✅ this waited for the io_uring diff --git a/quest/m4/relay-soak.md b/quest/m4/relay-soak.md index 14ea43db007..85096507e24 100644 --- a/quest/m4/relay-soak.md +++ b/quest/m4/relay-soak.md @@ -37,4 +37,4 @@ Public API: none. Wire: none. ## Related -- [Relay allocations](/quest/m4/perf/relay-allocs.md) - steady-state allocation cost, where this measures growth +- [Relay allocations](/quest/m9/perf/relay-allocs.md) - steady-state allocation cost, where this measures growth diff --git a/quest/m3/render-d3d11.md b/quest/m4/render-d3d11.md similarity index 100% rename from quest/m3/render-d3d11.md rename to quest/m4/render-d3d11.md diff --git a/quest/m4/resume-redesign.md b/quest/m4/resume-redesign.md index da8e089f847..7551e7bb760 100644 --- a/quest/m4/resume-redesign.md +++ b/quest/m4/resume-redesign.md @@ -37,10 +37,19 @@ in several owners, which must be kept in agreement by hand: - route failure versus content verdict: 39eacc807a, 109235732c, eb6ac57cb7, 5c31208409 - end agreement and budget/expiry: 25532eacc8, a602360c99, 097dc1ee62, f52848d634, b59fcefde3, e33fe24cc3 +The churn continues: `resume.rs` took 28 commits from 2026-11-01 to +2026-11-12, 9 of them fixes (for example e419224dc, f3edb0b8d). Since this +was written, `rs/moq-net/verify/resume/ResumeReader.tla` and +`rs/moq-net/verify/failover/FailoverChain.tla` model the per-reader path; a +redesign replaces or re-maps them (the `verify-resume` and `verify-failover` +skills name the code each covers). + Every `track::Consumer`/`Subscriber` method has two implementations (26 -`Inner::Resume` arms, 11 `serving()` delegations), so each feature (ranges, +`Inner::Resume` arms and 11 `serving()` delegations then; 45 arms and 26 +`serving()` calls in `track.rs` at 4bf501b535), so each feature (ranges, anchors, fills, compression, largest, drops, start, live) is written twice -and drifts. `js/net/src/resume.ts` (1,786 lines) mirrors it, and the resume floor fix +and drifts. `js/net/src/resume.ts` (1,786 lines then, 2,296 at 4bf501b535) +mirrors it, and the resume floor fix [moq-dev/moq#6759](https://github.com/moq-dev/moq/pull/6759), the resumed JS drops fix [moq-dev/moq#6758](https://github.com/moq-dev/moq/pull/6758), and the compress check [moq-dev/moq#6460](https://github.com/moq-dev/moq/pull/6460) are each a place the two @@ -73,7 +82,7 @@ b9398f7459, 68c73349a3). Session copies under a front must stop caching for themselves and only feed the logical track. Land it in Rust first, then mirror `js/net` unless -[JS-only @moq/net](/quest/m5/rs2ts-js.md) has replaced it. Extend +[the Rust model in the browser](/quest/m5/rs2ts-js.md) has replaced it. Extend `benches/origin.rs` with the reader times flap sweep before starting, to measure the change. @@ -88,6 +97,10 @@ reader's old `delay.max` after an update. It was found by review and not reproduced. The redesign's tests must cover a preference update after the serving copy's upstream closed, followed by a read. +## Waiting + +- Maintainer: whether this goes before [resume trace validation](/quest/m4/resume-trace-validation.md) and [mutation testing](/quest/m1/mutants.md), and whether it moves to m3 once the cut ships. Options: (a) redesign first, giving both of those `Required` on this quest (or limiting mutants to `route/`), and move this to m3 after the cut as the root fix for the resume path's recurring bugs in shipped code; (b) validate and mutate the current path first, keeping this in m4; (c) redesign first but keep it in m4. Recommended: (a), since the other two would spend their effort on the per-reader path this deletes. + ## Required - [Release cut](/quest/m3/release/cut.md) - wait for the release in flight to ship diff --git a/quest/m4/resume-trace-validation.md b/quest/m4/resume-trace-validation.md index dabe593fc8c..d0199cc2def 100644 --- a/quest/m4/resume-trace-validation.md +++ b/quest/m4/resume-trace-validation.md @@ -14,7 +14,7 @@ moq-net): every TLA+ model gets trace validation, not route alone, so none can keep passing on code that has moved on. Milestone m4, behind route's. - Reuse the trace format and the TLC recipe - [route trace validation](/quest/m1/route-trace-validation.md) builds. The + [route trace validation](/quest/m3/route-trace-validation.md) builds. The sweep records each step as the model action it corresponds to (the `verify-resume` skill's Coverage table is the mapping) with the variables the model keeps, and TLC checks the trace is a behaviour of the spec. @@ -32,7 +32,7 @@ Public API: none. Wire: none. ## Required -- [Route sweeps are checked against the TLA+ model](/quest/m1/route-trace-validation.md) - the trace format and TLC recipe this reuses +- [Route sweeps are checked against the TLA+ model](/quest/m3/route-trace-validation.md) - the trace format and TLC recipe this reuses ## Related diff --git a/quest/m4/route-wakes.md b/quest/m4/route-wakes.md index 5d3066acf3b..907b8a789bc 100644 --- a/quest/m4/route-wakes.md +++ b/quest/m4/route-wakes.md @@ -37,7 +37,9 @@ Decisions: (2026-09-30) - An index, not a payload filter on the watch: each route records the fronts it serves or is requesting through, and each prefix node records - only its waiting fronts and `routed_broadcast` waiters. A payload filter + only its waiting fronts and the pending `request` and `follow` resolutions + that retry when their covering routes change (`routed_broadcast` became + these in #7172). A payload filter would still walk every watch under the prefix, which stays linear in paths. (2026-09-30) - No pin sets (decided 2026-10-03: #4741 deletes `Pin`, so any covering route @@ -62,9 +64,7 @@ Decisions: the changed route now wins; only those fronts re-select. A leave wakes only the fronts the leaver served or was requesting through. - Builds on fronts keyed by effective exclusion, so the index hangs off the final front - identity. `origin-front-parks.md` replaces the `routed_broadcast` retry - loop, one of the watch consumers here; whichever lands second adapts it. - (2026-09-30) + identity. (2026-09-30) Decided 2026-10-13: coordinate the index with the per-origin route model and per-track active/standby selection. Routes and announces records identity @@ -80,6 +80,6 @@ served root is withdrawn. Public API: none. Wire: none. -## Related +## Required -- [Routes and announces](/quest/m2/cluster-routes/README.md) - the identity and selection views the index serves +- [Routes and announces](/quest/m2/cluster-routes/README.md) - settles the source-selection model the index serves, so it is built once on the final selection diff --git a/quest/m5/rs2ts-model-subset.md b/quest/m4/rs2ts-model-subset.md similarity index 92% rename from quest/m5/rs2ts-model-subset.md rename to quest/m4/rs2ts-model-subset.md index f5caadd99cd..ae60ce8a714 100644 --- a/quest/m5/rs2ts-model-subset.md +++ b/quest/m4/rs2ts-model-subset.md @@ -4,7 +4,7 @@ moq-net's model, and the kio (clock included) and moq-pattern code it calls, stay inside the subset [rust2-ts](https://github.com/moq-dev/rust2/blob/main/ts/README.md) translates, -so the [source-translator route](/quest/m5/rs2ts-wasm2js.md) can be measured on the model +so the [source-translator route](/quest/m5/rs2ts-js.md) can be measured on the model with no hand-written stand-ins for moq's own logic. ## Plan @@ -45,7 +45,3 @@ after rust2-ts's designs for addresses as identity this quest starts (rejected: deciding now). Public API: none. - -## Required - -- [rust2-ts reaches moq-net's model](/quest/m5/rs2ts-model-ready.md) - the translator is ready for the model diff --git a/quest/m4/shaper-socket.md b/quest/m4/shaper-socket.md index 03da4387618..87628b33607 100644 --- a/quest/m4/shaper-socket.md +++ b/quest/m4/shaper-socket.md @@ -8,7 +8,7 @@ moq-shaper's sockets are sized like the relay's, so they don't drop at the rates Found by [moq-dev/moq#7529](https://github.com/moq-dev/moq/pull/7529)'s UDP queue-loss survey: moq-shaper binds its listen socket (`UdpSocket::bind` in `rs/moq-shaper/src/lib.rs`, `Shaper::bind`) at the kernel's 208 KiB default. Once in a shaped run it dropped 10 datagrams at its own socket, which showed as unattributed, since the shaper reports only its queue overflows; a drill would read them as path loss. #7529 also measured the relay's 8 MiB buffers absorbing a 100 ms stall with no drops, where 208 KiB dropped about 13% under 20 ms stalls. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2. Do both halves, since they're each small: give the shaper's sockets the relay's buffer sizing (moq-sock's 8 MiB `UDP_BUFFER` sizing in `rs/moq-sock/src/bind.rs`; reuse it rather than a second constant), and count its own socket's `sk_drops` (`SO_MEMINFO`, as #7529's survey read it) into the shaper's stats so drills report them separately. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): do both halves, since they're each small: give the shaper's sockets the relay's buffer sizing (moq-sock's 8 MiB `UDP_BUFFER` sizing in `rs/moq-sock/src/bind.rs`; reuse it rather than a second constant), and count its own socket's `sk_drops` (`SO_MEMINFO`, as #7529's survey read it) into the shaper's stats so drills report them separately. Optional, only if it stays small: export per-socket `sk_drops` on the relay's `/metrics`. #7529 found nothing in production that needed it, so skip it if it needs new plumbing through both runtimes. diff --git a/quest/m4/shared-serve.md b/quest/m4/shared-serve.md index 42b682554f0..4d88356521a 100644 --- a/quest/m4/shared-serve.md +++ b/quest/m4/shared-serve.md @@ -31,11 +31,11 @@ cheaper, since epoch semantics already live in the model. Decision prompts: -- Plan a shared core: not now; ✅ in m2; in m1. +- Plan a shared core: not now; ✅ after the cut; before the cut. - Scope: Rust first, then JS, publisher only; ✅ also the subscriber side; Rust and JS together; Rust only. - Rank: ✅ after Perf and the transport benches, so they show no regression; - with the moq-transport work; end of m2. + with the moq-transport work; last before the cut. Implementation guidance: @@ -61,4 +61,4 @@ Public API: none. Wire: none. ## Related - [Shared group serving and ingest in JS](/quest/m4/shared-serve-js.md) - the JS mirror -- [Perf](/quest/m4/perf/README.md) - hot-path work in the same machines +- [Perf](/quest/m9/perf/README.md) - hot-path work in the same machines diff --git a/quest/m4/sim-credit-open.md b/quest/m4/sim-credit-open.md deleted file mode 100644 index 8169feb9c20..00000000000 --- a/quest/m4/sim-credit-open.md +++ /dev/null @@ -1,37 +0,0 @@ -# [S] In-memory transports refuse a request without stream credit - -## Goal - -moq-net's in-memory test transports model the peer's bidirectional stream -credit, and `poll_open_request` returns `Ok(None)` (refused, as -`Error::NoStreamCredit` upstream) when none is left, as native QUIC does. -Integration tests then catch code that spends its own credit before a -prerequisite stream opens. - -## Plan - -Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 -/quest-complete run), from -[moq-dev/moq#7332](https://github.com/moq-dev/moq/pull/7332) (merged): m2, as -recommended. - -[moq-dev/moq#7332](https://github.com/moq-dev/moq/pull/7332)'s review found that pipelined FETCHes could take every bidi stream -before TRACK opened; on native QUIC, TRACK's `poll_open_request` fails at -once (`rs/moq-net/src/coding/stream.rs:52-61` at be39188000), but the -in-memory transports never refuse. `transport::memory` already counts bidi -credit (`Config::max_concurrent_bi_streams`, default unlimited, returned as -the peer drops both halves), yet its `poll_open_request` -(`rs/moq-net/src/transport/memory.rs:360-366`) parks on it like -`poll_open_bi`. The integration mock wrapping it -(`rs/moq-net/tests/integration/support/mock.rs:576`) and the lite test -sessions (`rs/moq-net/src/lite/test_transport.rs:535`, `:776`, `:1237`) -map `poll_open_bi` to `Some`. - -- Make `memory`'s `poll_open_request` return `Ok(None)` when its bidi credit - is exhausted, and have the mock delegate to it instead of its own - `poll_open_bi`. Where a lite test session gates bidi opens - (`with_open_bi_gate` parks `poll_open_bi`), refuse the request instead. -- Regression: with credit for one stream and fetches queued before TRACK, - the track still opens (the order [moq-dev/moq#7332](https://github.com/moq-dev/moq/pull/7332) fixed). - -Public API: none (test support only). Wire: none. diff --git a/quest/m4/smoke-libressl.md b/quest/m4/smoke-libressl.md index 66fbe230870..73c8385ee7f 100644 --- a/quest/m4/smoke-libressl.md +++ b/quest/m4/smoke-libressl.md @@ -9,7 +9,7 @@ or a system `openssl.cnf`. ## Plan Decided 2026-11-03 (maintainer, /quest-plan of the follow-ups from -[moq-dev/moq#7700](https://github.com/moq-dev/moq/pull/7700)): milestone m2. +[moq-dev/moq#7700](https://github.com/moq-dev/moq/pull/7700)). - #7700's moxygen smoke passed only with an explicit `OPENSSL_CONF`, because the dev shell's LibreSSL has no system `openssl.cnf`. diff --git a/quest/m3/spawn-audit.md b/quest/m4/spawn-audit.md similarity index 100% rename from quest/m3/spawn-audit.md rename to quest/m4/spawn-audit.md diff --git a/quest/m4/sps-fuzz-h264-reader.md b/quest/m4/sps-fuzz-h264-reader.md deleted file mode 100644 index 0308cd121ef..00000000000 --- a/quest/m4/sps-fuzz-h264-reader.md +++ /dev/null @@ -1,13 +0,0 @@ -# [XS] Re-run the SPS fuzzer on h264-reader - -## Goal - -The H.264 SPS fuzz target runs against moq-mux's parser after [moq-dev/moq#7398](https://github.com/moq-dev/moq/pull/7398) moves it onto `h264-reader`. Any overflow or panic it finds there is fixed in moq-mux, with its input committed under `rs/moq-mux/fuzz/regressions/`. - -## Plan - -Decided 2026-10-30 (maintainer, follow-up planning): [moq-dev/moq#7313](https://github.com/moq-dev/moq/pull/7313) found the crop and width overflows in `h264-parser` ([therishidesai/h264-parser#9](https://github.com/therishidesai/h264-parser/issues/9)). #7398 replaces that crate for H.264, so the finding has to be rechecked on the new parser. Run the target long enough to cover the old reproducer and a fresh corpus. - -The H.265 half landed separately: moq-mux reads every H.265 SPS with its own bounded parser. For H.264, h264-reader 0.9 bounds `ue(v)` at 2^32 - 2 and computes the picture size with checked arithmetic, and the old reproducer is committed as `rs/moq-mux/fuzz/regressions/fmp4/h264-sps-width` and passes. That review was by reading the code, so this run is what confirms it. - -Public API: none. Wire: none. diff --git a/quest/m4/stream-js.md b/quest/m4/stream-js.md new file mode 100644 index 00000000000..2f802866922 --- /dev/null +++ b/quest/m4/stream-js.md @@ -0,0 +1,26 @@ +# [S] JS byte streams + +## Goal + +`@moq/net` dials and accepts the same byte streams as Rust, exposed as a +WHATWG `ReadableStream`/`WritableStream` pair, and interoperates with the Rust +side in both directions. + +## Plan + +Mirror the [Rust surface](/quest/m1/stream/rust.md) on `@moq/net`'s socket. Add +the Rust/JS pair to `just test interop`. + +`@moq/net`'s socket `TcpStream` already exists and the socket interop test +(`test/interop/socket.sh`) runs, so this adds only the WHATWG adapter. The +JS tail is unpaced today (`js/net/src/socket.ts`: a reader falling behind it +fails with eviction), so a slow JS reader is not lossless until JS paces by +its fastest reader. + +Moved to m4 2026-11-12 (maintainer, /quest-audit): no downstream application +waits on JS byte streams, and they are lossless only after JS pacing lands. + +## Required + +- [Rust byte streams](/quest/m1/stream/rust.md) - the reference surface +- [JS fastest-reader pacing](/quest/m4/js-fastest-reader.md) - a JS reader that falls behind would otherwise end evicted, not lossless diff --git a/quest/m4/subscribe-trace-validation.md b/quest/m4/subscribe-trace-validation.md index 4ca65440ba7..fbbbf2abfa2 100644 --- a/quest/m4/subscribe-trace-validation.md +++ b/quest/m4/subscribe-trace-validation.md @@ -15,7 +15,7 @@ moq-net): every TLA+ model gets trace validation, not route alone. Milestone m4, behind route's. - Reuse the trace format and the TLC recipe - [route trace validation](/quest/m1/route-trace-validation.md) builds, with + [route trace validation](/quest/m3/route-trace-validation.md) builds, with the `verify-subscribe` skill's Coverage table as the mapping from code to model actions. - The model's steps are spread over `lite/publisher.rs` (`TrackRun`, @@ -34,7 +34,7 @@ Public API: none. Wire: none. ## Required -- [Route sweeps are checked against the TLA+ model](/quest/m1/route-trace-validation.md) - the trace format and TLC recipe this reuses +- [Route sweeps are checked against the TLA+ model](/quest/m3/route-trace-validation.md) - the trace format and TLC recipe this reuses ## Related diff --git a/quest/m3/subscriber-fin-fold.md b/quest/m4/subscriber-fin-fold.md similarity index 100% rename from quest/m3/subscriber-fin-fold.md rename to quest/m4/subscriber-fin-fold.md diff --git a/quest/m4/tenant-leak-checks.md b/quest/m4/tenant-leak-checks.md index 78923b43158..3b044eb83d8 100644 --- a/quest/m4/tenant-leak-checks.md +++ b/quest/m4/tenant-leak-checks.md @@ -1,32 +1,27 @@ -# [XS] The tenant matrix's HTTP checks assert exact refusals +# [XS] The tenant matrix's listing check asserts an exact status ## Goal -The relay tenant matrix's HTTP check asserts a definite outcome for each -cross-tenant request, so it cannot pass by timing out or by an unexpected -status: B listing A's root gets an exact refusal or B's own listing, and B -fetching A's track gets an exact refusal, with no 2-second timeout standing -in for "nothing served". +The relay tenant matrix's HTTP check asserts a definite outcome when B lists +A's root: an exact refusal status or B's own listing, never whichever status +comes back. ## Plan Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7366](https://github.com/moq-dev/moq/pull/7366): m2, as -recommended. [moq-dev/moq#7366](https://github.com/moq-dev/moq/pull/7366) (merged) added +[moq-dev/moq#7366](https://github.com/moq-dev/moq/pull/7366), which added `rs/moq-relay/tests/integration/tenant_matrix.rs`. -Its `http_routes_keep_tenants_apart` accepts any status for -`/announced/tenant-a?jwt=b` (checking the body only when it is 200) and -treats a `/fetch` still waiting after 2 s as "served nothing", since the -handler waits up to a hard-coded 30 s (`rs/moq-relay/src/web.rs:1130` at -be39188000). +`http_routes_keep_tenants_apart` accepts any status for +`/announced/tenant-a?jwt=b`, checking only that the body omits A's broadcast +and, when the status is 200, that it lists B's probe +(`tenant_matrix.rs:543-549` at 4bf501b535). Pin the status the auth contract +gives B for a path outside its root, and assert it. -- Pin the expected status of each request (what the auth contract says B - gets for a path outside its root) and assert it. -- Make the fetch answer definitively: either the refusal comes back before - the wait (the path is outside B's root, so no wait is needed), or the wait - becomes configurable so the test can set it short. Pick the one that - needs no new knob if possible. +The fetch half is done: +[moq-dev/moq#8087](https://github.com/moq-dev/moq/pull/8087) publishes a +decoy at B's root and asserts B's `/fetch` of A's path returns 200 with the +decoy, so it no longer leans on a timeout. -Public API: none, unless the fetch wait becomes a config field. Wire: none. +Public API: none. Wire: none. diff --git a/quest/m4/test-inventory-followups.md b/quest/m4/test-inventory-followups.md index 2dbce22bfc2..b93e8a67762 100644 --- a/quest/m4/test-inventory-followups.md +++ b/quest/m4/test-inventory-followups.md @@ -11,7 +11,6 @@ each orphan once. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from [moq-dev/moq#7316](https://github.com/moq-dev/moq/pull/7316) (merged), which finished the opt-in test inventory quest. -Milestone m2 (user: "use recommended for any CI/flakes"). [moq-dev/moq#7316](https://github.com/moq-dev/moq/pull/7316)'s review fixed a false pass: reaching any `gh nightly` group counted every branch of `sh/gh/nightly.sh`. It now binds a `just` call's literal diff --git a/quest/m4/token-auth-followups.md b/quest/m4/token-auth-followups.md new file mode 100644 index 00000000000..af84651bc7e --- /dev/null +++ b/quest/m4/token-auth-followups.md @@ -0,0 +1,25 @@ +# [S] The multi-token request cost is profiled + +## Goal + +The remaining multi-token gap in +`cargo bench -p moq-net --bench auth -- auth/request_tokens` is profiled, +and its cause either fixed or recorded in this quest's PR with numbers. + +## Plan + +Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from [moq-dev/moq#7211](https://github.com/moq-dev/moq/pull/7211), +which finishes request tokens. + +Moved to m4 2026-11-12 (maintainer, /quest-audit): only the profiling remains, a benchmark nobody waits on. +The renewal half is gone: merged #7211 applies one rule to a REQUEST_UPDATE's +AUTHORIZATION TOKEN whether the subscription is paused or forwarding, refusing a +token on a request admitted without one (`TOKEN_ON_SESSION_REQUEST` in +`rs/moq-net/src/ietf/publisher.rs`). + +- Single-token rounds match main; multi-token rounds are 1.5 to + 2.2x slower and grow with tokens times requests. The suspected cost is + each announce polling its guard on the shared verdict. Profile with + `samply` or `perf` on the bench before changing anything. + +Public API: none. Wire: none. diff --git a/quest/m4/track-test-eviction-asserts.md b/quest/m4/track-test-eviction-asserts.md index 10061103684..8310238d0f9 100644 --- a/quest/m4/track-test-eviction-asserts.md +++ b/quest/m4/track-test-eviction-asserts.md @@ -8,11 +8,7 @@ eviction turned off instead of passing. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: - -- m0 -- m1 -- ✅ m2 +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs): a quest of its own, after the cut (rejected: m0, m1). Today they pass with eviction off, so they don't guard what they name. Candidates include "an evicted successor with unread frames still starts the diff --git a/quest/m4/upstream-reports.md b/quest/m4/upstream-reports.md index 4a0328a525f..1a0f702734c 100644 --- a/quest/m4/upstream-reports.md +++ b/quest/m4/upstream-reports.md @@ -41,11 +41,11 @@ Approved to file (search for an existing report first): for the next codec. - **Biome:** filed 2026-10-31 as [biomejs/biome#12796](https://github.com/biomejs/biome/issues/12796) - (open): `useMathMinMax` flags comparisons of `bigint` operands, and its - autofix to `Math.min`/`Math.max` throws `TypeError` on bigints. #7305 - suppressed the 9 sites in `js/qmux` and `js/net/src/hop.ts`; their - `biome-ignore` comments don't link the report yet, so add the link, and - remove them once it is fixed. The string and `Date` cases were fixed + (open, checked 2026-11-12): `useMathMinMax` flags comparisons of `bigint` + operands, and its autofix to `Math.min`/`Math.max` throws `TypeError` on + bigints. #7305 suppressed the sites in `js/qmux` and `js/net/src/hop.ts`; + link the report from their 8 `biome-ignore` comments, then remove them once + it is fixed. The string and `Date` cases were fixed upstream ([biomejs/biome#12624](https://github.com/biomejs/biome/issues/12624), [biomejs/biome#10272](https://github.com/biomejs/biome/issues/10272)). @@ -57,3 +57,7 @@ Not to be filed: [DTLS waits for ICE](/quest/m3/webrtc/dtls-after-ice.md) fixes it there. Public API: none. Wire: none. + +## Waiting + +- Maintainer: which of the reports with no recorded approval may be posted: the GStreamer `faad` frame drop, retina 0.4.21's idle `Pending`, and the mr-boxington eviction (which [Mbx R2 verify](/quest/m0/ci/mbx-r2-verify.md)'s upstream threads may already cover). Options: (a) approve some or all, each after reading that project's AI policy and searching for an existing report; (b) drop all three and keep the approved Chromium and Firefox reports. Recommended: (b), since each workaround already works and none blocks other work. diff --git a/quest/m4/uring-lifecycle-tests.md b/quest/m4/uring-lifecycle-tests.md index b47c3b4497b..1dd78c29cbb 100644 --- a/quest/m4/uring-lifecycle-tests.md +++ b/quest/m4/uring-lifecycle-tests.md @@ -1,15 +1,11 @@ -# [XS] moq-uring's handle lifecycle tests cover the newly spawned queue and a woken close watcher +# [XS] moq-uring's handle lifecycle tests cover a driver in the newly spawned queue ## Goal -Two regressions cover the shutdown paths +A regression covers the shutdown path [moq-dev/moq#7309](https://github.com/moq-dev/moq/pull/7309) added: a `moq-uring` worker dropped while a connection driver is still in its newly -spawned queue (never polled) fails that connection's handles promptly, and -a reset arriving while moq-tokio's adapter `RecvStream` sits at its -read-ahead cap wakes the parked close watcher through the backend's own -wake, not a manual repoll. moq-tokio's test-only adapter read path is -narrowed if that removes adapter state. +spawned queue (never polled) fails that connection's handles promptly. ## Plan @@ -17,18 +13,17 @@ Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /que (merged; its online review's non-blocking coverage suggestions), which finished the io_uring handle lifecycle quest. -- Newly spawned queue: [moq-dev/moq#7309](https://github.com/moq-dev/moq/pull/7309) makes `Worker::drop` also drop tasks spawned - since its last turn (`rs/moq-uring/src/worker.rs`, `spawn` at :697 and - :789 at be39188000) and gives `Driver` a `Drop` that fails the connection - (`rs/moq-uring/src/quic/driver/connection.rs:827`). - Spawn a driver, drop the worker before its next turn, and check a handle - fails with the worker-gone error instead of pending. -- Reset at the cap: [moq-dev/moq#7309](https://github.com/moq-dev/moq/pull/7309)'s `recv_closed_sees_a_reset_at_the_cap` and - `recv_closed_resumes_when_a_read_drains_the_cap` in - `rs/moq-tokio/src/transport.rs` repoll `poll_closed` by hand. Use a - backend (moq-uring or the real moq-tokio QUIC stream) whose reset fires - the registered waker, and assert the watcher wakes on its own. -- The adapter read path [moq-dev/moq#7309](https://github.com/moq-dev/moq/pull/7309) added is reached only by tests today; if a - narrower shape passes both tests, take it. +[moq-dev/moq#7309](https://github.com/moq-dev/moq/pull/7309) makes `Worker::drop` also drop tasks spawned +since its last turn (`rs/moq-uring/src/worker.rs`, the `spawns` queue taken in +`impl Drop for Worker`) and gives `Driver` a `Drop` that fails the connection +(`impl Drop for Driver` in `rs/moq-uring/src/quic/driver/connection.rs`). +Only `dropped_worker_rejects_operations` covers a dropped worker today. Spawn +a driver, drop the worker before its next turn, and check a handle fails with +the worker-gone error instead of pending. + +The other half of the original quest, a reset at moq-tokio's adapter +read-ahead cap waking the close watcher, is gone: +[moq-dev/moq#7771](https://github.com/moq-dev/moq/pull/7771) replaced that +adapter read path, and its `recv_closed_*_cap` tests with it. Public API: none. Wire: none. diff --git a/quest/m4/usage-settings-once.md b/quest/m4/usage-settings-once.md deleted file mode 100644 index 6034e0c96de..00000000000 --- a/quest/m4/usage-settings-once.md +++ /dev/null @@ -1,30 +0,0 @@ -# [S] Upstream single-declaration settings in usage-rs - -## Goal - -usage-rs ([jdx/usage](https://github.com/jdx/usage)) builds the settings -registry from `#[derive(Cli)]` fields that carry `setting = "key"`, so -`rs/moq-tokio/src/settings.rs` and `rs/moq-relay/src/settings.rs` (626 lines -declaring every merged flag a second time) are deleted after a bump. Source -presence, precedence, deprecated-name rejection, and provenance keep working. -CLI behavior and wire are unchanged. - -## Plan - -Filed 2026-11-02 (maintainer-approved, /quest-unblock), building on the -one-source-of-truth goal of jdx/usage#896, #1180 and #1202. - -Today `usage::Cli` and `usage::Config` reject each other's attributes, and -`Registry::drift` tests the two declarations agree. The Cli side already names -`env` and `setting` per field; the registry side adds the type and `parse`. -Propose inferring those from the field so one declaration serves both. If -upstream declines, delete this quest and keep the drift test. - -Shipped 2026-11-03 (checked 2026-11-04, /quest-unblock): usage-lib -[7.2.0](https://github.com/jdx/usage/releases/tag/v7.2.0) lets derive flags -declare their own settings ([jdx/usage#1610](https://github.com/jdx/usage/pull/1610), -closing jdx/usage#1608). - -## Related - -- [Upstream unknown-env option in usage-rs](/quest/m4/usage-unknown-env.md) - another jdx/usage ask; file both together if both are approved diff --git a/quest/m4/usage-unknown-env.md b/quest/m4/usage-unknown-env.md deleted file mode 100644 index d2194b7db8f..00000000000 --- a/quest/m4/usage-unknown-env.md +++ /dev/null @@ -1,22 +0,0 @@ -# [S] Upstream unknown-env option in usage-rs - -## Goal - -usage-rs ([jdx/usage](https://github.com/jdx/usage)) can refuse unknown -environment variables under a prefix, the way `unknown_flags = "error"` -refuses unknown flags. Once a release ships it, bump usage-rs and replace -`moq_tokio::cli::check_env` with the option. The option needs a way to -accept names read outside the spec, such as `moq_video::ENV`. - -## Plan - -Filed 2026-11-02 (maintainer-approved, /quest-unblock). No existing issue or -release covered it; the closest, clap's declined `env_prefix`, builds names -from a prefix, which this does not. - -Shipped 2026-11-03 (checked 2026-11-04, /quest-unblock): usage-lib -[7.2.0](https://github.com/jdx/usage/releases/tag/v7.2.0) adds the opt-in -`unknown_env` node ([jdx/usage#1609](https://github.com/jdx/usage/pull/1609), -closing jdx/usage#1607), with derive syntax -`#[usage(unknown_env(prefix = "MOQ_", allow(...)))]` for names read outside -the spec. diff --git a/quest/m4/watch-decode-gate.md b/quest/m4/watch-decode-gate.md index bbf97892ec7..798c75087c6 100644 --- a/quest/m4/watch-decode-gate.md +++ b/quest/m4/watch-decode-gate.md @@ -25,4 +25,4 @@ Public API: none. Wire: none. ## Required -- [Watch worker](/quest/m3/watch-worker.md) - moves decode and `Sync` into a worker +- [Watch worker](/quest/m4/watch-worker.md) - moves decode and `Sync` into a worker diff --git a/quest/m4/watch-test-waits.md b/quest/m4/watch-test-waits.md index dd881cad908..a16cb88e37a 100644 --- a/quest/m4/watch-test-waits.md +++ b/quest/m4/watch-test-waits.md @@ -15,7 +15,7 @@ Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 [moq-dev/moq#7312](https://github.com/moq-dev/moq/pull/7312): - Fixed yields in watch tests: - - ✅ Quest in m2, one quest across `js/watch` tests. Reason: a fixed + - ✅ One quest across `js/watch` tests. Reason: a fixed yield count is flaky by construction. - Drop. diff --git a/quest/m3/watch-worker.md b/quest/m4/watch-worker.md similarity index 92% rename from quest/m3/watch-worker.md rename to quest/m4/watch-worker.md index 97db1ca2710..c4b99ef5335 100644 --- a/quest/m3/watch-worker.md +++ b/quest/m4/watch-worker.md @@ -11,7 +11,7 @@ instead of falling back to the main thread. ## Plan -[Plan: watch worker](/quest/m3/plan-watch-worker.md) picks between an +[Plan: watch worker](/quest/m4/plan-watch-worker.md) picks between an invisible page-wide worker with on-demand signal handles and application-spawned workers, and rewrites this quest with the result. Until then, the settled constraints in that quest apply here too. @@ -32,4 +32,4 @@ Changed 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and r ## Required -- [Plan: watch worker](/quest/m3/plan-watch-worker.md) - picks the worker model and rewrites this quest +- [Plan: watch worker](/quest/m4/plan-watch-worker.md) - picks the worker model and rewrites this quest diff --git a/quest/m3/webrtc/browser-only.md b/quest/m4/webrtc-browser-only.md similarity index 87% rename from quest/m3/webrtc/browser-only.md rename to quest/m4/webrtc-browser-only.md index ac4db052d24..3defcf9687f 100644 --- a/quest/m3/webrtc/browser-only.md +++ b/quest/m4/webrtc-browser-only.md @@ -17,10 +17,10 @@ PSK. Inventory what Chrome, Firefox, and Safari offer and accept (cipher suites, extensions, SDP attributes, codecs' packetizers) against what the vendored code still carries, cite the evidence for each removal in the PR, and keep what a browser negotiates even if MoQ doesn't call it yet. The -epic's settled removals (its "What goes" list, such as G.7xx, DTMF, and +[WebRTC epic](/quest/m3/webrtc/README.md)'s settled removals (its "What goes" list, such as G.7xx, DTMF, and comfort noise) and non-goals (TURN, DTLS 1.3) stay as decided: this trim only removes more, never restores. API that moq-webrtc never calls but a browser could need stays for -[the dead-API trim](/quest/m3/webrtc/dead-api.md). +[the dead-API trim](/quest/m4/webrtc-dead-api.md). Public API: none. Wire: none toward browsers. diff --git a/quest/m3/webrtc/clock.md b/quest/m4/webrtc-clock.md similarity index 100% rename from quest/m3/webrtc/clock.md rename to quest/m4/webrtc-clock.md diff --git a/quest/m3/webrtc/dead-api.md b/quest/m4/webrtc-dead-api.md similarity index 92% rename from quest/m3/webrtc/dead-api.md rename to quest/m4/webrtc-dead-api.md index 4153d888898..e5fc6185791 100644 --- a/quest/m3/webrtc/dead-api.md +++ b/quest/m4/webrtc-dead-api.md @@ -19,4 +19,4 @@ Public API: none. Wire: none. ## Required -- [The stack takes kio clock time](/quest/m3/webrtc/clock.md) - settles the time seams +- [The stack takes kio clock time](/quest/m4/webrtc-clock.md) - settles the time seams diff --git a/quest/m3/webrtc/sctp-feature.md b/quest/m4/webrtc-sctp-feature.md similarity index 86% rename from quest/m3/webrtc/sctp-feature.md rename to quest/m4/webrtc-sctp-feature.md index 02b8c14253b..8be9ab15e64 100644 --- a/quest/m3/webrtc/sctp-feature.md +++ b/quest/m4/webrtc-sctp-feature.md @@ -10,7 +10,7 @@ build has no SCTP code in its tree. Planned 2026-11-07 (maintainer, /quest-complete), a follow-up of [moq-dev/moq#8163](https://github.com/moq-dev/moq/pull/8163), which vendors -the stack but still compiles SCTP into every build. The epic's settled +the stack but still compiles SCTP into every build. The [WebRTC epic](/quest/m3/webrtc/README.md)'s settled decision ("SCTP stays ... but compiles only with the data-channel feature") gets this owner. Gate the modules and the str0m paths that reach them, and add a check, in the feature matrix `just rs features` runs, that a diff --git a/quest/m3/webrtc/sdp-bandwidth.md b/quest/m4/webrtc-sdp-bandwidth.md similarity index 100% rename from quest/m3/webrtc/sdp-bandwidth.md rename to quest/m4/webrtc-sdp-bandwidth.md diff --git a/quest/m4/workflow-audit-followups.md b/quest/m4/workflow-audit-followups.md index a4f23bf41f9..512d5e1505c 100644 --- a/quest/m4/workflow-audit-followups.md +++ b/quest/m4/workflow-audit-followups.md @@ -12,7 +12,6 @@ offline zizmor run the merge gate already does. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from [moq-dev/moq#7317](https://github.com/moq-dev/moq/pull/7317), which finished the workflow security linting quest, and [moq-dev/moq#7202](https://github.com/moq-dev/moq/pull/7202). -Milestone m2 (user: "use recommended for any CI/flakes"). [moq-dev/moq#7317](https://github.com/moq-dev/moq/pull/7317) runs `zizmor --offline` in `just gh check`; the online audits need the network and a GitHub token, so they belong in Nightly next to @@ -29,7 +28,7 @@ caught by hand. actionlint 1.7.12 (the pinned dev shell's) already fails `just gh check` on both main and `release` for a duplicated key in a workflow job and in a local action's `action.yml` (checked 2026-10-31), so the gate would have refused them. The rest of the tracked YAML is not -workflow code; the [formatter migration](/quest/m4/formatter-defaults.md) will +workflow code; the [formatter migration](/quest/m5/formatter-defaults.md) will parse it. Public API: none. Wire: none. diff --git a/quest/m5/README.md b/quest/m5/README.md index 39cc0a6f982..91e5217ee3c 100644 --- a/quest/m5/README.md +++ b/quest/m5/README.md @@ -9,6 +9,13 @@ speculative feature or study with no named consumer yet. ## Plan +Decided 2026-11-13 (maintainer, /quest-plan): this milestone requires the +[release cut](/quest/m3/release/cut.md), so until it lands spare capacity goes +to m0 to m3 and then to [m9](/quest/m9/README.md)'s lab work (rejected: +requiring m0, m1, and m2, which lifts only when they empty, and m0 rarely +does; also gating m3, which holds the cut). The cut's PR deletes the cut +quest, which lifts this gate. + A quest gated on the outside world has a condition quest beside it, as the root [questline](/quest/README.md) describes. A speculative feature or study with no consumer needs no condition quest: it parks here rather than in m3 or m4 @@ -33,3 +40,14 @@ in this repository and in moq.pro's plans, which kept the io_uring tuning quests (fleet relays on io_uring workers), broadcast fairness, and the transcode and NVENC quests (production transcoding). Re-plan one when a consumer asks. + +Parked 2026-11-12 (maintainer, /quest-audit before the cut): 29 quests the +audit judged not worth their weight (no consumer in this repository or the +downstream plans, and no bug in shipped code) are kept here instead of +deleted (rejected: deleting them; leaving them in m3 and m4). Each says so in +its Plan. The tuning experiments that touch no published surface moved to +the lab, [m9](/quest/m9/README.md). + +## Required + +- [Release cut](/quest/m3/release/cut.md) - the cut lands before this milestone's work starts diff --git a/quest/m4/aac-sbr-config.md b/quest/m5/aac-sbr-config.md similarity index 100% rename from quest/m4/aac-sbr-config.md rename to quest/m5/aac-sbr-config.md diff --git a/quest/m3/advertise-credit-timeout.md b/quest/m5/advertise-credit-timeout.md similarity index 76% rename from quest/m3/advertise-credit-timeout.md rename to quest/m5/advertise-credit-timeout.md index fa852968b8e..9b21cfafec8 100644 --- a/quest/m3/advertise-credit-timeout.md +++ b/quest/m5/advertise-credit-timeout.md @@ -15,14 +15,16 @@ Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 [moq-dev/moq#7414](https://github.com/moq-dev/moq/pull/7414): - Bound a stuck advertisement under a zero-credit peer: - - ✅ Quest in m2. Withdrawal already frees it after [moq-dev/moq#7414](https://github.com/moq-dev/moq/pull/7414), so the + - ✅ Plan it as a quest. Withdrawal already frees it after [moq-dev/moq#7414](https://github.com/moq-dev/moq/pull/7414), so the remaining cost is one buffered request per namespace, not a leak. - Drop it. +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. + [moq-dev/moq#7414](https://github.com/moq-dev/moq/pull/7414) (merged) buffers each message and starts `ADVERTISE_TIMEOUT` -(`rs/moq-net/src/ietf/publisher.rs:563` at be39188000, 5 s) only once the bytes -reach the transport (`:3749`); its `NamespacePending::Writing` state has no deadline, and the PR +(`rs/moq-net/src/ietf/publisher.rs:636`, 5 s) only once the bytes +reach the transport (`:3959-3962`); its `NamespacePending::Writing` state (`:8396`) still has no deadline, and the PR notes "send backpressure alone does not start answer timeouts or retries". Build on it: @@ -38,4 +40,4 @@ Public API: none. Wire: none. ## Related -- [One request lease](/quest/m3/ietf-request-lease.md) - other request lifetimes on the same streams +- [One request lease](/quest/m4/ietf-request-lease.md) - other request lifetimes on the same streams diff --git a/quest/m3/announce-revoke-reset.md b/quest/m5/announce-revoke-reset.md similarity index 88% rename from quest/m3/announce-revoke-reset.md rename to quest/m5/announce-revoke-reset.md index afe79144f54..4fb872cd39b 100644 --- a/quest/m3/announce-revoke-reset.md +++ b/quest/m5/announce-revoke-reset.md @@ -9,11 +9,7 @@ ending with a FIN. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: - -- m0 -- m1 -- ✅ m2 +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. From [moq-dev/moq#7703](https://github.com/moq-dev/moq/pull/7703)'s follow-ups. That PR (merged) resets a lapsed announce with EXPIRED_AUTH_TOKEN (0x7) diff --git a/quest/m4/api-diff-gst.md b/quest/m5/api-diff-gst.md similarity index 100% rename from quest/m4/api-diff-gst.md rename to quest/m5/api-diff-gst.md diff --git a/quest/m4/apple-sink-thread.md b/quest/m5/apple-sink-thread.md similarity index 100% rename from quest/m4/apple-sink-thread.md rename to quest/m5/apple-sink-thread.md diff --git a/quest/m5/archive-time-followups.md b/quest/m5/archive-time-followups.md index 3b7fb8dc68c..5b137ee3e68 100644 --- a/quest/m5/archive-time-followups.md +++ b/quest/m5/archive-time-followups.md @@ -48,4 +48,4 @@ Public API: `track::Fill` gains a lookup refusal; moq-ffi `MoqFill` gains ## Related -- [DVR rewind](/quest/m3/archive-dvr.md) - the consumer of time seeks +- [DVR rewind](/quest/m4/archive-dvr.md) - the consumer of time seeks diff --git a/quest/m4/askama-maintained.md b/quest/m5/askama-maintained.md similarity index 100% rename from quest/m4/askama-maintained.md rename to quest/m5/askama-maintained.md diff --git a/quest/m4/audio-visqol.md b/quest/m5/audio-visqol.md similarity index 100% rename from quest/m4/audio-visqol.md rename to quest/m5/audio-visqol.md diff --git a/quest/m4/bench-epoch-pin.md b/quest/m5/bench-epoch-pin.md similarity index 100% rename from quest/m4/bench-epoch-pin.md rename to quest/m5/bench-epoch-pin.md diff --git a/quest/m4/bench-latency-resolution.md b/quest/m5/bench-latency-resolution.md similarity index 91% rename from quest/m4/bench-latency-resolution.md rename to quest/m5/bench-latency-resolution.md index 1a561d1f34d..1f6a739f588 100644 --- a/quest/m4/bench-latency-resolution.md +++ b/quest/m5/bench-latency-resolution.md @@ -8,11 +8,8 @@ percentiles instead of rounding to the same whole millisecond. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: - -- m0 -- m1 -- ✅ m2 +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs). +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. `rs/moq-bench/src/stats.rs` records latency with `as_millis()` into one-millisecond buckets (`LATENCY_BUCKETS`, 60,001 buckets through 60 s). diff --git a/quest/m4/bench-transport-sweep.md b/quest/m5/bench-transport-sweep.md similarity index 69% rename from quest/m4/bench-transport-sweep.md rename to quest/m5/bench-transport-sweep.md index de4cc8ccd26..00902477d16 100644 --- a/quest/m4/bench-transport-sweep.md +++ b/quest/m5/bench-transport-sweep.md @@ -8,14 +8,18 @@ framing, flow control) shows up as its own series in the regression table. ## Plan -Decided 2026-10-21 with the maintainer: m2. `rs/moq-net/benches/session.rs` +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. + +`rs/moq-net/benches/session.rs` sweeps `VERSIONS` (lite-06, lite-07-wip, moq-transport-22) over the in-memory transport only. Add the transports a production session uses, moq-quic and qmux at least, keeping each run deterministic: drive QUIC over an in-memory datagram link rather than real sockets if the noise demands it. Keep existing bench IDs stable so their history carries over; new -transports add series rather than renaming old ones. This bench owns the -qmux series in the regression table; the stream transport micro-bench +transports add series rather than renaming old ones. Transport costs alone +already have Nightly series: qmux's `receive` bench and moq-quic's `streams`, +`receive`, and `endpoint` benches (`sh/gh/nightly.sh`); this sweep would add +only the session over each transport. The stream transport micro-bench (`rs/qmux/examples/bench.rs`) stays a local profiling tool (decided 2026-10-22, maintainer). diff --git a/quest/m4/bench-uring-pool.md b/quest/m5/bench-uring-pool.md similarity index 63% rename from quest/m4/bench-uring-pool.md rename to quest/m5/bench-uring-pool.md index a71321e4a8a..6ed7f964480 100644 --- a/quest/m4/bench-uring-pool.md +++ b/quest/m5/bench-uring-pool.md @@ -1,10 +1,11 @@ -# [XS] The relay bench reports io_uring pool size and RSS +# [XS] The relay bench reports io_uring pool size ## Goal The `bench/` relay summary reports each run's io_uring buffer pool gauges -(receive and send, total and in use) and the relay's RSS, so a regression in -pool growth shows up in normal bench runs. +(receive and send, total and in use), so a regression in pool growth shows +up in normal bench runs. The summary already reports the relay's RSS +(`bench/run.sh`, the `RSS-MiB` column). ## Plan diff --git a/quest/m5/browser-mesh.md b/quest/m5/browser-mesh.md index 5edfc2fcca1..907c7629bde 100644 --- a/quest/m5/browser-mesh.md +++ b/quest/m5/browser-mesh.md @@ -15,7 +15,7 @@ might forward too, honestly it would be lit to have P2P meshes for browsers on the local network so please plan that." - ✅ The forwarding engine is the Rust router, run in the tab through the - [Rust-to-TypeScript port](/quest/m5/rs2ts-js.md) (rejected: handwriting + [Rust model in the browser](/quest/m5/rs2ts-js.md), WebAssembly first since 2026-11-09, (rejected: handwriting the forwarding rule in `js/net`, which was feasibility, route requests, holds, and echoes when this was decided and is paths, withdrawals, holds, and the doubt since the @@ -28,7 +28,7 @@ on the local network so please plan that." This lifts three statements once it lands: the P2P epic's "A browser never carries another peer's traffic", the -[cluster routing](/quest/m2/cluster-routing/README.md) plan's browser tabs +[cluster routing](/quest/m3/cluster-routing/README.md) plan's browser tabs keeping `proxy` off, and `js/net`'s non-transit tab. Until then each stays true. @@ -60,5 +60,5 @@ relay does. ## Required -- [Ship JS-only @moq/net from Rust](/quest/m5/rs2ts-js.md) - the Rust router and model running in the tab +- [@moq/net runs the Rust model in the browser](/quest/m5/rs2ts-js.md) - the Rust router and model running in the tab - [P2P](/quest/m4/p2p/README.md) - direct browser sessions, their signaling, and `Peers` diff --git a/quest/m5/carla-bridge.md b/quest/m5/carla-bridge.md index 496a2c14a88..250901a4757 100644 --- a/quest/m5/carla-bridge.md +++ b/quest/m5/carla-bridge.md @@ -55,7 +55,7 @@ probe opens the camera's own backend with a CPU picture. ## Required -- [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - the AMD encoder this exercises +- [Vulkan Video encode on AMD](/quest/m5/vulkan-encode.md) - the AMD encoder this exercises ## Related diff --git a/quest/m4/chromium-opus-timestamps.md b/quest/m5/chromium-opus-timestamps.md similarity index 92% rename from quest/m4/chromium-opus-timestamps.md rename to quest/m5/chromium-opus-timestamps.md index 45975da2420..61e11d05485 100644 --- a/quest/m4/chromium-opus-timestamps.md +++ b/quest/m5/chromium-opus-timestamps.md @@ -26,9 +26,10 @@ and closed by the maintainer for that reason. Decided 2026-11-04 (maintainer): file the Chromium bug ([chromium#581458361](https://issues.chromium.org/issues/581458361), Blink>Media>WebCodecs, -submitted by the maintainer) and no spec issue; track it here in m4, since +submitted by the maintainer) and no spec issue; track it here, since push-to-talk works around it by gating silence before the encoder and stamping chunks from its own inputs, so nothing depends on this. +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. ## Waiting diff --git a/quest/m4/cli-catalog-updates.md b/quest/m5/cli-catalog-updates.md similarity index 85% rename from quest/m4/cli-catalog-updates.md rename to quest/m5/cli-catalog-updates.md index 59cdf521dd1..a25f0c35c44 100644 --- a/quest/m4/cli-catalog-updates.md +++ b/quest/m5/cli-catalog-updates.md @@ -10,7 +10,9 @@ resuming mid-update) with stock tools instead of hand-built clients. ## Plan -Decided 2026-11-03 (maintainer, `/quest-plan`): m2. A relay-replacement +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. + +Planned 2026-11-03 (maintainer, `/quest-plan`). A relay-replacement replay of catalog resume had to build scratch publisher clients because the CLI publishes one catalog per broadcast (`rs/moq-cli/src/publish.rs`, `PublishCatalog`). diff --git a/quest/m4/corpus-fixtures.md b/quest/m5/corpus-fixtures.md similarity index 84% rename from quest/m4/corpus-fixtures.md rename to quest/m5/corpus-fixtures.md index fa2990abe1b..08e695da1f9 100644 --- a/quest/m4/corpus-fixtures.md +++ b/quest/m5/corpus-fixtures.md @@ -11,8 +11,8 @@ directory, so a contributor finds them in one place. ## Plan -Planned 2026-11-08 (maintainer, approved follow-up). Milestone: m4, tooling -nobody waits on. +Planned 2026-11-08 (maintainer, approved follow-up). +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. - Move the file and update its readers: `rs/moq-auth/src/corpus.rs` (`include_str!` and the module doc) and `js/auth/src/corpus.test.ts`. diff --git a/quest/m4/crossbeam-pin-fence.md b/quest/m5/crossbeam-pin-fence.md similarity index 100% rename from quest/m4/crossbeam-pin-fence.md rename to quest/m5/crossbeam-pin-fence.md diff --git a/quest/m5/e2ee/README.md b/quest/m5/e2ee/README.md index 8c0a0cfc7ed..3de4f4b064e 100644 --- a/quest/m5/e2ee/README.md +++ b/quest/m5/e2ee/README.md @@ -23,7 +23,7 @@ The contract is [draft-lcurley-moq-e2ee](/drafts/draft-lcurley-moq-e2ee.md), pro - An encrypting publisher refuses a shared or explicit epoch (maintainer, 2026-10-08). Two producers handed one epoch would derive the same keys and repeat every nonce. Both cores produce only under an epoch they minted themselves; consuming still binds a discovered one. Rejected: a per-instance salt. State the rule in the draft's epoch section. - The epoch is untrusted and unauthenticated. A wrong epoch fails authentication and a withheld one denies service; neither can make a nonce repeat, because only the publisher instance chooses what it encrypts under. This is the same trust a cache needs to serve the right instance. First-party plaintext publishers already mint the same route epoch. - Nothing in the path says the bytes are encrypted. The format after decryption (`meeting.hang`) is inside the opaque name; a plaintext player, exporter, or matcher that opens a protected broadcast finds no catalog it can read and fails with its usual typed refusal, the same as for any format it does not support. -- One 32-byte secret authorizes the whole broadcast; HKDF-SHA-256 derives separate AES-128-GCM keys for each physical track and for grouped-frame versus datagram domains. A grouped frame uses the 96-bit nonce `uint64_be(group) || uint32_be(frame)`; a datagram uses its sequence with frame zero under the datagram domain. Empty AAD: every immutable end-to-end field is in the HKDF info or the nonce. +- One 32-byte secret authorizes the whole broadcast; HKDF-SHA-256 derives separate AES-128-GCM keys for each physical track and for grouped-frame versus datagram domains. A grouped frame uses a 96-bit nonce of a 64-bit group identity and `uint32_be(frame)`. The group identity becomes the group's start timestamp, not its sequence, once [the E2EE nonce binds the group start](/quest/m2/time-addressing/e2ee.md) lands; until then the draft and both cores use `uint64_be(group)`. A datagram uses its sequence with frame zero under the datagram domain. Empty AAD: every immutable end-to-end field is in the HKDF info or the nonce. - Within an instance, group and datagram sequences are allocated monotonically and frames are numbered in write order, which is the whole reuse rule. There is no ciphertext retention or retransmission API: relays forward bytes unchanged, and an application that needs to resend encrypts nothing twice because it never gets the same identity twice. - Per-key limits are `2^24` AEAD invocations (failed opens included) and `2^36` plaintext bytes. The generation handle owns each key's accounting, shared by every track and subscription through it and never evicted while the application holds it; binding the epoch again starts fresh, since remembering retired epochs would be state a relay grows (maintainer, 2026-10-24). Rejected: a credential-level registry of live generations, and refusing to re-bind a released epoch. Datagram plaintext is capped at `1200 - 24 - 16` bytes against the widest moq-lite header; a publisher cannot see the Subscribe ID each hop encodes, so the library does not pretend to budget it. diff --git a/quest/m5/e2ee/cli.md b/quest/m5/e2ee/cli.md index 42b9ce3b986..18a7215c31e 100644 --- a/quest/m5/e2ee/cli.md +++ b/quest/m5/e2ee/cli.md @@ -8,8 +8,8 @@ Existing plaintext commands remain the default; an explicit application credenti ## Plan -Make `moq-crypto` publishable with this CLI consumer; the crate currently sets -`publish = false` while it is workspace-only. +`moq-crypto` is already publishable: its manifest (`rs/moq-crypto/Cargo.toml`) +no longer sets `publish = false`. - Hand the [trait-generic CLI paths](/quest/m5/net-wrap/cli.md) an encrypted broadcast for publication and playback. Keep codec, capture, playback, and synchronization logic outside the crypto layer. - Accept the credential through a dedicated file descriptor, otherwise-unused stdin, or a permission-checked file (`0600` on Unix), never a command argument or environment variable. Redact errors and tracing, zeroize owned secret bytes, and document shell-safe invocation. diff --git a/quest/m5/e2ee/js-undeclared-names.md b/quest/m5/e2ee/js-undeclared-names.md index e2be23f53dc..fcac24009c1 100644 --- a/quest/m5/e2ee/js-undeclared-names.md +++ b/quest/m5/e2ee/js-undeclared-names.md @@ -13,7 +13,7 @@ broadcast the same way. Decided with the maintainer on 2026-10-24: - One quest, both halves; amended 2026-11-03 (maintainer): the `@moq/net` - half moved to [its own quest](/quest/m3/net-broadcast-requests.md), which + half moved to [its own quest](/quest/m4/net-broadcast-requests.md), which plaintext users (Kixel's T3 host) need first, and this keeps the `@moq/crypto` half. - `@moq/crypto` mirrors the Rust handler exactly: it opens each request's @@ -34,5 +34,5 @@ Public API: a request handler on `@moq/crypto`'s broadcast producer. Wire: none. ## Required -- [`@moq/net` broadcasts expose their requests](/quest/m3/net-broadcast-requests.md) - the public handler this wraps; split out 2026-11-03 so plaintext users get it first +- [`@moq/net` broadcasts expose their requests](/quest/m4/net-broadcast-requests.md) - the public handler this wraps; split out 2026-11-03 so plaintext users get it first - [JS mirror](/quest/m5/net-wrap/js.md) - the `@moq/net` interfaces the `@moq/crypto` wrappers implement diff --git a/quest/m3/errqueue-truncation.md b/quest/m5/errqueue-truncation.md similarity index 69% rename from quest/m3/errqueue-truncation.md rename to quest/m5/errqueue-truncation.md index 538b046e39f..fab62c36388 100644 --- a/quest/m3/errqueue-truncation.md +++ b/quest/m5/errqueue-truncation.md @@ -8,7 +8,9 @@ moq-sock's `MSG_ERRQUEUE` reader (`rs/moq-sock/src/udp/linux.rs`) handles `MSG_C Found by [moq-dev/moq#7543](https://github.com/moq-dev/moq/pull/7543), which made the Windows receive path keep a datagram with truncated control data, matching what [moq-dev/moq#7394](https://github.com/moq-dev/moq/pull/7394) did for `recvmmsg`, `recvmsg`, and Apple's `recvmsg_x`. The error-queue reader is a different path and still turns `MSG_CTRUNC` into `io::ErrorKind::InvalidData` (about line 55). It only matters if error-queue metadata ever grows past `ERR_CMSG_LEN`, so this is defensive. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, matching the receive paths. Walk the cmsgs that fit (the iterator is already bounded by the returned control length), return the decoded entry if one is complete, and warn once through a `Once` as `decode_recv` in `unix.rs` does. Extend the existing mocked `MSG_ERRQUEUE` test (`LinuxError::decode`, about line 309) with a truncated case. +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. + +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): match the receive paths. Walk the cmsgs that fit (the iterator is already bounded by the returned control length), return the decoded entry if one is complete, and warn once through a `Once` as `decode_recv` in `unix.rs` does. Extend the existing mocked `MSG_ERRQUEUE` test (`LinuxError::decode`, about line 309) with a truncated case. Public API: none. Wire: none. diff --git a/quest/m5/ffi-e2ee.md b/quest/m5/ffi-e2ee.md index 84d4632e6dc..a0b680b7c6a 100644 --- a/quest/m5/ffi-e2ee.md +++ b/quest/m5/ffi-e2ee.md @@ -9,11 +9,14 @@ and typed authentication failures as Rust and JS. ## Plan -Decided 2026-10-28 with the maintainer (moq.pro T3 planning interview): this +Decided 2026-10-28 with the maintainer (a downstream planning interview): this is the quest the [encrypted consumer](/quest/m5/e2ee/consumer.md) defers FFI -to, adding both halves. The named consumer is moq.pro's -[T3 Code over MoQ](https://github.com/moq-dev/moq.pro/blob/main/quest/m3/t3/README.md) -mobile app, which receives its credential in a pairing link. +to, adding both halves. + +No named consumer remains (checked 2026-11-12, /quest-audit). It was planned +for a downstream application's mobile client, which has since decided on no +end-to-end encryption and on a transport that does not go through moq-ffi. +It stays parked here until a binding user asks for encrypted broadcasts. Wrap `moq-crypto`'s producer and consumer entry points without new crypto in the binding layer, and keep the encrypting publisher's refusal of a shared or diff --git a/quest/m5/ffi-stream.md b/quest/m5/ffi-stream.md index 02d4bb02700..a38c19447b1 100644 --- a/quest/m5/ffi-stream.md +++ b/quest/m5/ffi-stream.md @@ -10,14 +10,17 @@ and writes the stream in-process. ## Plan -Decided 2026-10-28 with the maintainer (moq.pro T3 planning interview): the -named consumer is moq.pro's -[T3 Code over MoQ](https://github.com/moq-dev/moq.pro/blob/main/quest/m3/t3/README.md) -line, whose React Native app reaches its host through a native module over -these bindings instead of the WebSocket fallback, so a lossy mobile link keeps -QUIC's per-stream recovery. - -Wrap the [Rust surface](/quest/m3/stream/rust.md) in `moq-ffi` with the same +Decided 2026-10-28 with the maintainer (a downstream planning interview): +planned for a downstream application's React Native app, to reach its host +through a native module over these bindings instead of the WebSocket +fallback, so a lossy mobile link keeps QUIC's per-stream recovery. + +No named consumer remains (checked 2026-11-12, /quest-audit): that +application's mobile module now wraps a WebTransport binding rather than +moq-ffi, and carries its traffic in groups rather than byte streams. It stays +parked here until a binding user asks for byte streams. + +Wrap the [Rust surface](/quest/m1/stream/rust.md) in `moq-ffi` with the same path layout and priority knob, and surface it through the generated Swift and Kotlin packages. Use an async read and write shape that matches how the bindings already expose frames; report the choice. Add a Kotlin or Swift @@ -25,5 +28,5 @@ binding test against a JS peer to the interop suite. ## Required -- [Rust byte streams](/quest/m3/stream/rust.md) - the surface these bindings wrap -- [JS byte streams](/quest/m3/stream/js.md) - the peer the interop test runs against +- [Rust byte streams](/quest/m1/stream/rust.md) - the surface these bindings wrap +- [JS byte streams](/quest/m4/stream-js.md) - the peer the interop test runs against diff --git a/quest/m4/fill-wait-cancel-test.md b/quest/m5/fill-wait-cancel-test.md similarity index 86% rename from quest/m4/fill-wait-cancel-test.md rename to quest/m5/fill-wait-cancel-test.md index 024533d9a27..876d71ad568 100644 --- a/quest/m4/fill-wait-cancel-test.md +++ b/quest/m5/fill-wait-cancel-test.md @@ -14,7 +14,9 @@ has delivered or failed before sending a FETCH, and the wait also ends when no reader wants the group or the track is aborted. That last part has no test: the PR found no way for a test to observe a released task. -Decided 2026-10-29 (maintainer): m2. Add the smallest test-only hook that can +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. + +Decided 2026-10-29 (maintainer): add the smallest test-only hook that can see the wait end (for example a counter of live waits in `rs/moq-net/src/ietf/subscriber.rs`, compiled only for tests), then cover a cancelled reader and an aborted track. Rejected the same day: a time limit diff --git a/quest/m4/fmp4-byte-chunk-perf.md b/quest/m5/fmp4-byte-chunk-perf.md similarity index 100% rename from quest/m4/fmp4-byte-chunk-perf.md rename to quest/m5/fmp4-byte-chunk-perf.md diff --git a/quest/m4/formatter-defaults.md b/quest/m5/formatter-defaults.md similarity index 95% rename from quest/m4/formatter-defaults.md rename to quest/m5/formatter-defaults.md index af3fcaa64e5..5a33fa74070 100644 --- a/quest/m4/formatter-defaults.md +++ b/quest/m5/formatter-defaults.md @@ -57,7 +57,9 @@ Decided with the maintainer on 2026-11-03: - YAML and Markdown: keep their earlier separate schedule / ✅ fold both into this migration, avoiding repeated reformats. Their formatter choice is now part of the Bun evaluation; retiring remark remains settled. -- Milestone: m3 / ✅ m2, moved here at the maintainer's request. + +Moved to m5 2026-11-12 (maintainer, /quest-audit): blocked on a stable +`bun lint`, an outside release. ## Required @@ -66,4 +68,4 @@ Decided with the maintainer on 2026-11-03: ## Related -- [Typecheck with bun check](/quest/m3/bun-check.md) - the same Bun toolchain +- [Typecheck with bun check](/quest/m4/bun-check.md) - the same Bun toolchain diff --git a/quest/m4/gpu-encoder-death-test.md b/quest/m5/gpu-encoder-death-test.md similarity index 78% rename from quest/m4/gpu-encoder-death-test.md rename to quest/m5/gpu-encoder-death-test.md index 852fe5db9c8..71581c9f2f8 100644 --- a/quest/m4/gpu-encoder-death-test.md +++ b/quest/m5/gpu-encoder-death-test.md @@ -11,10 +11,10 @@ under a fresh track name on the new broadcast with no carried-over failure. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7371](https://github.com/moq-dev/moq/pull/7371): m2, as -recommended. [moq-dev/moq#7371](https://github.com/moq-dev/moq/pull/7371) (merged) moved encoder failure into +[moq-dev/moq#7371](https://github.com/moq-dev/moq/pull/7371). [moq-dev/moq#7371](https://github.com/moq-dev/moq/pull/7371) (merged) moved encoder failure into each publication's slot, so a new network broadcast starts clean, and covered that only with unit tests and fake encoders. +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. - Put it in `test/interop/clients/js/` beside `publish-ui.browser.ts`. Inject the failure through the encoder (closing the `VideoEncoder` or a diff --git a/quest/m3/gpu-release.md b/quest/m5/gpu-release.md similarity index 84% rename from quest/m3/gpu-release.md rename to quest/m5/gpu-release.md index d0e2260f4bf..30f237d0ca9 100644 --- a/quest/m3/gpu-release.md +++ b/quest/m5/gpu-release.md @@ -15,8 +15,8 @@ GPU capacity and health reporting remains a separate additive follow-up. ## Required -- [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - AMD proof -- [VA-API Vulkan import](/quest/m3/vaapi-vulkan-import.md) - Intel proof +- [Vulkan Video encode on AMD](/quest/m5/vulkan-encode.md) - AMD proof +- [VA-API Vulkan import](/quest/m5/vaapi-vulkan-import.md) - Intel proof ## Related diff --git a/quest/m4/h265-layered-aud.md b/quest/m5/h265-layered-aud.md similarity index 88% rename from quest/m4/h265-layered-aud.md rename to quest/m5/h265-layered-aud.md index bf5cef302e8..0d4e7b42fab 100644 --- a/quest/m4/h265-layered-aud.md +++ b/quest/m5/h265-layered-aud.md @@ -9,8 +9,9 @@ that names the cause, rather than splitting each view into its own frame. ## Plan Planned 2026-11-08 (maintainer, approved follow-up of -[moq-dev/moq#8199](https://github.com/moq-dev/moq/pull/8199)). Milestone: -m4, a feature with no named consumer yet. +[moq-dev/moq#8199](https://github.com/moq-dev/moq/pull/8199)). +Moved to m5 2026-11-12 (maintainer, /quest-audit): a feature with no named +consumer. x265 4.2 with `--num-views 2 --aud` writes a layer-0 AUD between the base and enhancement views. `rs/moq-mux/src/codec/h265/split.rs` closes the open diff --git a/quest/m4/kt-version-catalog.md b/quest/m5/kt-version-catalog.md similarity index 100% rename from quest/m4/kt-version-catalog.md rename to quest/m5/kt-version-catalog.md diff --git a/quest/m5/link-quality.md b/quest/m5/link-quality.md index d3c801377f9..8bc39e2b0f5 100644 --- a/quest/m5/link-quality.md +++ b/quest/m5/link-quality.md @@ -14,7 +14,7 @@ Decided 2026-10-01: the wire carries one additive metric and each node computes its own link costs. Decided 2026-10-08: m2, and Related rather than Required for the -[cluster routing line](/quest/m2/cluster-routing/README.md): static costs are +[cluster routing line](/quest/m3/cluster-routing/README.md): static costs are the default and measured cost is opt-in, so the line ships without it. - Derive the cost from what QUIC already measures (smoothed RTT, loss); an diff --git a/quest/m5/lite07-mesh.md b/quest/m5/lite07-mesh.md index 2f89d919292..5a680be302d 100644 --- a/quest/m5/lite07-mesh.md +++ b/quest/m5/lite07-mesh.md @@ -13,7 +13,7 @@ this quest and every `Required` entry that links it. moq.pro's lite-07 rollout waits on [Finalize moq-lite-07](/quest/m2/lite07-finalize.md), so this condition cannot clear before a release carries the final version. -Decided 2026-10-05: [Cluster routing](/quest/m2/cluster-routing/README.md)'s +Decided 2026-10-05: [Cluster routing](/quest/m3/cluster-routing/README.md)'s route layer lands in lite-07, so the final lite-07 carries ROUTE and path-less ANNOUNCE instead of native hop lists (a temporary lite-06 compatibility chain rides until lite-08), and loses the `Hop Base`/`Hop Keep` diff --git a/quest/m3/mesh/guide.md b/quest/m5/mesh-guide.md similarity index 100% rename from quest/m3/mesh/guide.md rename to quest/m5/mesh-guide.md diff --git a/quest/m3/mesh/js-enroll.md b/quest/m5/mesh-js-enroll.md similarity index 54% rename from quest/m3/mesh/js-enroll.md rename to quest/m5/mesh-js-enroll.md index 1347a2fdde1..c6a8bb7614a 100644 --- a/quest/m3/mesh/js-enroll.md +++ b/quest/m5/mesh-js-enroll.md @@ -10,14 +10,17 @@ implements the same contract. ## Plan -Planned 2026-11-03 (maintainer) from Kixel's `@kixel/mesh` -([moq-dev/kixel.ai#16](https://github.com/moq-dev/kixel.ai/pull/16)), which -implements this contract in Kixel today; once this lands, `@kixel/mesh` -becomes a thin wrapper over it. Port `@kixel/mesh`'s hardening where it is +Planned 2026-11-03 (maintainer) from a downstream application's JS mesh +client, which implemented this contract then. Port its hardening where it is generic: https except on loopback, no redirects, bounded time and reply size, and every reply field checked. Test against `moq auth serve` the way `rs/moq-auth/src/mesh/contract_tests.rs` does. +Moved to m5 2026-11-12 (maintainer, /quest-audit): that client has since +moved to a certificate enrollment reply, which `moq_auth::mesh::Client`'s +token-only `Enrollment` (`rs/moq-auth/src/mesh/mod.rs`) does not carry, so it +no longer wraps this contract and nothing else asks for a JS client. + ## Required -- [`@moq/crypto` loads a stored signing key](/quest/m3/crypto-key-load.md) - the endpoint key this proves +- [`@moq/crypto` loads a stored signing key](/quest/m4/crypto-key-load.md) - the endpoint key this proves diff --git a/quest/m5/miri-genmc.md b/quest/m5/miri-genmc.md index 311958f7d01..3995bfc0360 100644 --- a/quest/m5/miri-genmc.md +++ b/quest/m5/miri-genmc.md @@ -46,4 +46,4 @@ Public API: none. Wire: none. ## Related -- [Drop kio's epoch fence](/quest/m4/crossbeam-pin-fence.md) - the ordering in kio's log that only a weak-memory check can confirm +- [Drop kio's epoch fence](/quest/m5/crossbeam-pin-fence.md) - the ordering in kio's log that only a weak-memory check can confirm diff --git a/quest/m4/mux-vr-ambisonic-fmp4.md b/quest/m5/mux-vr-ambisonic-fmp4.md similarity index 100% rename from quest/m4/mux-vr-ambisonic-fmp4.md rename to quest/m5/mux-vr-ambisonic-fmp4.md diff --git a/quest/m5/net-wrap/README.md b/quest/m5/net-wrap/README.md index 63d0557208f..b0163c0814f 100644 --- a/quest/m5/net-wrap/README.md +++ b/quest/m5/net-wrap/README.md @@ -15,7 +15,7 @@ Compression is not a wrapper: it moves into moq-net as a track flag moq-flate, so moq-flate is out of scope. Non-goals: -- A runtime seam: libraries poll and spawn nothing ([the remaining library spawns](/quest/m3/spawn-audit.md)). +- A runtime seam: libraries poll and spawn nothing ([the remaining library spawns](/quest/m4/spawn-audit.md)). - The transport seam: custom transports implement moq-net's `transport::poll` traits, including both native WebTransport backends. ## Plan diff --git a/quest/m4/nix-relay-peers-test.md b/quest/m5/nix-relay-peers-test.md similarity index 68% rename from quest/m4/nix-relay-peers-test.md rename to quest/m5/nix-relay-peers-test.md index f9884520cea..f9224ca67d5 100644 --- a/quest/m4/nix-relay-peers-test.md +++ b/quest/m5/nix-relay-peers-test.md @@ -11,17 +11,18 @@ root and two leaves from the module and sees them connect. Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from -[moq-dev/moq#7343](https://github.com/moq-dev/moq/pull/7343): m2, as -recommended. +[moq-dev/moq#7343](https://github.com/moq-dev/moq/pull/7343). + +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. [moq-dev/moq#7343](https://github.com/moq-dev/moq/pull/7343) made `services.moq-relay.cluster.rootUrl` a -`nullOr (nonEmptyListOf str)` (`nix/modules/moq-relay.nix:153`), joined into -`MOQ_CLUSTER_CONNECT` at `:380`, and verified it only by a manual `nix eval`. +`nullOr (nonEmptyListOf str)` (`nix/modules/moq-relay.nix:120` at 4bf501b535), joined into +`MOQ_CLUSTER_CONNECT` at `:365`, and verified it only by a manual `nix eval`. - Add an eval check beside `nix/runner/config-test.nix`, wired into - `flake.nix`'s `checks` (`:729` as of be39188000): `null` sets no `MOQ_CLUSTER_CONNECT`, one + `flake.nix`'s `checks` (`:750` at 4bf501b535): `null` sets no `MOQ_CLUSTER_CONNECT`, one and two URLs join with commas, `[ ]` fails the type check, and a leaf - without peers trips the assertion at `:219`. + without peers trips the assertion at `:216`. - A VM test (root plus two leaves) only if it stays cheap in CI; otherwise leave it out and say so. diff --git a/quest/m5/nvenc-cuda-arrays.md b/quest/m5/nvenc-cuda-arrays.md index f1e0ce192a0..061d497ba6a 100644 --- a/quest/m5/nvenc-cuda-arrays.md +++ b/quest/m5/nvenc-cuda-arrays.md @@ -20,5 +20,5 @@ Public API: possibly a CUDA-array frame variant in moq-video's GPU frame types. ## Related -- [NVENC reuses its input and output buffers](/quest/m5/nvenc-pool.md) - per-frame allocation on the encode side, distinct from this copy +- [NVENC reuses its input and output buffers](/quest/m9/nvenc-pool.md) - per-frame allocation on the encode side, distinct from this copy - [GPU CI](/quest/m3/gpu-ci.md) - the NVIDIA host this would validate on once its driver is new enough diff --git a/quest/m3/obs-studio/README.md b/quest/m5/obs-studio/README.md similarity index 95% rename from quest/m3/obs-studio/README.md rename to quest/m5/obs-studio/README.md index 2d503ee65f6..0754107e1e2 100644 --- a/quest/m3/obs-studio/README.md +++ b/quest/m5/obs-studio/README.md @@ -12,7 +12,7 @@ keeps shipping. This README holds the port and merge work: carry the plugin's output and service into obs-studio's tree in the shape its maintainers ask for, then land it upstream. Shape it around their answer to -[the condition](/quest/m3/obs-studio/maintainers.md), not ahead of it. +[the condition](/quest/m5/obs-studio/maintainers.md), not ahead of it. Agents may prepare the proposal and the code, but never post to obs-studio (issues, PRs, comments, forums) without the maintainer's explicit approval diff --git a/quest/m3/obs-studio/maintainers.md b/quest/m5/obs-studio/maintainers.md similarity index 91% rename from quest/m3/obs-studio/maintainers.md rename to quest/m5/obs-studio/maintainers.md index ed97963a108..d62f9950a56 100644 --- a/quest/m3/obs-studio/maintainers.md +++ b/quest/m5/obs-studio/maintainers.md @@ -14,6 +14,6 @@ this quest and every `Required` entry that links it. To check: ask the maintainer whether obs-studio has agreed, and record any constraints they set (dependency policy, scope, review owners) in the -[questline](/quest/m3/obs-studio/README.md) before deleting this quest. To +[questline](/quest/m5/obs-studio/README.md) before deleting this quest. To advance: offer the maintainer a drafted proposal covering what the output does, how it builds, and who maintains it. diff --git a/quest/m4/peer-grant-bound.md b/quest/m5/peer-grant-bound.md similarity index 100% rename from quest/m4/peer-grant-bound.md rename to quest/m5/peer-grant-bound.md diff --git a/quest/m3/pes-hold-open-gop.md b/quest/m5/pes-hold-open-gop.md similarity index 100% rename from quest/m3/pes-hold-open-gop.md rename to quest/m5/pes-hold-open-gop.md diff --git a/quest/m3/probe-padding.md b/quest/m5/probe-padding.md similarity index 92% rename from quest/m3/probe-padding.md rename to quest/m5/probe-padding.md index 25a0ec4c82b..0d4bda6ced0 100644 --- a/quest/m3/probe-padding.md +++ b/quest/m5/probe-padding.md @@ -14,15 +14,17 @@ probe to the larger of their send target and the peer's PROBE target. ## Plan -Parked 2026-11-06 (maintainer, /quest-unblock): ✅ moved from the m2 probe -epic to m3, because the maintainer doesn't think padding is the right -approach. Rethink how a browser raises its estimate before building this; the -padding overhead budget's denominator was left open for that rethink. Its draft, +Moved to m5 2026-11-12 (maintainer, /quest-audit): parked for a rethink with no named consumer. + +Parked 2026-11-06 (maintainer, /quest-unblock): the maintainer doesn't think +padding is the right approach. Rethink how a browser raises its estimate +before building this; the padding overhead budget's denominator was left open +for that rethink. Its draft, [moq-dev/moq#7390](https://github.com/moq-dev/moq/pull/7390), was closed; its send target and ALR schedule remain at `refs/pull/7390/head` (ca3813d7e1). Decided with the maintainer on 2026-10-17 (see the -[epic](/quest/m2/probe/README.md)): +[epic](/quest/m3/probe/README.md)): - **Padding is the browser's sender.** Native sessions retransmit in-flight data instead (native probing, [moq-dev/moq#6635](https://github.com/moq-dev/moq/pull/6635)). Padding is @@ -96,4 +98,4 @@ Tests, on `@moq/clock`'s `Manual`: ## Required -- [Browser send estimate](/quest/m2/probe/browser-send-estimate.md) - a probe cluster needs a browser `getStats()` send estimate to pace against, which shipping Chrome lacks and Safari reports as null +- [Browser send estimate](/quest/m3/probe/browser-send-estimate.md) - a probe cluster needs a browser `getStats()` send estimate to pace against, which shipping Chrome lacks and Safari reports as null diff --git a/quest/m3/probe-rung-disable-js.md b/quest/m5/probe-rung-disable-js.md similarity index 88% rename from quest/m3/probe-rung-disable-js.md rename to quest/m5/probe-rung-disable-js.md index 10dff4968f6..d4cd757725e 100644 --- a/quest/m3/probe-rung-disable-js.md +++ b/quest/m5/probe-rung-disable-js.md @@ -12,9 +12,9 @@ lower renditions keep their share instead of every rung degrading together. Split from the native rung disable by the maintainer on 2026-10-20: the native publishers moved to m1 on native probing, while a browser can only probe with -[padding](/quest/m3/probe-padding.md). Moved to m3 with padding on 2026-11-06 -(maintainer, /quest-unblock), since it recovers only through padding, which -the maintainer parked for a rethink. +[padding](/quest/m5/probe-padding.md). + +Moved to m5 2026-11-12 (maintainer, /quest-audit): it recovers only through padding, parked for a rethink. - Mirror the native quest's rule and constants from `moq_mux::rate`, rather than tuning a JS-only one. Its decisions hold here too: catalog state @@ -38,7 +38,7 @@ Public API: none. Wire: none. ## Required -- [Padding probes](/quest/m3/probe-padding.md) - the browser's sender, which raises the estimate so a disabled rung can resume +- [Padding probes](/quest/m5/probe-padding.md) - the browser's sender, which raises the estimate so a disabled rung can resume ## Related diff --git a/quest/m3/publish-done-codes.md b/quest/m5/publish-done-codes.md similarity index 66% rename from quest/m3/publish-done-codes.md rename to quest/m5/publish-done-codes.md index 54fad80e086..7794df1b1b4 100644 --- a/quest/m3/publish-done-codes.md +++ b/quest/m5/publish-done-codes.md @@ -2,18 +2,16 @@ ## Goal -A moq-transport PUBLISH_DONE carries GOING_AWAY (0x4), TOO_FAR_BEHIND (0x5), and EXPIRED (0x6) when those are the reason, and a subscriber reads each back as its stream code, in Rust and JS, so the bindings report why instead of an unspecified remote error. A relay that ends a relayed track because its upstream ended (a malformed track it aborted, say) forwards that reason instead of INTERNAL_ERROR. +A moq-transport PUBLISH_DONE carries GOING_AWAY (0x4), TOO_FAR_BEHIND (0x5), and EXPIRED (0x6) when those are the reason, and a subscriber reads each back as its stream code, in Rust and JS, so the bindings report why instead of an unspecified remote error. A relay that ends a relayed track because its upstream ended with one of those reasons forwards it instead of INTERNAL_ERROR. ## Plan -Follow-up of [moq-dev/moq#7396](https://github.com/moq-dev/moq/pull/7396), which did this for MALFORMED_TRACK (0x12). The stream-reset codes already map both ways in `rs/moq-net/src/ietf/error.rs:30-75`; only PUBLISH_DONE doesn't use them. +Follow-up of [moq-dev/moq#7396](https://github.com/moq-dev/moq/pull/7396), which did this for MALFORMED_TRACK (0x12). The stream-reset codes already map both ways in `rs/moq-net/src/ietf/error.rs` (`GOING_AWAY`, `TOO_FAR_BEHIND`, `MALFORMED_TRACK`); PUBLISH_DONE (`Served::status` in `rs/moq-net/src/ietf/publisher.rs`, `PublishDoneStatus` in `ietf/publish.rs`) still sends INTERNAL_ERROR for every error but authorization and MALFORMED_TRACK. Decided 2026-10-30 (maintainer, follow-up planning after the /quest-afk run). Reuse #7396's path: the publisher picks the PUBLISH_DONE code from the same error-to-code mapping that resets the track's streams, and the subscriber reads it back as the peer's stream code. Cover drafts 18, 20, and 22, both bindings, and JS, each test failing without its fix. Draft 20 added 2026-11-03 (maintainer, /quest-plan of the follow-ups from [moq-dev/moq#7704](https://github.com/moq-dev/moq/pull/7704)), since draft 20 joined the covered drafts in [interop coverage](https://github.com/moq-dev/moq/blob/acef90cf0425d9f0e4e699759cede07adfb2ecfb/quest/m1/interop-coverage/README.md). -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs): add the relayed case here, from [moq-dev/moq#7135](https://github.com/moq-dev/moq/pull/7135)'s follow-ups. When the relay aborts a malformed upstream track it still ends the downstream subscription with PUBLISH_DONE `INTERNAL_ERROR`, so bindings show `Internal`. Map the upstream end's reason through the same path, after checking whether draft-22's PUBLISH_DONE registry has a code that fits each case. Options shown: +Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs): add the relayed case here, from [moq-dev/moq#7135](https://github.com/moq-dev/moq/pull/7135)'s follow-ups. A relay already forwards a malformed upstream as MALFORMED_TRACK since #7396; map the other upstream end reasons through the same path, after checking whether draft-22's PUBLISH_DONE registry has a code that fits each case. -- m0 -- m1 -- ✅ m2, in this quest +Moved to m5 2026-11-12 (maintainer, /quest-audit): moq-transport only, with no named consumer. Public API: none. Wire: existing registered codes only. diff --git a/quest/m5/quic-dev-dep-path.md b/quest/m5/quic-dev-dep-path.md new file mode 100644 index 00000000000..00906fdd77f --- /dev/null +++ b/quest/m5/quic-dev-dep-path.md @@ -0,0 +1,15 @@ +# [XS] Dev-dependencies take the workspace version + +## Goal + +External dev-dependencies take the workspace version, so moq-tokio no longer pins its own `tempfile`, and internal dev-dependencies use a path. + +## Plan + +Widened 2026-10-31 from [moq-dev/moq#7260](https://github.com/moq-dev/moq/pull/7260): `rs/moq-tokio/Cargo.toml` pins its dev-dependency `tempfile = "3"` directly, while every other crate takes `tempfile = { workspace = true }`. Switch it to the workspace entry, and check the dev-dependency tables for other external crates pinned outside the workspace, and for internal crates taken with `workspace = true` (`cargo metadata`, or grep the `[dev-dependencies]` tables). + +The original case, moq-quic's dev-dependency on moq-congestion (found on [moq-dev/moq#7377](https://github.com/moq-dev/moq/pull/7377)), is gone: moq-congestion is now a normal dependency of moq-quic (`rs/moq-quic/Cargo.toml:39` at 4bf501b535), where `workspace = true` is right. + +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. + +Public API: none. Wire: none. diff --git a/quest/m5/quic-io-boundary.md b/quest/m5/quic-io-boundary.md index da4863e6152..f24ab6da587 100644 --- a/quest/m5/quic-io-boundary.md +++ b/quest/m5/quic-io-boundary.md @@ -13,8 +13,12 @@ those buffers with the ring is Decided in the 2026-09-30 plan: commit to the redesign, but start only after a `moq-quic` profile shows where ingress and egress spend their time (`just bench-profile cpu` on `io-uring-workers`). -No current profile is from `moq-quic`, and moq-uring already avoids every -other copy the current API allows. +`moq-quic` is moq-uring's only backend, so the io_uring relay profiles the +[perf epic](/quest/m9/perf/README.md) cites (such as +[moq-dev/moq#7505](https://github.com/moq-dev/moq/pull/7505)'s) are `moq-quic` +profiles, but none recorded so far splits ingress and egress across this +boundary; take that profile first. moq-uring already avoids every other copy +the current API allows. Candidates, to be ranked by the profile: diff --git a/quest/m2/cluster-routes/browser-proof.md b/quest/m5/routes-browser-proof.md similarity index 94% rename from quest/m2/cluster-routes/browser-proof.md rename to quest/m5/routes-browser-proof.md index 4f231a94a96..1fca454aec1 100644 --- a/quest/m2/cluster-routes/browser-proof.md +++ b/quest/m5/routes-browser-proof.md @@ -32,6 +32,8 @@ connects unproven rather than being refused. ## Plan +Moved to m5 2026-11-12 (maintainer, /quest-audit): it waits on a Firefox release and the lite-07 cut ships Firefox unproven (2026-11-07, below), so it no longer gates lite-07 through the routes epic. + Decided 2026-11-06 (maintainer, /quest-complete while queueing [moq-dev/moq#8022](https://github.com/moq-dev/moq/pull/8022)): cover Firefox and Chromium, and never send a proof the relay will reject. diff --git a/quest/m5/rs2ts-browser-report.md b/quest/m5/rs2ts-browser-report.md deleted file mode 100644 index 79b70be0ce3..00000000000 --- a/quest/m5/rs2ts-browser-report.md +++ /dev/null @@ -1,27 +0,0 @@ -# [M] Compare Rust-backed @moq/net in media workloads - -## Goal - -A reproducible browser report compares Rust-backed `@moq/net` with its -retained handwritten baseline through full publish/watch media workloads, -including bundle size, CPU, and first-frame latency. - -## Plan - -Decided 2026-10-12: this broader media report remains in m2 (m5 since the -2026-11-05 re-cut). It supplements -the focused browser comparison required before each backend adoption; it -cannot replace or postpone that decision evidence. Measure whichever backend -is deployed and name the exact revisions and build inputs. - -Run it as a variant of `bench/browser` (`just bench-browser BASE`), whose A/A -and injected-copy controls bound variability and which already separates -network, decode, and render costs and records unsupported metrics. Keep -baseline artifacts even after deleting handwritten code. Report regressions -and accepted tradeoffs without assuming that generated code is faster. - -Public API: none. Wire: none. - -## Required - -- [JS-only @moq/net](/quest/m5/rs2ts-js.md) - the adopted implementation and retained baseline diff --git a/quest/m5/rs2ts-js.md b/quest/m5/rs2ts-js.md index bfba7dd61f8..a71a2352d60 100644 --- a/quest/m5/rs2ts-js.md +++ b/quest/m5/rs2ts-js.md @@ -1,39 +1,48 @@ -# [L] Ship JS-only @moq/net from Rust +# [L] @moq/net runs the Rust model in the browser ## Goal -`@moq/net` ships the Rust protocol and model implementation as JavaScript -without requiring a WASM runtime. Publishing and consumption, lite and -moq-transport, and supported browser transports retain parity. The public -TypeScript interface is preserved where practical. +`@moq/net` ships the Rust protocol and model implementation to browsers, Node, +Bun, and Deno through rust2's WebAssembly export, with a generated TypeScript +binding. Publishing and consumption, lite and moq-transport, and supported +browser transports retain parity, and the public TypeScript interface is +preserved where practical. Translation to plain TypeScript replaces the +WebAssembly build only for a crate where it measures better. ## Plan -Decided 2026-10-12: WASM is transitional and JS-only is the final goal. -Decided 2026-10-31 (maintainer): moq's WASM support is removed, since nobody -used it and it is being replaced, so no production WASM backend precedes this. -The evaluation scopes this quest for the chosen backend and adds any required -research or implementation blockers before it becomes executable. Rejecting -wasm2js does not automatically select Charon or abandon the JS-only goal. - -Compare against the handwritten `@moq/net` in a complete browser workload -before adoption. Present measured size, startup, CPU, -memory, and usability tradeoffs for the maintainer to accept; preserve wire -and correctness. Run the browser suite and `just test interop --all` in CI. -Ensure shipped output loads without WebAssembly and preserves full-width -integer values and explicit lifetime/cancellation behavior. - -Pin the generation toolchain and verify reproducible package artifacts. -Normal JS contributors and consumers should not need that toolchain. Update -existing API docs and package READMEs for loading, deployment, and regeneration; -any necessary public API break is proposed separately and migrates consumers. - -Retire the handwritten implementation only after JS-only parity and adoption. -Preserve shared browser tests. - -Public API: `@moq/net` remains; enumerate any approved changes and obsolete -unpublished package removals. Wire: none. - -## Required - -- [WASM-to-JS evaluation](/quest/m5/rs2ts-wasm2js.md) - the selected route, evidence, and any additional prerequisites +Decided with the maintainer on 2026-11-09, in rust2's quest audit +([rust2-wasm](https://github.com/moq-dev/rust2/blob/main/quest/m2/wasm/README.md)): +✅ WASM ships, and TypeScript must beat it (rejected: JS-only staying the +goal with WebAssembly only a baseline; pausing the translator until +WebAssembly is measured). This supersedes the 2026-10-12 decision here that +WASM is transitional and JS-only is the final goal. Recorded here 2026-11-12 +(/quest-audit); the wasm2js evaluation and the separate browser report quests +were deleted, since rust2-wasm's gate replaces both: one report compares the +same `pub` API through rust2-wasm, through rust2-ts's translation, and +hand-written `@moq/net`, on CPU time, shipped bytes, and startup, in Chrome +and Firefox, and the maintainer decides from it. + +Still decided 2026-10-31 (maintainer): moq's old wasm-bindgen support stays +removed; rust2-wasm is the only WebAssembly route. + +- Adopt only after the gate's report, against the handwritten `@moq/net` in a + complete browser workload. Preserve wire and correctness, full-width + integer values, and explicit lifetime and cancellation behavior. Run the + browser suite and `just test interop --all` in CI. +- Pin the generation toolchain and verify reproducible package artifacts. + Normal JS contributors and consumers should not need that toolchain. Update + API docs and package READMEs for loading, deployment, and regeneration; any + necessary public API break is proposed separately and migrates consumers. +- Retire the handwritten implementation only after parity and adoption. + Preserve shared browser tests. + +Public API: `@moq/net` remains; enumerate any approved changes. Wire: none. + +## Waiting + +- moq-dev/rust2: its rust2-wasm epic reaches its gate, moq-net running in a browser with the three-way report. Check that rust2's `quest/m2/wasm/README.md` has no children left and its report exists. + +## Related + +- [Keep moq-net's model in rust2-ts's subset](/quest/m4/rs2ts-model-subset.md) - keeps the translation route measurable diff --git a/quest/m5/rs2ts-model-ready.md b/quest/m5/rs2ts-model-ready.md deleted file mode 100644 index cc4dad73704..00000000000 --- a/quest/m5/rs2ts-model-ready.md +++ /dev/null @@ -1,10 +0,0 @@ -# [XS] rust2-ts reaches moq-net's model - -## Goal - -[rust2's model epic](https://github.com/moq-dev/rust2/blob/main/quest/m2/ts-model/README.md) -has landed the translation of safe dependencies and bindings, so moq-net's model is the -next thing it translates. Check by reading that epic's Required list. - -This quest tracks a condition outside the repository. When it holds, delete -this quest and every `Required` entry that links it. diff --git a/quest/m5/rs2ts-wasm2js.md b/quest/m5/rs2ts-wasm2js.md deleted file mode 100644 index 0f67fe2acf6..00000000000 --- a/quest/m5/rs2ts-wasm2js.md +++ /dev/null @@ -1,85 +0,0 @@ -# [M] Evaluate WASM-to-JS for @moq/net - -## Goal - -A reproducible report evaluates JS-only output from the production Rust -browser implementation and settles the next route with the maintainer. -Try Binaryen wasm2js first; source translation is an alternative if its -measured cost is justified. The report updates the -[JS-only adoption quest](/quest/m5/rs2ts-js.md), rather than shipping an -unvalidated replacement or treating WASM as the final destination. - -## Plan - -Decided 2026-10-31 (maintainer): moq's WASM support (`rs/moq-wasm`, -`@moq/wasm`, moq-ffi's wasm32 build, and the browser backend baseline) is -removed, since nobody used it and it is being replaced. There is no production -WASM backend to follow or compare against. wasm2js needs a wasm32 build of the -Rust implementation again; confirm that route with the maintainer before -restoring one, or start from a source translator. Compare against the -handwritten `@moq/net`, measuring complete artifacts and matching startup paths. Require full-width u64, -lifetime, cancellation, backpressure, malformed-input, and memory-growth -coverage; the limited fixtures in #5337 do not prove these generally. - -[Binaryen wasm2js](https://github.com/WebAssembly/binaryen#wasm2js) emits JS, -but retains linear-memory machinery and documents differences in trapping -and numeric conversions. Validate the reachable behavior of our build rather -than assuming Rust compilation alone proves equivalence. Pin compatible -Rust, wasm-bindgen, and Binaryen versions; check feature lowering, import -bindings, legalized i64 interfaces, and cached views after memory growth. -The attachment's bindgen text patches are prototype evidence, not a stable -integration contract. Resolve failures at the tool or binding source. - -Options to retain in the report: - -| Route | Why consider it | Cost to establish | -|---|---|---| -| Existing Rust -> WASM -> JS | Reuses production Rust, including async, and its tests | Size, startup, linear memory, tool compatibility, and semantic differences | -| Coarser Rust/browser boundary | Can reduce crossings, copies, and Promise work for WASM and WASM-to-JS | Measure latency/backpressure as well as throughput; no assumed win | -| Charon MIR/LLBC -> TS | Control JS representations and preserve Rust ownership semantics | A custom emitter/runtime, extraction coverage, source maps, and pinned toolchain | -| rust-js | Reuse an existing Rust-to-JS compiler and runtime | Audit its supported subset and semantic differences, especially eager Promise execution versus lazy Rust futures | -| Component Model + jco | Reusable typed binding and distribution machinery | Additional interface/adapters; default output still uses WASM, while `--js` offers JS-only output | -| Own MIR/THIR backend | Control the extraction and lowering choices; under research in moq-dev/rust2 | Highest direct ownership of compiler integration and runtime semantics | - -Sources: [Charon](https://github.com/AeneasVerif/charon), -[rust-js](https://github.com/rust-js-lang/rust-js), and -[jco](https://bytecodealliance.github.io/jco/transpiling.html). -Jco is an interface/distribution alternative, not evidence of faster output. -A JS-only result need not be readable TypeScript. - -If wasm2js is acceptable, scope its adoption. If not, record the concrete -reason and plan the smallest source-translator investigation that addresses -it. Do not require a new repo, a shared npm runtime, a particular integer -representation, or codec/model/session rewrites before this evidence exists. -Split independently landable prerequisites into quests. - -Decided with the maintainer on 2026-10-23: an own rustc-driver translator is -developed as a parallel research track in -[moq-dev/rust2](https://github.com/moq-dev/rust2), starting with moq-net's -codec ([its model epic](https://github.com/moq-dev/rust2/blob/main/quest/m2/ts-model/README.md)). -It does not replace wasm2js as the first route here. Once it translates -reachable moq-net code, compare it as the source-translator row below; moq -consumes it as a pinned flake input, never by vendoring. - -Preserve earlier source-translation evidence: - -- [#5049](https://github.com/moq-dev/moq/pull/5049), summarized in - [#5341](https://github.com/moq-dev/moq/pull/5341), used a fail-closed Python - LLBC emitter on Charon 0.1.284/nightly-2026-09-17. Real zigzag/unzigzag - matched native Rust over 15,900 cases; the integer entry point was 840 B - gzip. Bun timings were a browser proxy. Ownership probes were handwritten, - not proof of general copy/drop lowering; varint::size was extracted but - its remaining constructs were not translated. -- [#5337](https://github.com/moq-dev/moq/issues/5337) reports closure-body - retention and control-flow duplication fixes in Charon - [#1513](https://github.com/AeneasVerif/charon/pull/1513) and - [#1514](https://github.com/AeneasVerif/charon/pull/1514). Both merged - upstream; pin a Charon revision carrying them before relying on them; its emitter benchmark does not establish - the performance of a future emitter. -- Source translation must establish precise Drop/Arc/Weak behavior, kio - wakers and guards, unsafe storage, nested Option states, trait objects, - value-keyed and ordered collections, and exact integer/overflow semantics. - Codec success alone cannot establish model/session feasibility. Keep any - proposed safe-storage or sans-IO rewrite measured and separately justified. - -Public API: none; the report scopes any later changes. Wire: none. diff --git a/quest/m5/safari-webtransport.md b/quest/m5/safari-webtransport.md index e3deb862bbf..5aec53a2f2b 100644 --- a/quest/m5/safari-webtransport.md +++ b/quest/m5/safari-webtransport.md @@ -30,7 +30,7 @@ tracking since PR #3345 and closes with this. Playwright has a worker `MediaStreamTrackProcessor` or transfers a `MediaStreamTrack`; the harness's capability probe already reports the path taken, and the fixture must take it. -- Safari publishers then pace from `getStats()`'s `estimatedSendRate`, which Safari 26.4 still reports as null, so they hold their rate until WebKit fills it (decided 2026-11-07 in [Browser publishers adapt to their send estimate](/quest/m2/probe/browser-send-estimate.md)). Check the `return_object_property_estimatedSendRate` entry under `api.WebTransport.getStats` in browser-compat-data, or a Safari publisher's `Reservation.rate`, and record which. +- Safari publishers then pace from `getStats()`'s `estimatedSendRate`, which Safari 26.4 still reports as null, so they hold their rate until WebKit fills it (decided 2026-11-07 in [Browser publishers adapt to their send estimate](/quest/m3/probe/browser-send-estimate.md)). Check the `return_object_property_estimatedSendRate` entry under `api.WebTransport.getStats` in browser-compat-data, or a Safari publisher's `Reservation.rate`, and record which. - Update `doc/lib/js/index.md`. - `parseUserAgent` in `js/net/src/connection/browser.ts` already returns the Safari and iOS versions (bowser left in diff --git a/quest/m5/teleop-robot.md b/quest/m5/teleop-robot.md index 2d10477c0ea..e813c769696 100644 --- a/quest/m5/teleop-robot.md +++ b/quest/m5/teleop-robot.md @@ -1,13 +1,14 @@ -# [M] Robot teleoperation conventions +# [S] Robot teleoperation conventions ## Goal -The track shapes and discovery every teleoperated machine needs are -documented conventions in `doc/`, built on moq-net and moq-json, so an -integrator writes an adapter rather than a stack. There is no `moq-robot` -crate, and this quest ships documentation only; [the MAVLink -gateway](/quest/m5/teleop-mavlink.md) implements the conventions in its own -quest. It is the gate for the rest of the questline. +The track layout, discovery, and timing every teleoperated machine needs are +documented conventions in `doc/`, beside the delivery shapes the +[teleoperation guide](/doc/concept/use-case/teleoperation.md) already +documents, so an integrator writes an adapter rather than a stack. There is no +`moq-robot` crate, and this quest ships documentation only; +[the MAVLink gateway](/quest/m5/teleop-mavlink.md) implements the conventions +in its own quest. It is the gate for the rest of the questline. ## Plan @@ -16,12 +17,20 @@ crate, too niche for a published library with no second consumer). The shapes below become documented conventions, and `moq-mavlink` ships as a moq-cli gateway like moq-srt. If a second consumer appears, generalize an announce-prefix fan-in helper into moq-net instead of adding a robot -crate. Where the text below says "the crate", read "the conventions". +crate. -Decided 2026-10-08: parked in m5. No teleoperation consumer needs the crate -yet; telemetry beside video needs [data sync](/quest/m3/watch-data-sync.md), +Decided 2026-10-08: parked in m5. No teleoperation consumer needs it yet; +telemetry beside video needs [data sync](/quest/m5/watch-data-sync.md), not this primitive. +Narrowed 2026-11-12 (/quest-audit): the teleoperation guide +(`doc/concept/use-case/teleoperation.md`, with `rs/moq-json/examples/teleop.rs`) +already documents the two delivery shapes (moq-json snapshot for replaceable +state, an ordered moq-json stream for commands), the group cache bound and +`Evicted`, a new command track per control session, command IDs and +acknowledgments, and that a connected subscriber is not an authorized +operator. What remains extends that guide rather than restating it. + ### Direction The operator publishes and the robot subscribes. The robot serves its own @@ -34,79 +43,30 @@ exist. | `robot/` | `video`, `telemetry` (lossy), `rpc` (reliable) | | `control//` | `command` (lossy), `rpc` (reliable) | -### Two delivery classes, and both already exist - -`moq-json` implements exactly this split, and the crate reuses it rather than -re-deriving it from track knobs: `snapshot` is lossy (one value updated over -time, intermediate updates collapsed, older groups dropped) and `stream` is -lossless (an ordered append-log where nothing is superseded). - -The use-cases draft says the same thing in terms of framing -(`drafts/draft-lcurley-moq-use-cases.md`, "Interaction"): a GROUP per input for -latency, a single GROUP with a FRAME per input for reliability. A group is one -QUIC stream (`open_uni`, `rs/moq-net/src/lite/publisher.rs`), so frames inside -it are ordered and delivered exactly once. - -That guarantee is scoped, and the crate must say so rather than promise -losslessness. A group holds at most `MAX_CACHE_BYTES` and `MAX_GROUP_FRAMES` -(`rs/moq-net/src/model/group.rs`). A raw group write past either aborts it -with `Error::GroupTooLarge`, but the JSON stream helper refuses an append -that might exceed the budget before encoding, leaving the log intact and -writable (`rs/moq-json/src/stream/mod.rs`). It never rolls its one group; -when its budget is spent the publisher opens a new track. A link drop or -reconnect loses what was in flight. -Those caps hold until [Unbounded groups](/quest/m2/unbounded-groups/README.md) -lands; after it, a lagging reader boards a flight if the publisher serves -replay, and otherwise refreshes or sees `Evicted`. -So the contract is ordered, gap-free delivery for a live reader within one -bounded log, and recovery after a reconnect belongs to the application -protocol. That is an acceptable division for MAVLink, whose mission, parameter -and file-transfer services already carry their own stop-and-wait -retransmission, but it must be stated, not assumed. - -The framing is where the guarantee lives, not the subscription flags: - -- `track::Subscriber::ordered()` returns an `Ordered` handle that reads the - groups it receives in sequence order. It is a local cursor, not a delivery - guarantee: groups that aged out or were skipped never arrive. A class built - on it would not be reliable, which is why the reliable class is a single - group instead. -- What makes the lossy class lossy on the wire is the publisher's - `Info::max_age`: `evict_expired_scan` aborts an aged-out group with `Error::Old` - (`rs/moq-net/src/model/track.rs`), and an abort resets the QUIC stream, so - stale bytes stop being retransmitted. -- A subscriber cannot weaken either class. `clamp_combined` - (`rs/moq-net/src/model/track.rs`) clamps the aggregate window down to - `Info::max_age`, and `Subscription::max_delay` already defaults to - `Duration::ZERO`, so a raw observer subscribing through `moq-net` neither - widens the window nor has to be prevented from trying. - ### Contents +- The broadcast and track layout above, and how a robot learns its operators + by announce prefix, documented against moq-net's origin announce stream + (`origin::Consumer`'s announcements under a prefix). moq-boy no longer reads + viewer broadcasts (its viewers dial a socket endpoint), so there is no + fan-in code to move; a helper lands in moq-net only when a second consumer + needs one. - The catalog entries. The hang catalog already advertises data tracks in its - `json` and `binary` sections (`rs/hang/src/catalog/{json,binary}.rs`), - written by the `moq-mux` data producers, and each entry carries `extra` - fields. Decide whether the robot's fields ride there or in a namespaced - root section through `moq-mux`'s `CatalogExt` and `RenditionConfig`. - No hang schema change. -- Announce-prefix fan-in, documented against moq-net's origin announce - stream (`origin::Consumer`'s announcements under a prefix). moq-boy no - longer reads viewer broadcasts (its viewers dial a socket endpoint), so - there is no fan-in code to move; a helper lands in moq-net only when a - second consumer needs one. -- The two delivery classes with the group structure and `Info::max_age` each - one needs: `moq-json`'s snapshot and stream modes for JSON; plain tracks - for opaque bytes, one group per latest value or one bounded ordered group - for the append-log. The [compression migration](https://github.com/moq-dev/moq/pull/7046) - deleted moq-flate's snapshot/stream helpers, so use the track's `compression` - flag when compression is useful, not those helpers. + `json` and `binary` sections (`rs/hang/src/catalog/{json,binary}.rs`), and + each entry carries `extra` fields. Decide whether the robot's fields ride + there or in a namespaced root section through `moq-mux`'s catalog + extension. No hang schema change. +- Opaque bytes: plain tracks with one group per latest value for the lossy + class, or one bounded ordered group for the append-log, setting the track's + `compression` flag when useful. The lossy class is lossy on the wire because + the publisher's `Info::max_age` evicts an aged-out group with `Error::Old` + (`rs/moq-net/src/model/track.rs`), resetting its stream. - Per-stage timestamp instrumentation, generalised from moq-boy's `status` - track. Check it against the publisher-reported stats track - ([media stats](/doc/concept/stats.md#media-reports), moq#2734) before adding a - second stats surface. Capability only: publishing a competitive benchmark - is out of scope, because Transitive's breakdown puts camera plus USB at - roughly 100 ms of a 130 ms glass-to-glass total, so we would mostly be - measuring somebody's webcam. + track (`rs/moq-boy/src/status.rs`). Check it against the publisher-reported + stats track ([media stats](/doc/concept/stats.md#media-reports)) before + adding a second stats surface. Capability only: publishing a competitive + benchmark is out of scope, since camera and USB dominate glass-to-glass + latency. ### Cross-track correlation @@ -115,16 +75,16 @@ its own quest. A command, the telemetry sample it produced, and the video frame showing the result share one timebase, so a recording is usable as training data and an operator sees what the machine actually saw. -The bridge is each broadcast's fixed catalog `media.clock: { wall, timescale }`, -documented in `doc/concept/hang.md` (#4461). Media and command tracks keep -their own timescales; convert PTS explicitly into the broadcast clock before -joining samples. The robot's video and telemetry share a clock; the operator's -command broadcast supplies its own mapping. Command and telemetry tracks -declare `Timescale::MILLI` explicitly: an undeclared Rust timescale means -untimed, so a track that leaves it out carries no timestamps to join. +The bridge is each broadcast's fixed catalog `media.clock`, documented in +`doc/concept/hang.md`. Media and command tracks keep their own timescales; +convert PTS explicitly into the broadcast clock before joining samples. The +robot's video and telemetry share a clock; the operator's command broadcast +supplies its own mapping. Command and telemetry tracks declare +`Timescale::MILLI` explicitly: an undeclared Rust timescale means untimed, so a +track that leaves it out carries no timestamps to join. The clock assumption: the two hosts' wall clocks are synchronized by the -deployment, not by the library. State this beside the API, since a join across +deployment, not by the library. State this in the docs, since a join across unsynchronized hosts looks valid while being wrong. Report whether a mapping is present, but never infer synchronization from its presence. No per-record anchors or clock-sync mechanism; see @@ -132,13 +92,13 @@ anchors or clock-sync mechanism; see ### Check against moq-boy -With no crate there is nothing to port. Instead, check the conventions -against `moq-boy`, the only existing consumer, as the no-arbitration case: it -merges viewers' JSON button states, where the gateway routes MAVLink frames, -so if the documented shapes cannot describe its crowd control they are the -wrong shapes. moq-boy changes only where it contradicts a convention. +Check the conventions against `moq-boy`, the only existing consumer, as the +no-arbitration case: it merges viewers' JSON button states, where the gateway +routes MAVLink frames, so if the documented shapes cannot describe its crowd +control they are the wrong shapes. moq-boy changes only where it contradicts a +convention. ## Related -- [Unbounded groups](/quest/m2/unbounded-groups/README.md) - removes the group caps that bound the reliable class's one log +- [Unbounded groups](/quest/m3/unbounded-groups/README.md) - changes the group cache bound the reliable class's one log relies on - [arbitration](/quest/m5/teleop-arbitration.md) - which controller is obeyed diff --git a/quest/m5/transcode-parallel-open.md b/quest/m5/transcode-parallel-open.md index ccf642020e8..832efcaccf2 100644 --- a/quest/m5/transcode-parallel-open.md +++ b/quest/m5/transcode-parallel-open.md @@ -11,12 +11,12 @@ rather than the sum of them. ## Plan Decided 2026-10-31 (user, /quest-plan of the follow-ups from the 2026-10-31 /quest-complete run), from [moq-dev/moq#7375](https://github.com/moq-dev/moq/pull/7375) (merged), -which finished the transcode rungs quest and lists this as a follow-up. Milestone m3: [moq-dev/moq#7375](https://github.com/moq-dev/moq/pull/7375)'s probe memo +which finished the transcode rungs quest and lists this as a follow-up. It is deferred because [moq-dev/moq#7375](https://github.com/moq-dev/moq/pull/7375)'s probe memo already makes each shape probe once, so this only matters for a source that keeps changing to new sizes. The resolve for an unmemoized shape runs serially inside the `select!` in -`Transcoder::run` (`rs/moq-transcode/src/lib.rs:117`, loop at `:170`), +`Transcoder::run` (`rs/moq-transcode/src/lib.rs:119`, loop at `:168`), blocking every rung meanwhile. Open the per-rung encoders concurrently (a `kio::Tasks` or joined futures), and have the loop keep encoding while the resolve is pending. Only the resolve for the newest source shape may diff --git a/quest/m4/uring-bench-followups.md b/quest/m5/uring-bench-followups.md similarity index 91% rename from quest/m4/uring-bench-followups.md rename to quest/m5/uring-bench-followups.md index be7b23fbc19..b0976ed66d6 100644 --- a/quest/m4/uring-bench-followups.md +++ b/quest/m5/uring-bench-followups.md @@ -8,9 +8,9 @@ Run-to-quiescence on the io_uring workers has acceptance numbers from a quiet ho Follow-ups of [moq-dev/moq#7530](https://github.com/moq-dev/moq/pull/7530), which runs a turn's tasks to quiescence before submitting. Its deterministic counts were the evidence (fanout 40.2 to 20.9 turns per round, -3.6% instructions); its `just bench-runtime` rows were indicative only, because host load put every runtime's p99 in seconds and moved the unchanged tokio rows by up to 8%. #7530 merged 2026-11-02. -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, one quest for both. +Moved to m5 2026-11-12 (maintainer, /quest-audit): no consumer; parked rather than deleted. -- **Quiet re-run.** Record `/proc/loadavg` and run `just bench-runtime` for fanout and fanout-heavy, base against [moq-dev/moq#7530](https://github.com/moq-dev/moq/pull/7530), alternating. [moq-dev/moq#7533](https://github.com/moq-dev/moq/pull/7533) and [moq-dev/moq#7534](https://github.com/moq-dev/moq/pull/7534) saw the load generator's p99 sit near 1.4 s in every run, a startup artifact inside the throughput window, so check whether a quiet host alone fixes p99 or the generator's latency window needs to exclude startup first. Record the result in the [perf epic](/quest/m4/perf/README.md). +- **Quiet re-run.** Record `/proc/loadavg` and run `just bench-runtime` for fanout and fanout-heavy, base against [moq-dev/moq#7530](https://github.com/moq-dev/moq/pull/7530), alternating. [moq-dev/moq#7533](https://github.com/moq-dev/moq/pull/7533) and [moq-dev/moq#7534](https://github.com/moq-dev/moq/pull/7534) saw the load generator's p99 sit near 1.4 s in every run, a startup artifact inside the throughput window, so check whether a quiet host alone fixes p99 or the generator's latency window needs to exclude startup first. Record the result in the [perf epic](/quest/m9/perf/README.md). The QUIC request-reply benchmark is now part of `worker_turns` in [moq-dev/moq#7530](https://github.com/moq-dev/moq/pull/7530). The nightly `rust-shared` group retains its timing series and turn/enter counter deltas in diff --git a/quest/m3/vaapi-vulkan-import.md b/quest/m5/vaapi-vulkan-import.md similarity index 96% rename from quest/m3/vaapi-vulkan-import.md rename to quest/m5/vaapi-vulkan-import.md index f8a49539447..70207f8a2e7 100644 --- a/quest/m3/vaapi-vulkan-import.md +++ b/quest/m5/vaapi-vulkan-import.md @@ -7,7 +7,7 @@ DMA-BUF with a DRM format modifier, from a producer that synchronizes explicitly through a timeline, encodes through VA-API on Intel (iHD) without a CPU round trip. An ignored hardware test proves it on an Arrow Lake iGPU, and `Kind::Auto` picks VA-API for a surface on an Intel device, through the same -seam as [Vulkan encode](/quest/m3/vulkan-encode.md): `encode::Config::input` +seam as [Vulkan encode](/quest/m5/vulkan-encode.md): `encode::Config::input` and a VA-API candidate with `vulkan: true` in `rs/moq-video/src/encode/backend/mod.rs`. @@ -69,6 +69,6 @@ Public API: none; the surface landed in #4975. Wire: none. ## Related -- [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - the AMD half of the same proof +- [Vulkan Video encode on AMD](/quest/m5/vulkan-encode.md) - the AMD half of the same proof - [GPU capacity and health](/quest/m3/gpu-health.md) - reports per device keyed by the same render node `dev_t` - [Video hardware validation](/quest/m5/video-hardware.md) - VA-API so far ran only on Meteor Lake diff --git a/quest/m5/video-hardware-access.md b/quest/m5/video-hardware-access.md index a220234d56f..10c0f057c6c 100644 --- a/quest/m5/video-hardware-access.md +++ b/quest/m5/video-hardware-access.md @@ -17,8 +17,8 @@ Check by asking the maintainer which machines are available. As of 2026-10-05 only part is on hand. The maintainer's desktop has an Intel Arrow Lake iGPU (iHD 26.1.2, `renderD128`) and an AMD RX 9070 (RADV, Mesa 26.0.8, `renderD129`), so it has a second render node; that check moved to -[VA-API external images](/quest/m3/vaapi-vulkan-import.md). The -[multi-vendor GPU quests](/quest/m3/gpu-release.md) test on it. +[VA-API external images](/quest/m5/vaapi-vulkan-import.md). The +[multi-vendor GPU quests](/quest/m5/gpu-release.md) test on it. Checked 2026-10-08 on that desktop: `vainfo` lists only `VAEntrypointEncSlice` on the iGPU (H.264, HEVC, HEVC Main10, AV1), no `EncSliceLP`, so the @@ -30,6 +30,13 @@ Decided 2026-10-08: the embedded-device condition folds in here, since a Pi 4 or 5 serves both the PipeWire camera pass and the embedded video check. Record the device, OS image, and driver when one is on hand. +Checked 2026-11-12 (/quest-audit): part of the Windows condition holds. The +maintainer's `windows` host (Windows 11, NVIDIA RTX 2070) has hosted agents +natively since 2026-11-10, so the Windows GPU probes and the Media Foundation +encode of a GPU export can run there. It has no camera, so the MJPEG and YUY2 +camera items still wait. Whether the maintainer's macOS laptop is Apple Silicon +is not recorded; ask. + The validation can start with whatever subset is available; split the rest out of [Video hardware validation](/quest/m5/video-hardware.md) when that happens. diff --git a/quest/m5/video-hardware.md b/quest/m5/video-hardware.md index e19a36d5d48..b508fd7df53 100644 --- a/quest/m5/video-hardware.md +++ b/quest/m5/video-hardware.md @@ -16,7 +16,7 @@ V4L2 `VIDIOC_EXPBUF` source is dropped with the export itself (see Decided 2026-10-08: the VAAPI low-power entrypoint check is dropped. Neither Meteor Lake nor the Arrow Lake test host exposes `EncSliceLP`, so no machine on hand reaches that fallback. A render node other than the first is checked -by [VA-API external images](/quest/m3/vaapi-vulkan-import.md). +by [VA-API external images](/quest/m5/vaapi-vulkan-import.md). - **Windows Media Foundation capture**: the camera opens while capture is enabled and closes when it is disabled, so the LED is off whenever capture @@ -35,7 +35,8 @@ by [VA-API external images](/quest/m3/vaapi-vulkan-import.md). - **OBS follows its grant on a shaped uplink**: stream from a real OBS build through `tc netem` (or `rs/moq-shaper`), narrow and reopen the link, and confirm the Stats tab's video target drops at once, climbs back over a few - seconds, and the encoder output follows (`cpp/obs/src/moq-rate.h`). + seconds, and the encoder output follows (the grant handling in + `cpp/obs/src/moq-output.cpp`). ### PipeWire DMA-BUF capture on KDE @@ -81,7 +82,7 @@ runner hosts (found by the m0 coverage audit): - WGC conversion and encode: `wgc_conversion_crops_odd_edges_and_preserves_color`, `wgc_conversion_workload`, and `wgc_nv12_encodes_with_matching_color_on_gpu_and_software`. The `wgc_*` - capture tests belong to [Windows.Graphics.Capture](/quest/m4/capture-wgc.md). + capture tests belong to [Windows.Graphics.Capture](/quest/m3/capture-wgc.md). - Media Foundation encode: `mediafoundation_cpu_rgba` and `mediafoundation_camera_texture`. @@ -92,6 +93,10 @@ validation on an RTX 3070 Ti found that NVENC rejects stream-ordered pool memory, so buffers registered with it must come from plain `cuMemAlloc`. That is not a bug any amount of review finds. +## Waiting + +- Maintainer: whether to split the items the `windows` host (RTX 2070, no camera; see [hardware access](/quest/m5/video-hardware-access.md)) can run today, the Windows GPU probes and the two dx12 export tests, into their own quest with `Hosts: windows`. Options: split them into m4 (or into lab, since running ignored tests changes no API), or keep everything here until the whole hardware set is on hand. Recommendation: split into m4 with `Hosts: windows`, since a downstream studio's Windows run lists this quest as related and those tests need nothing else. + ## Required - [Video validation hardware is on hand](/quest/m5/video-hardware-access.md) - the machines to run it on diff --git a/quest/m3/vulkan-encode.md b/quest/m5/vulkan-encode.md similarity index 97% rename from quest/m3/vulkan-encode.md rename to quest/m5/vulkan-encode.md index 3610e43ce4c..579808a809f 100644 --- a/quest/m3/vulkan-encode.md +++ b/quest/m5/vulkan-encode.md @@ -88,5 +88,5 @@ Public API: a new encoder backend and feature. Wire: none. ## Related -- [VA-API encodes an external Vulkan image](/quest/m3/vaapi-vulkan-import.md) - the Intel half of the same proof -- [Intra-refresh GOPs](/quest/m3/intra-refresh/README.md) - import and playback of intra-refresh video, which this encoder does not produce +- [VA-API encodes an external Vulkan image](/quest/m5/vaapi-vulkan-import.md) - the Intel half of the same proof +- [Intra-refresh GOPs](/quest/m4/intra-refresh/README.md) - import and playback of intra-refresh video, which this encoder does not produce diff --git a/quest/m3/watch-data-sync.md b/quest/m5/watch-data-sync.md similarity index 76% rename from quest/m3/watch-data-sync.md rename to quest/m5/watch-data-sync.md index 388cbf2fb28..15f56be54d5 100644 --- a/quest/m3/watch-data-sync.md +++ b/quest/m5/watch-data-sync.md @@ -21,11 +21,9 @@ releases it. they arrive. Since 2026-10-05 a track is all timed or all untimed (the untimed model, [#4822](https://github.com/moq-dev/moq/pull/4822)), so no stream mixes the two. - This replaces the sync code a web frontend writes today to hold KLV or - MAVLink telemetry back to the video playhead. In m3 rather than m1, since - that workaround exists. -- Decided 2026-10-08: the A/V clock and the watch worker land first, since - they change what `Sync` waits on and where it runs. + MAVLink telemetry back to the video playhead. +- Register with `Sync` where it runs when this starts. It no longer waits on + the [watch worker](/quest/m4/watch-worker.md), which moves `Sync` but + changes nothing this reader needs; whichever lands second adapts. -## Required - -- [Watch worker](/quest/m3/watch-worker.md) - moves `Sync` into a worker, where the reader registers +Moved to m5 2026-11-12 (maintainer, /quest-audit): no named consumer, since that workaround exists. diff --git a/quest/m9/README.md b/quest/m9/README.md new file mode 100644 index 00000000000..1a37fd292e6 --- /dev/null +++ b/quest/m9/README.md @@ -0,0 +1,36 @@ +# m9: lab experiments that merge whenever + +## Goal + +Experiments and measurements that touch no published crate's or package's +public API, no wire format, and no CLI flag, config key, or environment +variable: prototypes, spikes, benchmarks, profiles, and tuning inside one +implementation. Each can start at any time, run beside any other work, and +merge into `main` whenever it is ready, because nothing a release publishes +changes shape when it lands. + +## Plan + +Decided 2026-11-13 (maintainer, /quest-plan): renamed from `lab` to m9 +(rejected: keeping `lab`). `quest unblocked` orders milestones by name, so +`lab` ranked ahead of m0; m9 ranks last and leaves m6 to m8 free for the +shifts after a cut. m9 is never gated, so while m4 and m5 wait for the cut, +spare capacity comes here. + +Decided 2026-11-12 (maintainer, /quest-audit before the cut): lab sits beside +the milestones instead of in their order. A cut never waits on it, and the +shift after a cut leaves it in place. + +- ✅ A quest belongs here when its whole outcome is internal: an unpublished + crate, a benchmark or test harness, a measurement, a model, or a change + behind an existing API that no caller can observe except as speed or cost + (rejected: leaving such work in m4 and m5, where it queues behind a cut it + cannot affect). +- A lab quest that finds it needs a public API, wire, or CLI change stops and + brings that to the maintainer; the change is planned as its own quest in the + milestone its kind and urgency call for, and the lab quest keeps only the + internal half. +- A lab quest with a named consumer waiting on its result says so in its Plan + and still stays here; `quest unblocked` ranks it by what it unblocks. +- A lab quest nobody has started and nobody waits on is deleted when it goes + stale, like the rest of the deferred work; git history keeps it. diff --git a/quest/m9/archive-store-bench.md b/quest/m9/archive-store-bench.md new file mode 100644 index 00000000000..cc0b7839cfe --- /dev/null +++ b/quest/m9/archive-store-bench.md @@ -0,0 +1,25 @@ +# [S] Archive reads count their store round trips + +## Goal + +A test sweeps `moq_archive::Consumer` over track count and objects per track +and asserts how many LIST and GET calls each refresh and read makes, so a +change that adds store round trips to archive reads fails. + +## Plan + +- The cost is store round trips (per refresh, one LIST of the archive prefix + plus one per track; one GET per object), which an `InMemory` Criterion + bench would hide ([moq-dev/moq#6262](https://github.com/moq-dev/moq/pull/6262)). +- Count them with the crate's test store, `rs/moq-archive/src/mock.rs`, which + already logs every GET, PUT, DELETE, and LIST (`Mock::take`), as + `consumer/tests.rs` does for one case. Deterministic, so it needs no real + object store, credentials, or Nightly job. +- Sweep both axes, track count and objects per track, per the fan-out + benchmark rule. + +Rewritten 2026-11-12 (/quest-audit): the probe against a real S3, R2, or +MinIO store is dropped. Its latency is the store's, and the round-trip count +it was after is measured exactly in-process. + +Public API: none. Wire: none. diff --git a/quest/m4/kio-channel-contention/README.md b/quest/m9/kio-channel-contention/README.md similarity index 99% rename from quest/m4/kio-channel-contention/README.md rename to quest/m9/kio-channel-contention/README.md index 9ca86e55cf8..542496db149 100644 --- a/quest/m4/kio-channel-contention/README.md +++ b/quest/m9/kio-channel-contention/README.md @@ -112,7 +112,7 @@ Decided (maintainer, 2026-10-14): an append-only log primitive with front eviction (Append-only log (`kio::log`)), which much of moq-net could adopt - ([Log adoption](/quest/m4/kio-channel-contention/log-adoption.md)). + ([Log adoption](/quest/m9/kio-channel-contention/log-adoption.md)). - Rejected, so later sessions don't re-propose them: - Left-right (two copies, writer swaps the active one): before reusing the old copy the writer must see every reader leave it, an O(consumers) scan @@ -137,7 +137,7 @@ Decided (maintainer, 2026-10-14, per-thread follow-up): - Per-worker sharding is no longer ruled out; two spikes decide. Both run on the thread-per-core workers (`--runtime-workers`, the default once - [Shard by default](/quest/m4/perf/shard-default.md) lands) and link workers over an + [Shard by default](/quest/m9/perf/shard-default.md) lands) and link workers over an in-memory transport (`moq_net::transport::memory`), never real QUIC. The replica spike replicates the data plane and shares announcements; the cluster spike shares nothing (both results below). A @@ -220,7 +220,7 @@ landing the end-state storage once. So the log moved to m1, ahead of Front eviction, scoped to `GroupState`'s frames with the above as its expected win, and no longer waits on the spikes: they change how much it wins, not the storage Front eviction needs. `TrackState` arrival order moved to -[Log adoption](/quest/m4/kio-channel-contention/log-adoption.md), which +[Log adoption](/quest/m9/kio-channel-contention/log-adoption.md), which waited on them until both reported (below). A 2026-10-22 staging CDN chat profile (default tokio runtime, `f57c0aad`, DWARF stacks, 36% incomplete) put waiter registration at 1.97% self-time, the @@ -433,7 +433,7 @@ The heap repeats across runs and grows with W squared: per worker it is a full mesh re-announces every route to every peer (not checked route by route). At 16 workers a relay with 10,000 broadcasts would hold about 9 GB of announce state that a shared origin holds in 49 MB. -[Cluster route memory](/quest/m4/perf/cluster-route-memory.md) breaks that +[Cluster route memory](/quest/m9/perf/cluster-route-memory.md) breaks that per-route state down and shrinks it, profiles the hop's group cost, and fixes the `bench-profile lock` failures both spikes hit. diff --git a/quest/m4/kio-channel-contention/log-adoption.md b/quest/m9/kio-channel-contention/log-adoption.md similarity index 87% rename from quest/m4/kio-channel-contention/log-adoption.md rename to quest/m9/kio-channel-contention/log-adoption.md index 18f0072388a..247859875a6 100644 --- a/quest/m4/kio-channel-contention/log-adoption.md +++ b/quest/m9/kio-channel-contention/log-adoption.md @@ -19,13 +19,13 @@ stays on its mutex. per-thread sessions every consumer of a replica is on one thread, so its mutex is uncontended and a lock-free multi-reader log only pays off if the shared model survives. The replica spike (2026-11-01, in the - [epic](/quest/m4/kio-channel-contention/README.md)) does not settle + [epic](/quest/m9/kio-channel-contention/README.md)) does not settle it: replicas lose on low fan-out, so shared tracks would stay, and the shared track it keeps is still the most contended channel, with one reader per worker. Nor does the cluster spike (2026-11-01, same epic): a relay per thread took every channel's lock wait to zero, but its win is conditional on a worker mesh that never re-announces peer routes, which - [its plan](/quest/m4/kio-channel-contention/relay-per-thread.md) + [its plan](/quest/m9/kio-channel-contention/relay-per-thread.md) weighs, deciding whether adoption waits on it. If relay-per-thread lands, adoption is likely abandoned; until then the shared model stays. - Audit moq-net (and moq-relay's `critical.rs`) for other state that is @@ -35,3 +35,7 @@ stays on its mutex. - Maps and arbitrarily mutated state (`OriginState` routes, `BroadcastState` requests) are not log-shaped and stay put. - Split the PR per channel if the audit finds more than a couple. + +## Required + +- [Plan relay-per-thread](/quest/m9/kio-channel-contention/relay-per-thread.md) - decides whether the shared model, and so this adoption, survives diff --git a/quest/m4/kio-channel-contention/relay-per-thread.md b/quest/m9/kio-channel-contention/relay-per-thread.md similarity index 74% rename from quest/m4/kio-channel-contention/relay-per-thread.md rename to quest/m9/kio-channel-contention/relay-per-thread.md index 249d253bdc6..83d0d164d10 100644 --- a/quest/m4/kio-channel-contention/relay-per-thread.md +++ b/quest/m9/kio-channel-contention/relay-per-thread.md @@ -7,16 +7,17 @@ one cluster per worker thread linked by a mesh that never re-announces peer routes to peers. It weighs that design against the replica spike's per-track replicas, sets how routes from outside the relay reach every worker, measures announce state and propagation with no-re-announce, and -decides whether [Log adoption](/quest/m4/kio-channel-contention/log-adoption.md) +decides whether [Log adoption](/quest/m9/kio-channel-contention/log-adoption.md) waits on it. ## Plan -Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), milestone: +Planned 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-afk and /quest-complete runs), in this epic. -- m0 -- m1 -- ✅ m2, in this epic +Moved to lab 2026-11-12 (maintainer, /quest-audit): a plan and spikes that never merge. + +`origin::Producer::with_proxy` (`rs/moq-net/src/model/origin.rs`) now carries +the per-connection `proxy` preference the worker links would turn off. From the epic's 2026-11-01 cluster-spike verdict ([moq-dev/moq#7534](https://github.com/moq-dev/moq/pull/7534)): a conditional @@ -41,5 +42,5 @@ Public API: none (a plan). Wire: none. ## Related -- [Cluster route memory](/quest/m4/perf/cluster-route-memory.md) - shrinks the per-route state a full mesh multiplies -- [Shard by default](/quest/m4/perf/shard-default.md) - the thread-per-core workers this runs on +- [Cluster route memory](/quest/m9/perf/cluster-route-memory.md) - shrinks the per-route state a full mesh multiplies +- [Shard by default](/quest/m9/perf/shard-default.md) - the thread-per-core workers this runs on diff --git a/quest/m5/nvenc-pool.md b/quest/m9/nvenc-pool.md similarity index 100% rename from quest/m5/nvenc-pool.md rename to quest/m9/nvenc-pool.md diff --git a/quest/m4/perf/README.md b/quest/m9/perf/README.md similarity index 86% rename from quest/m4/perf/README.md rename to quest/m9/perf/README.md index f089a7a53d6..4b7f662a196 100644 --- a/quest/m4/perf/README.md +++ b/quest/m9/perf/README.md @@ -31,7 +31,7 @@ hot-path survey, so quests don't re-litigate them: one `origin::Producer` spans all workers, so a subscriber on worker B reads `kio::Lock` state written on worker A. These quests shrink that cost in place; they do not attempt per-worker model sharding, unless the - [kio channel contention](/quest/m4/kio-channel-contention/README.md) + [kio channel contention](/quest/m9/kio-channel-contention/README.md) spikes measure a win (decided 2026-10-14). - The batched write/read machinery (`frame::Buffer`, `write_frames`, the egress `Prefetch`) already exists in moq-net; egress is amortized, @@ -70,7 +70,7 @@ recipe it needs has shipped. It stays in later work (m4), since the release defe optimization. Decided 2026-10-22: the WebSocket and TCP stream transports join this line -as the [Stream transports](/quest/m4/stream-transports/README.md) epic, ranked +as the [Stream transports](/quest/m9/stream-transports/README.md) epic, ranked second, and the io_uring TCP line moves in from m3. io_uring pays far more for TCP than for UDP, so it no longer waits on fleet demand. @@ -79,7 +79,12 @@ tokio, so an io_uring quest measures with `just bench-runtime` at both revisions on the same host and worker count, and profiles with `just bench-profile MODES WORKLOAD WORKERS io-uring-workers [BASE]`. -## Required +Moved to lab 2026-11-12 (maintainer, /quest-audit): every child measures +first and touches no published API or wire. The epic no longer lists its two +sibling epics under `Required`, which held all fourteen children behind an +L-sized port; they are part of this line, not blockers of it. -- [kio channel contention](/quest/m4/kio-channel-contention/README.md) - workers spend less time blocked on kio channel-state mutexes (up to 215% of worker CPU on chat at 16 tokio workers), via a version counter, per-thread spikes, and an append-only log -- [Stream transports](/quest/m4/stream-transports/README.md) - qmux drives its own WebSocket and TCP sockets in one task, measured end to end, then serves them from the io_uring workers +## Related + +- [kio channel contention](/quest/m9/kio-channel-contention/README.md) - workers spend less time blocked on kio channel-state mutexes +- [Stream transports](/quest/m9/stream-transports/README.md) - qmux drives its own WebSocket and TCP sockets, then serves them from the io_uring workers diff --git a/quest/m4/perf/bench-loaded-cells.md b/quest/m9/perf/bench-loaded-cells.md similarity index 95% rename from quest/m4/perf/bench-loaded-cells.md rename to quest/m9/perf/bench-loaded-cells.md index d7c3b9ba023..c8df83f2691 100644 --- a/quest/m4/perf/bench-loaded-cells.md +++ b/quest/m9/perf/bench-loaded-cells.md @@ -10,7 +10,7 @@ Today each cell records the load it ran under (`host_load_1m` in `bench/instruct Decided 2026-11-12 (maintainer, `/quest-plan`), from an agent memory note that recorded the workaround ("compare only quiet cells within one run") instead of a fix: - **Scope.** ✅ Fix the bench (rejected: keep the workaround as a paragraph in `bench/README.md`). -- **Milestone.** ✅ the m4 perf epic (rejected: m0, m1). +- **Milestone.** ✅ the perf epic (rejected: m0, m1). Moved to lab 2026-11-12 (maintainer, /quest-audit): bench tooling only. What was measured on the shared 24-CPU host, 2026-11-02 and 2026-11-03: instructions per delivered frame read 5 to 20% low once the one-minute load passed about 9, because a contended relay batches more frames per wakeup, and above about 20 cells also lost groups. One revision read 142.7k under load and 169k to 175k quiet. diff --git a/quest/m4/perf/cluster-route-memory.md b/quest/m9/perf/cluster-route-memory.md similarity index 56% rename from quest/m4/perf/cluster-route-memory.md rename to quest/m9/perf/cluster-route-memory.md index b848b961eac..1b6933033dd 100644 --- a/quest/m4/perf/cluster-route-memory.md +++ b/quest/m9/perf/cluster-route-memory.md @@ -11,15 +11,17 @@ Two follow-ups of the kio channel-contention spikes, each measured: From [moq-dev/moq#7534](https://github.com/moq-dev/moq/pull/7534) (the relay-per-thread cluster spike) and [moq-dev/moq#7533](https://github.com/moq-dev/moq/pull/7533) (the replica spike). -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, in the perf epic, one quest for the spikes' measurement follow-ups. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): in the perf epic, one quest for the spikes' measurement follow-ups. -- **Route memory.** #7534 meshed W origins over moq-lite-06 `memory` sessions, as dialed cluster peers do. Announce heap per broadcast went from 4.9 KB with one origin to 45 KB at 4 workers and 886 KB at 16 (240 routes, about 3.7 KB each), and allocations from 64 to 14,600; 10,000 broadcasts at 16 would hold about 9 GB. Real full-mesh relay clusters pay the same per route. Measure it on a relay cluster with the counting allocator (#7534's `rs/moq-relay/examples/spike_announce.rs` is in its prototype commit [`ddd8b04`](https://github.com/moq-dev/moq/commit/ddd8b042ab82c7a4d53957e31fe35942957b7dfb), not on `main`), break the 3.7 KB down by structure (origin route entry, per-session announce state, hop lists, cursors), and shrink what's duplicated per route. Check against [cluster routing](/quest/m2/cluster-routing/README.md)'s per-origin routes first, since they change what a route holds. -- **Hop cost.** About 140 thousand user instructions per one-frame group across one in-memory hop; fanout's 64 subscribers spread over 15 hops raised user instructions 19%, while chat amortized it. Take a `cpu` profile of one hop and attribute it before fixing; a per-reader part belongs to [mesh session cost](/quest/m4/perf/mesh-session-cost.md), which owns the per-reader group cost. -- **`bench-profile lock`**: moved to [its own quest](/quest/m4/perf/profile-lock-switches.md) on 2026-11-03 (maintainer, /quest-plan of the follow-ups from [moq-dev/moq#7742](https://github.com/moq-dev/moq/pull/7742)), which keeps refusing nonalternating switch records and finds their cause instead of tolerating them. +Moved to lab 2026-11-12 (maintainer, /quest-audit): a measurement with no public API or wire change. + +- **Route memory.** #7534 meshed W origins over moq-lite-06 `memory` sessions, as dialed cluster peers do. Those numbers predate the lite-07 route layer (`rs/moq-net/src/route/`), so measure again on lite-07 before breaking anything down. Announce heap per broadcast went from 4.9 KB with one origin to 45 KB at 4 workers and 886 KB at 16 (240 routes, about 3.7 KB each), and allocations from 64 to 14,600; 10,000 broadcasts at 16 would hold about 9 GB. Real full-mesh relay clusters pay the same per route. Measure it on a relay cluster with the counting allocator (#7534's `rs/moq-relay/examples/spike_announce.rs` is in its prototype commit [`ddd8b04`](https://github.com/moq-dev/moq/commit/ddd8b042ab82c7a4d53957e31fe35942957b7dfb), not on `main`), break the 3.7 KB down by structure (origin route entry, per-session announce state, hop lists, cursors), and shrink what's duplicated per route. Check against [cluster routing](/quest/m3/cluster-routing/README.md)'s per-origin routes first, since they change what a route holds. +- **Hop cost.** About 140 thousand user instructions per one-frame group across one in-memory hop; fanout's 64 subscribers spread over 15 hops raised user instructions 19%, while chat amortized it. Take a `cpu` profile of one hop and attribute it before fixing; a per-reader part belongs to [mesh session cost](/quest/m9/perf/mesh-session-cost.md), which owns the per-reader group cost. +- **`bench-profile lock`**: moved to [its own quest](/quest/m9/perf/profile-lock-switches.md) on 2026-11-03 (maintainer, /quest-plan of the follow-ups from [moq-dev/moq#7742](https://github.com/moq-dev/moq/pull/7742)), which keeps refusing nonalternating switch records and finds their cause instead of tolerating them. Public API: none expected. Wire: none. ## Related -- [kio channel contention](/quest/m4/kio-channel-contention/README.md) - the epic both spikes reported to +- [kio channel contention](/quest/m9/kio-channel-contention/README.md) - the epic both spikes reported to - [Announce interest streams share one route table](/quest/m4/announce-interest-memory.md) - the same per-stream announce state, from a subscriber's side diff --git a/quest/m9/perf/cursor-version-skip.md b/quest/m9/perf/cursor-version-skip.md new file mode 100644 index 00000000000..026cfc72144 --- /dev/null +++ b/quest/m9/perf/cursor-version-skip.md @@ -0,0 +1,29 @@ +# [M] Track-state reads skip when nothing changed + +## Goal + +`Cursor::probe` and `GroupExpiry::is_expired` skip their track-state reads +when the track's version is unchanged, and fanout's mutex wait falls with +them. `Subscriber::poll_recv_group` then moves onto a `kio::Watch`. The watch +resets when the tail-grace deadline arms or the floor moves. Each part lands +only if a fresh lock profile still shows its wait. + +## Plan + +Planned 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-complete run). Found by [moq-dev/moq#7532](https://github.com/moq-dev/moq/pull/7532). With +`poll_finished` on the watch, `Cursor::sync`'s track-state read was the largest +fanout wait (52% at 16 workers), followed by `probe`. + +Since then, [moq-dev/moq#7493](https://github.com/moq-dev/moq/pull/7493) +made `Cursor::sync` (`rs/moq-net/src/model/track.rs`) return before locking +the track once its window is settled, and +[moq-dev/moq#7605](https://github.com/moq-dev/moq/pull/7605) made a live +reader's `probe` skip the fills and the subscription when it misses nothing. +`probe` still takes the track lock once for its judgement. + +- Re-profile first: `just bench-profile lock` against main, fanout and chat at + 4 and 16 workers, on a quiet host. #7532 couldn't capture 16-worker chat. + Drop any part below whose wait no longer shows. +- Make sure each read depends only on state that bumps the version. + `poll_recv_group` reads the wall clock (a copy's tail grace), and its floor + changes without a `Ready`, so it moves only with explicit resets. diff --git a/quest/m4/perf/hot-path-waste.md b/quest/m9/perf/hot-path-waste.md similarity index 90% rename from quest/m4/perf/hot-path-waste.md rename to quest/m9/perf/hot-path-waste.md index 0570b3fa944..abd864df6a1 100644 --- a/quest/m4/perf/hot-path-waste.md +++ b/quest/m9/perf/hot-path-waste.md @@ -11,9 +11,11 @@ Four small hot-path costs are removed or measured, each with a before/after coun ## Plan -Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): m2, in the perf epic, one quest because each part is small and found by reading the same profiles. +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): in the perf epic, one quest because each part is small and found by reading the same profiles. -- **Lite subscriber track name** ([moq-dev/moq#7500](https://github.com/moq-dev/moq/pull/7500)): #7500 brought the lite publisher's request serve to one copy per request; its allocation test's backtraces showed the subscriber also copying the name twice per subscription, in `SubscriberDriver::poll` and `TrackServe::prepare_establish` (`rs/moq-net/src/lite/subscriber.rs`). Count it the way #7500's `a_lite_request_keeps_one_copy_of_its_track_name` does, on the subscriber's side. +Moved to lab 2026-11-12 (maintainer, /quest-audit): internal costs with no public API or wire change. + +- **Lite subscriber track name** ([moq-dev/moq#7500](https://github.com/moq-dev/moq/pull/7500)): #7500 brought the lite publisher's request serve to one copy per request; its allocation test's backtraces showed the subscriber also copying the name twice per subscription, in `SubscriberDriver::poll` and `TrackServe::prepare_establish` (`rs/moq-net/src/lite/subscriber.rs`). Count it the way #7500's `a_lite_request_keeps_one_copy_of_its_track_name` (`rs/moq-net-tests/tests/announce_allocs.rs`) does, on the subscriber's side. - **`Subscriber.#drift`** ([moq-dev/moq#7523](https://github.com/moq-dev/moq/pull/7523)): `js/net/src/track.ts` walks back from the cap for the newest stamped group, and when `info` is undefined and no group carries a timestamp, the loop never breaks. In #7523's first fixture (an un-accepted, unstamped `Track.Producer`) it took 74% of steady-state CPU at 1024 groups. First check whether a real subscriber can read with history while its track's info is still undefined (for example before the info arrives, or on a version or path that never sets it), and how many groups it holds then. If it can, remember where the last walk stopped or skip the walk until a stamped group arrives; if not, record why and drop this part. - **Header encoding** ([moq-dev/moq#7526](https://github.com/moq-dev/moq/pull/7526)): `encodeVarint` (`js/net/src/stream.ts`) allocates an `ArrayBuffer(9)` and a view per varint, and an IETF `Frame.encode` makes five varints plus two `concat`s per object, about 12% self time in #7526's ietf 128x8 case and part of the GC share that grows with streams. Lite's `frameHeader` (`js/net/src/lite/publisher.ts`) does the same. Encode into one reused scratch buffer and copy out once. Measure with `just js serve-bench` in Chromium, as #7526 did. - **Fetch paths**: lite's `#runFetchGroup` and the IETF fetch writer still write each frame separately with no give-up watch; #7526 didn't measure them. Measure first, then apply #7526's live-path treatment if the cost is the same. diff --git a/quest/m4/perf/ingest-batch.md b/quest/m9/perf/ingest-batch.md similarity index 100% rename from quest/m4/perf/ingest-batch.md rename to quest/m9/perf/ingest-batch.md diff --git a/quest/m4/perf/js-subscribe-drop-single.md b/quest/m9/perf/js-subscribe-drop-single.md similarity index 100% rename from quest/m4/perf/js-subscribe-drop-single.md rename to quest/m9/perf/js-subscribe-drop-single.md diff --git a/quest/m9/perf/kio-waiter-costs.md b/quest/m9/perf/kio-waiter-costs.md new file mode 100644 index 00000000000..529e7cdd50f --- /dev/null +++ b/quest/m9/perf/kio-waiter-costs.md @@ -0,0 +1,29 @@ +# [S] kio waiters stop churning on long-lived drivers + +## Goal + +Two kio waiter costs are measured and fixed where they are real, each with before/after instruction counts: + +- A long-lived waiter parked on more than `PARKED` (16) lists is no longer discarded and rebuilt on every poll, measured on a relay session driver. +- A short-lived `kio::log` costs measurably less than about 1.3k instructions to set up and clean up, or the measurement shows why it can't. + +## Plan + +From [moq-dev/moq#7556](https://github.com/moq-dev/moq/pull/7556) (a measured no-win on adapter waker allocation) and [moq-dev/moq#7531](https://github.com/moq-dev/moq/pull/7531) (borrow-free QUIC state prototypes). + +Decided 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-afk run): in the perf epic, one quest for these, since they share kio's waiter and log internals and one measurement setup. + +Moved to lab 2026-11-12 (maintainer, /quest-audit): internal costs with no public API or wire change. + +- **`Park::hold`'s retire rule** (`rs/kio/src/waiter.rs`): `hold` replaces a waiter that is not `reusable()`, and a waiter parked on more than `PARKED = 16` lists is not ([moq-dev/moq#7789](https://github.com/moq-dev/moq/pull/7789) raised it from 8 to fix subscription ordering bugs). #7556 traced about 4 `Arc` per group on a session driver to this, with the new waiter's first registration landing on the session-close `Bridge` of the async-to-poll adapters. [moq-dev/moq#7771](https://github.com/moq-dev/moq/pull/7771) has since removed those adapters, so re-trace on the native poll transports first. A relay session driver parks on many more lists, so measure there (allocations and instructions per delivered group over `just bench-runtime` or a session bench). Then choose between raising `PARKED`, changing the retire rule, or leaving it, on the numbers. +- **`kio::log` lifetime cost**: #7531 attributed about 1.3k instructions per short-lived log to its epoch pin, flush, and collect, plus segment retirement on drop. That matters to any design holding one log per stream or per group. Profile a create/append/drop microbench, and cut what doesn't need to run for a log that never grew past its first segment. + +Dropped 2026-11-12 (/quest-audit): the IETF fetch close-watch part. The fetch path became a poll machine in [moq-dev/moq#7284](https://github.com/moq-dev/moq/pull/7284) (`FetchPayloadWrite` in `rs/moq-net/src/ietf/publisher.rs`), so no chunk builds its own `kio::wait` any more. + +A measured no-win for either part is a valid outcome; record it here and drop that part. + +Public API: none expected. Wire: none. + +## Related + +- [kio channel contention](/quest/m9/kio-channel-contention/README.md) - the append-only log design that would pay `kio::log`'s cost per stream diff --git a/quest/m4/perf/mesh-session-cost.md b/quest/m9/perf/mesh-session-cost.md similarity index 89% rename from quest/m4/perf/mesh-session-cost.md rename to quest/m9/perf/mesh-session-cost.md index 207c864368e..13ce5461c89 100644 --- a/quest/m4/perf/mesh-session-cost.md +++ b/quest/m9/perf/mesh-session-cost.md @@ -30,8 +30,6 @@ Decided 2026-11-03 (maintainer, follow-up planning after the 2026-11-03 /quest-a any floor change sends a batch on each route stream, even when no announce changed. Measure that floor-only batch cost against session count, and batch or -coalesce floor changes if it grows with sessions. Options shown: +coalesce floor changes if it grows with sessions, in this quest. -- m0 -- m1 -- ✅ m2, in this quest +Moved to lab 2026-11-12 (maintainer, /quest-audit): an internal optimization with no public API or wire change. diff --git a/quest/m4/perf/profile-lock-switches.md b/quest/m9/perf/profile-lock-switches.md similarity index 80% rename from quest/m4/perf/profile-lock-switches.md rename to quest/m9/perf/profile-lock-switches.md index 122ca0b1bed..b3143559da0 100644 --- a/quest/m4/perf/profile-lock-switches.md +++ b/quest/m9/perf/profile-lock-switches.md @@ -11,9 +11,9 @@ number lost is reported beside the result. ## Plan Decided 2026-11-03 (maintainer, /quest-plan of the follow-ups from -[moq-dev/moq#7742](https://github.com/moq-dev/moq/pull/7742)): milestone m2, -in the perf epic. The `bench-profile lock` bullet moves here from -[cluster route memory](/quest/m4/perf/cluster-route-memory.md), since it has +[moq-dev/moq#7742](https://github.com/moq-dev/moq/pull/7742)): in the perf +epic. The `bench-profile lock` bullet moves here from +[cluster route memory](/quest/m9/perf/cluster-route-memory.md), since it has nothing to do with route memory. - Decision prompt (maintainer, 2026-11-03): cluster-route-memory said to @@ -28,11 +28,12 @@ nothing to do with route memory. other agent was running, so outside load is not an established cause. - [moq-dev/moq#7533](https://github.com/moq-dev/moq/pull/7533) could not record base chat at 16 workers. - - The [kio channel contention](/quest/m4/kio-channel-contention/README.md) + - The [kio channel contention](/quest/m9/kio-channel-contention/README.md) notes, and the log lock profile quest #7742 deleted, also record refusals. - Check: the refusal is at `bench/profile.py:131`, with its test at - `bench/profile.test.py:279`. Capture the raw records of a refused run and + `bench/profile.test.py:459`. Duplicate records at one timestamp are + already skipped; records at distinct timestamps still refuse. Capture the raw records of a refused run and see whether they come from perf losing records, from records out of order across CPUs, or from a parse error. Fix the cause in the capture or the parser. diff --git a/quest/m4/perf/quic-lock-contention.md b/quest/m9/perf/quic-lock-contention.md similarity index 98% rename from quest/m4/perf/quic-lock-contention.md rename to quest/m9/perf/quic-lock-contention.md index 1f8ee755610..fadb477343c 100644 --- a/quest/m4/perf/quic-lock-contention.md +++ b/quest/m9/perf/quic-lock-contention.md @@ -64,4 +64,4 @@ staging queue and batch dispatch should need none. ## Related -- [Relay allocations](/quest/m4/perf/relay-allocs.md) - the allocation profile of the same workload +- [Relay allocations](/quest/m9/perf/relay-allocs.md) - the allocation profile of the same workload diff --git a/quest/m4/perf/relay-allocs.md b/quest/m9/perf/relay-allocs.md similarity index 60% rename from quest/m4/perf/relay-allocs.md rename to quest/m9/perf/relay-allocs.md index c85120c8685..05403c66d29 100644 --- a/quest/m4/perf/relay-allocs.md +++ b/quest/m9/perf/relay-allocs.md @@ -3,20 +3,21 @@ ## Goal The fanout relay's heap churn per delivered frame drops measurably, starting -with the per-stream `Session::clone` and `SendStream::poll_write` sites, and -the heap profile tracks it so a regression shows. +from a fresh heap profile of the native poll transports, and the heap profile +tracks it so a regression shows. ## Plan -Measured by #5830's heap profile (`just bench-profile heap fanout-heavy 4`): -about 1.08M allocations/s, roughly 70 per delivered frame. Top churn: +Re-profile first. #5830's heap profile (`just bench-profile heap fanout-heavy 4`) +measured about 1.08M allocations/s, roughly 70 per delivered frame, with `SendStream::poll_write` 41.5%, the web transport `Session::clone` 15.9%, moq-net `SendStream` poll 10.3%, `Session` poll 8.2%, -`group::Consumer::with_recover` 5.9%, kio `Tasks` 5.5%. - -`moq-tokio`'s `poll_open_uni` clones the session and boxes a future per -opened stream, and the web `Session` holds header `Vec`s and optional -connect request/response, so each clone allocates. Start there. +`group::Consumer::with_recover` 5.9%, and kio `Tasks` 5.5%. Those +numbers predate [moq-dev/moq#7771](https://github.com/moq-dev/moq/pull/7771), +which replaced the async-to-poll adapters with native poll transports: +`moq-tokio`'s `poll_open_uni` (`rs/moq-tokio/src/quic/driver/connection.rs`) +no longer clones the session or boxes a future per opened stream. Fix what the +new profile shows, not the old list. A 2026-10-22 staging CDN SD profile (`f57c0aad`) put memory moves at 3.78% self-time; `SendStream::poll_write` was the nearest resolved caller for 1.43% @@ -38,12 +39,6 @@ Validate #5830's CPU profile mode with a real capture before relying on its stacks. Sweep the fanout benchmark over publishers and subscribers so a cost that grows with the table shows as a slope. -Decided 2026-10-17 (maintainer): m2, under perf. - -Decided 2026-10-22 (maintainer): this waits on the work that replaces the -session clones and boxed opens at the top of this profile; re-profile after -it lands before fixing what remains. The QUIC adapters are now removed by -[moq-dev/moq#7771](https://github.com/moq-dev/moq/pull/7771); re-profile the -native poll implementations before fixing what remains. +Moved to lab 2026-11-12 (maintainer, /quest-audit): an internal optimization with no public API or wire change. Public API: none expected. Wire: none. diff --git a/quest/m4/perf/route-lock.md b/quest/m9/perf/route-lock.md similarity index 90% rename from quest/m4/perf/route-lock.md rename to quest/m9/perf/route-lock.md index 9f11250e7a3..50d74bcc53c 100644 --- a/quest/m4/perf/route-lock.md +++ b/quest/m9/perf/route-lock.md @@ -11,4 +11,4 @@ lock's share of mutex wait falls. Planned 2026-11-01 (maintainer, follow-up planning after the 2026-11-01 /quest-complete run). Profile with `just bench-profile lock fanout`, and remove each acquisition that reads state already held. Also make `bench/profile.py lock` accept `tokio-rt-worker` threads, which #7493 flagged. `GroupExpiry::is_expired` is -covered by [track-state reads](/quest/m4/perf/cursor-version-skip.md). +covered by [track-state reads](/quest/m9/perf/cursor-version-skip.md). diff --git a/quest/m4/perf/shard-default.md b/quest/m9/perf/shard-default.md similarity index 93% rename from quest/m4/perf/shard-default.md rename to quest/m9/perf/shard-default.md index 51e0ef0d6f5..e283c69b29d 100644 --- a/quest/m4/perf/shard-default.md +++ b/quest/m9/perf/shard-default.md @@ -10,8 +10,8 @@ shared runtime stays the default. ## Plan -Sharding is opt-in: `runtime.workers` is unset by default, and -`runtime.pin` defaults off. The maintainer decided on #5688 to keep it that +Sharding is opt-in: `runtime.workers` defaults to `"shared"` +(`rs/moq-relay/src/runtime.rs`), and `runtime.pin` defaults off. The maintainer decided on #5688 to keep it that way until a measurement justifies the flip. - Extend `just bench-shard` to load that saturates the shared runtime, for @@ -63,4 +63,4 @@ Update `doc/bin/relay/` for the default change. ## Related -- [kio channel contention](/quest/m4/kio-channel-contention/README.md) - its per-thread spikes run on the workers +- [kio channel contention](/quest/m9/kio-channel-contention/README.md) - its per-thread spikes run on the workers diff --git a/quest/m4/perf/ts-idle-jitter.md b/quest/m9/perf/ts-idle-jitter.md similarity index 100% rename from quest/m4/perf/ts-idle-jitter.md rename to quest/m9/perf/ts-idle-jitter.md diff --git a/quest/m3/stream/bench.md b/quest/m9/stream-bench.md similarity index 88% rename from quest/m3/stream/bench.md rename to quest/m9/stream-bench.md index ab61793b75c..86390ce820b 100644 --- a/quest/m3/stream/bench.md +++ b/quest/m9/stream-bench.md @@ -15,4 +15,4 @@ gate here; the result informs the docs and the mesh guide. ## Required -- [Rust byte streams](/quest/m3/stream/rust.md) - the stream under test +- [Rust byte streams](/quest/m1/stream/rust.md) - the stream under test diff --git a/quest/m4/stream-transports/README.md b/quest/m9/stream-transports/README.md similarity index 83% rename from quest/m4/stream-transports/README.md rename to quest/m9/stream-transports/README.md index 0d11bb92fc3..1b2fb7e82fc 100644 --- a/quest/m4/stream-transports/README.md +++ b/quest/m9/stream-transports/README.md @@ -22,26 +22,22 @@ Facts from the 2026-10-22 audit, so quests don't re-litigate them: over loopback, swept over transport x sessions x message size, and reports process CPU per message, which is one record up to the 16382-byte default maximum record size. `bench/README.md` covers both. -- The relay's WebSocket path is axum (hyper, `axum-server`, tokio-rustls), - then axum's tungstenite 0.29, converted zero-copy into qmux's tungstenite - 0.30 (`rs/moq-relay/src/websocket.rs`). The relay therefore compiles two - tungstenite versions. qmux names `tokio_tungstenite` and `tungstenite` types - in its public API (`rs/qmux/src/ws.rs`). +- The relay accepts WebSocket through axum and hyper (`axum-server`, + tokio-rustls), answers the upgrade itself, and hands the upgraded socket to + qmux (`rs/moq-relay/src/websocket.rs`). Since + [moq-dev/moq#7494](https://github.com/moq-dev/moq/pull/7494) the relay + compiles one tungstenite, 0.30, and since + [moq-dev/moq#8300](https://github.com/moq-dev/moq/pull/8300) qmux's public + API names no tungstenite type. - tungstenite itself costs a zero-filled 128 KiB read buffer, a copy into - its `out_buffer` on write, and a memmove after a partial write. The layers - qmux adds around it look heavier, though none is measured yet: - - on WebSocket, a `futures` split BiLock shared by reader and writer, a - `ws_pump` task, and an `mpsc(16)` channel (`rs/qmux/src/transport.rs`) - - a `run_reader` task taking the session `Mutex` per record - (`rs/qmux/src/session.rs`) - - a new `Vec` per outbound record from `mux::fill` - - on plain TCP, a `BytesMut::zeroed` allocation per inbound record and a - `BytesMut` per outbound size varint, behind the same shape as WebSocket: - `tokio::io::split`, a `reader_loop` task, and an `mpsc` channel -- qmux is hard-wired to tokio: `tokio::spawn`, `tokio::sync::mpsc`, - `tokio::select!`, and `tokio::time` for the WebSocket keep-alive. That is - why it cannot run on an io_uring worker today, and why the driver is a - prerequisite of the ring port rather than an optional win. + its `out_buffer` on write, and a memmove after a partial write. +- Before the driver, qmux wrapped the socket in a split BiLock, a `ws_pump` + task, an `mpsc` channel, a `run_reader` task taking the session `Mutex` per + record, and on TCP a `tokio::io::split` with a `reader_loop` task, all + hard-wired to tokio. #7494 replaced them with one caller-spawned `Driver` + (`rs/qmux/src/driver.rs`) that owns the socket and needs only + `tokio::io::{AsyncRead, AsyncWrite}`, so a worker's local task set can run + it. That driver is the ring port's prerequisite, and it has landed. - WebSocket framing is message-oriented: one Binary message is one qmux record (one frame on qmux-00), at most `max_record_size` (16382 by default), and Text is a protocol violation. The browser @@ -96,7 +92,7 @@ Against the base, every case runs fewer: tcp 13-18%, tls 6-14%, ws 17-21%, and w 1-13%, the smaller numbers at 16000 B where TLS dominates. Syscalls per message (strace, 16 sessions, 1024 B) stayed at about 2.55 on every transport. Instructions undercount tungstenite's zero-fill (`rep stos` is one instruction), -so the [codec](/quest/m4/stream-transports/codec.md) gate must re-profile CPU time. Two +so the [codec](/quest/m9/stream-transports/codec.md) gate must re-profile CPU time. Two paired relay runs (`chat` and `fanout` on `ws`, `wss`, `tcp`, `tls`) were inconclusive: the QUIC control, untouched by the driver, moved -8% to +25%, and most cases changed sign between runs. Only `tls` `fanout` rose in all six pairs @@ -118,7 +114,7 @@ Decisions (2026-10-22), each with its reason: bench attributes a material share to tungstenite once the driver has landed. tungstenite runs over any tokio-IO stream, the ring's included, so the codec is not on the ring's critical path. -- **The ring line moved in from m3.** io_uring pays far more for TCP than for +- **The ring line moved into this line.** io_uring pays far more for TCP than for UDP, so it no longer waits on fleet demand. Its ablation picks mechanisms; it no longer gates the line. This is a deliberate exemption from the perf line's "a measured no-win abandons the quest" rule (reaffirmed in review): @@ -127,8 +123,10 @@ Decisions (2026-10-22), each with its reason: right. The before/after is still recorded. - **The ablation reuses the bench harness** rather than building a second one. -- **Ranked second in m4/perf**, after kio channel contention, which is the - largest cost measured so far. +- **Ranked second in the perf line**, after kio channel contention, which is + the largest cost measured so far. + +Moved to lab 2026-11-12 (maintainer, /quest-audit): measured transport work with no wire change. Decided 2026-10-31 (maintainer), in [moq-dev/moq#7086](https://github.com/moq-dev/moq/pull/7086): diff --git a/quest/m4/stream-transports/ablation.md b/quest/m9/stream-transports/ablation.md similarity index 90% rename from quest/m4/stream-transports/ablation.md rename to quest/m9/stream-transports/ablation.md index 7c5c71eb075..3723126681d 100644 --- a/quest/m4/stream-transports/ablation.md +++ b/quest/m9/stream-transports/ablation.md @@ -4,7 +4,7 @@ A written-down number comparing ring TCP against tokio TCP under the relay's actual stream workload, on the same hardware. It picks which ring mechanisms -the [port](/quest/m4/stream-transports/ring.md) uses and records the expected win. +the [port](/quest/m9/stream-transports/ring.md) uses and records the expected win. ## Plan @@ -29,4 +29,4 @@ Vary at least: Record CPU per message, CPU per Gbps, syscalls per message, ring submissions and completions, and latency, in the same machine-readable form the UDP matrix uses. Attribute the gain (or its absence) to specific mechanisms, and record -the result in the [epic](/quest/m4/stream-transports/README.md) README. +the result in the [epic](/quest/m9/stream-transports/README.md) README. diff --git a/quest/m4/stream-transports/codec.md b/quest/m9/stream-transports/codec.md similarity index 87% rename from quest/m4/stream-transports/codec.md rename to quest/m9/stream-transports/codec.md index 94ed4008cab..bb52bbaeb3b 100644 --- a/quest/m4/stream-transports/codec.md +++ b/quest/m9/stream-transports/codec.md @@ -4,7 +4,7 @@ qmux implements the RFC 6455 subset it uses, and tungstenite and tokio-tungstenite leave the dependency tree, but only if the stream -transport bench (see the [epic](/quest/m4/stream-transports/README.md)) attributes a +transport bench (see the [epic](/quest/m9/stream-transports/README.md)) attributes a material share of WebSocket CPU to tungstenite now that qmux's `Driver` owns the socket. If the bench shows no such share, delete this quest. @@ -20,9 +20,10 @@ moved here when [moq-dev/moq#7494](https://github.com/moq-dev/moq/pull/7494) landed the driver and kept tungstenite): tungstenite's default read buffer dominates its share of the profile. Shrink `WebSocketConfig::read_buffer_size` to near qmux's maximum record size -(about 16 KiB) in `qmux::ws` (`capped` in `rs/qmux/src/ws.rs`) and on the -relay's axum `WebSocketUpgrade` (`rs/moq-relay/src/websocket.rs`) until this -codec replaces tungstenite. That setting is the buffer tungstenite allocates +(about 16 KiB) in `capped` (`rs/qmux/src/ws.rs`) until this codec replaces +tungstenite. That one site covers the relay too: it no longer uses axum's +`WebSocketUpgrade`, and hands its upgraded socket to `qmux::ws::Upgraded` +(`rs/moq-relay/src/websocket.rs`), which builds its socket with `capped`. That setting is the buffer tungstenite allocates up front, not a memory cap: a larger frame still grows past it, so the existing message and frame size limits stay the bound. The stopgap lowers the baseline allocation per connection and the zero-fill per read attempt. diff --git a/quest/m4/stream-transports/relay.md b/quest/m9/stream-transports/relay.md similarity index 78% rename from quest/m4/stream-transports/relay.md rename to quest/m9/stream-transports/relay.md index 3aecdb5aef9..846de7f1986 100644 --- a/quest/m4/stream-transports/relay.md +++ b/quest/m9/stream-transports/relay.md @@ -15,7 +15,7 @@ relay serves `listen`'s `tcp`/`unix` listeners beside it from the shared tokio runtime, through `init_streams` (`rs/moq-relay/src/relay.rs`). Move the TCP listeners onto the workers: each worker binds its own listener in the reuseport group and runs the router from -[ring TCP](/quest/m4/stream-transports/ring.md) on it. A Unix socket has no reuseport +[ring TCP](/quest/m9/stream-transports/ring.md) on it. A Unix socket has no reuseport group, so a `unix` listener keeps its one shared acceptor; the integration test covers it too. @@ -39,7 +39,11 @@ CORS scoping, the same landing-page fallback, and the same to prove a WebSocket session and an HTTP route both work when served from a worker. +## Waiting + +- Maintainer: how moq-tokio hands a WebSocket session's driver back, since this lab quest must not change a published API without a decision. Options: add an additive sibling beside `moq_tokio::websocket::accept` that returns the session and its driver, or reshape `websocket::accept` itself, which breaks moq-tokio and would have to land before the cut. Recommendation: the additive sibling, so nothing here waits on the cut. + ## Required -- [Ring TCP](/quest/m4/stream-transports/ring.md) - the module and adapters this +- [Ring TCP](/quest/m9/stream-transports/ring.md) - the module and adapters this serves from diff --git a/quest/m4/stream-transports/ring.md b/quest/m9/stream-transports/ring.md similarity index 96% rename from quest/m4/stream-transports/ring.md rename to quest/m9/stream-transports/ring.md index 3c3643b4cb5..ec8fcd512d6 100644 --- a/quest/m4/stream-transports/ring.md +++ b/quest/m9/stream-transports/ring.md @@ -33,4 +33,4 @@ existing suite does. ## Required -- [Ring TCP ablation](/quest/m4/stream-transports/ablation.md) - picks the mechanisms before the worker implementation starts +- [Ring TCP ablation](/quest/m9/stream-transports/ablation.md) - picks the mechanisms before the worker implementation starts diff --git a/quest/m5/uring-cluster-dial.md b/quest/m9/uring-cluster-dial.md similarity index 91% rename from quest/m5/uring-cluster-dial.md rename to quest/m9/uring-cluster-dial.md index 4f8b94a201f..a57b45f6532 100644 --- a/quest/m5/uring-cluster-dial.md +++ b/quest/m9/uring-cluster-dial.md @@ -24,11 +24,11 @@ tokio, swept over links and subscribers. Decided 2026-10-22 (maintainer), from the io_uring audit: cluster dials stay measure-gated. WebRTC data channels stay on tokio. TCP and WebSocket move onto the workers through -[Stream transports](/quest/m4/stream-transports/README.md), which its own +[Stream transports](/quest/m9/stream-transports/README.md), which its own decision exempts from a gate. Public API: none until the port. Wire: none. ## Related -- [Stream transports](/quest/m4/stream-transports/README.md) - the TCP and WebSocket half, ported in m4 +- [Stream transports](/quest/m9/stream-transports/README.md) - the TCP and WebSocket half, ported in m4 diff --git a/quest/m5/3202-moq-uring-use-fixed-file-slots-for-worker-udp-sockets.md b/quest/m9/uring-fixed-file-slots.md similarity index 100% rename from quest/m5/3202-moq-uring-use-fixed-file-slots-for-worker-udp-sockets.md rename to quest/m9/uring-fixed-file-slots.md diff --git a/quest/m5/3200-moq-uring-batch-completion-wakeups-with-min-timeout.md b/quest/m9/uring-min-timeout-wakeups.md similarity index 100% rename from quest/m5/3200-moq-uring-batch-completion-wakeups-with-min-timeout.md rename to quest/m9/uring-min-timeout-wakeups.md diff --git a/quest/m5/3199-moq-uring-remove-sq-indirection-and-per-enter-ring-fd.md b/quest/m9/uring-sq-indirection.md similarity index 100% rename from quest/m5/3199-moq-uring-remove-sq-indirection-and-per-enter-ring-fd.md rename to quest/m9/uring-sq-indirection.md diff --git a/quest/m5/uring-transport-machines.md b/quest/m9/uring-transport-machines.md similarity index 95% rename from quest/m5/uring-transport-machines.md rename to quest/m9/uring-transport-machines.md index 263ca7c21aa..2c773cc6abb 100644 --- a/quest/m5/uring-transport-machines.md +++ b/quest/m9/uring-transport-machines.md @@ -27,4 +27,4 @@ Public API: none unless the check goes, which removes ## Related -- [Relay io_uring packages](/quest/m3/relay-io-uring-package.md) - not blocked on this; `Local` already gives parity, and only its cost is measured here +- [Relay io_uring packages](/quest/m4/relay-io-uring-package.md) - not blocked on this; `Local` already gives parity, and only its cost is measured here diff --git a/quest/m3/video-quality/README.md b/quest/m9/video-quality/README.md similarity index 92% rename from quest/m3/video-quality/README.md rename to quest/m9/video-quality/README.md index e1f1b819bc9..49da274a12e 100644 --- a/quest/m3/video-quality/README.md +++ b/quest/m9/video-quality/README.md @@ -27,13 +27,12 @@ Decided in planning (2026-10-28): fetched fails the run instead of being skipped (✅ "Manifest + R2" over "Private repo (moq.pro)" and "Synthetic + open only"). - Gating: each row (host, backend, clip, rung) gets a VMAF floor and a pop - ceiling measured from nightlies, the way [ViSQOL](/quest/m4/audio-visqol.md) + ceiling measured from nightlies, the way [ViSQOL](/quest/m5/audio-visqol.md) keeps `budgets.json`. Scores also feed the [benchmark table](/quest/m4/bench-sources.md) for trends (✅ "Floors + trend" over "Trend only" and "Golden diff", which is brittle across drivers). -- Milestone m2, ranked right after release execution (✅ "Near the top" - over "Next to GPU CI" and "Next to audio ViSQOL"). Rate control under - on-demand encoding is unmeasured, and moq-boy is visibly bad. +- Moved to lab 2026-11-12 (maintainer, /quest-audit): a test harness with no + public API and no consumer waiting on it. - Split (✅ "Epic, 4 quests" over "Epic, 3 quests" and "Single [XL] quest"): the harness, then on-demand pops, hardware rows, and moq-boy, each requiring the harness. diff --git a/quest/m3/video-quality/gpu-runners.md b/quest/m9/video-quality/gpu-runners.md similarity index 100% rename from quest/m3/video-quality/gpu-runners.md rename to quest/m9/video-quality/gpu-runners.md diff --git a/quest/m3/video-quality/hardware.md b/quest/m9/video-quality/hardware.md similarity index 84% rename from quest/m3/video-quality/hardware.md rename to quest/m9/video-quality/hardware.md index 356e7ae7ed4..52633d97a3e 100644 --- a/quest/m3/video-quality/hardware.md +++ b/quest/m9/video-quality/hardware.md @@ -11,7 +11,7 @@ prints the same table locally on any GPU host. ## Plan -Decisions are recorded in the [epic](/quest/m3/video-quality/README.md): +Decisions are recorded in the [epic](/quest/m9/video-quality/README.md): every row gets floors, and an offline host fails its job. Local runs write their table to `.tmp/` and upload nothing. @@ -24,13 +24,13 @@ their table to `.tmp/` and upload nothing. - Select the backend explicitly per row. moq-transcode's default prefers hardware, so a missing driver must fail the row, not fall back to OpenH264. -- When [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) lands, its +- When [Vulkan Video encode on AMD](/quest/m5/vulkan-encode.md) lands, its backend joins the desktop's rows. Public API: none. Wire: none. ## Required -- [Harness](/quest/m3/video-quality/harness.md) - the scorer and row format +- [Harness](/quest/m9/video-quality/harness.md) - the scorer and row format - [GPU CI](/quest/m3/gpu-ci.md) - `nixos`'s runner gains its GPU -- [GPU runners](/quest/m3/video-quality/gpu-runners.md) - `nixos2` and the desktop host runners +- [GPU runners](/quest/m9/video-quality/gpu-runners.md) - `nixos2` and the desktop host runners diff --git a/quest/m3/video-quality/harness.md b/quest/m9/video-quality/harness.md similarity index 96% rename from quest/m3/video-quality/harness.md rename to quest/m9/video-quality/harness.md index e9398b418e0..daa45780246 100644 --- a/quest/m3/video-quality/harness.md +++ b/quest/m9/video-quality/harness.md @@ -10,7 +10,7 @@ scores reach the benchmark table. ## Plan -Decisions are recorded in the [epic](/quest/m3/video-quality/README.md). +Decisions are recorded in the [epic](/quest/m9/video-quality/README.md). - Layout: `test/video-quality/`, mirroring `test/audio-quality/`, with per-row floors and ceilings in a budgets file measured from nightlies by the @@ -60,5 +60,5 @@ Public API: none. Wire: none. ## Related -- [Audio ViSQOL](/quest/m4/audio-visqol.md) - the same floor rule for audio +- [Audio ViSQOL](/quest/m5/audio-visqol.md) - the same floor rule for audio - [Benchmark sources](/quest/m4/bench-sources.md) - the history format the scores join diff --git a/quest/m3/video-quality/moq-boy.md b/quest/m9/video-quality/moq-boy.md similarity index 96% rename from quest/m3/video-quality/moq-boy.md rename to quest/m9/video-quality/moq-boy.md index 8f1e3009e43..eaa9d25cd31 100644 --- a/quest/m3/video-quality/moq-boy.md +++ b/quest/m9/video-quality/moq-boy.md @@ -39,4 +39,4 @@ Public API: possibly, see above. Wire: none. ## Required -- [Harness](/quest/m3/video-quality/harness.md) - the scorer this fix is measured with +- [Harness](/quest/m9/video-quality/harness.md) - the scorer this fix is measured with diff --git a/quest/m3/video-quality/on-demand.md b/quest/m9/video-quality/on-demand.md similarity index 93% rename from quest/m3/video-quality/on-demand.md rename to quest/m9/video-quality/on-demand.md index 6e8c9aeb4c0..37314803804 100644 --- a/quest/m3/video-quality/on-demand.md +++ b/quest/m9/video-quality/on-demand.md @@ -24,7 +24,7 @@ Where rate control starts cold today: - A color-space change reopens the encoder. - moq-video's own capture publisher also starts its encoder on demand. -Scenarios, each a transcode row in the [harness](/quest/m3/video-quality/harness.md): +Scenarios, each a transcode row in the [harness](/quest/m9/video-quality/harness.md): - Steady live rung from the first frame, as the baseline. - Subscribe mid-clip, unsubscribe, resubscribe. @@ -45,4 +45,4 @@ Public API: none expected. Wire: none. ## Required -- [Harness](/quest/m3/video-quality/harness.md) - the scorer and pop metric +- [Harness](/quest/m9/video-quality/harness.md) - the scorer and pop metric diff --git a/rs/moq-quic/README.md b/rs/moq-quic/README.md index de05d2873c1..f1ef948a470 100644 --- a/rs/moq-quic/README.md +++ b/rs/moq-quic/README.md @@ -29,7 +29,7 @@ Ported since the fork point: - [`aea1e538`](https://github.com/quinn-rs/quinn/commit/aea1e5384927258887119fa1dc4fe3f23f1182c7) ([quinn#2981](https://github.com/quinn-rs/quinn/pull/2981), and [`77a51bdc`](https://github.com/quinn-rs/quinn/commit/77a51bdc5dc8b63bfba30429ab79fbb142bc8e2a) on `0.11.x`): send an MTU probe only on a validated path. A probe is built after the anti-amplification check, so a server probed a migrated peer address beyond three times what it had received from it, spending the budget its `PATH_CHALLENGE` needed. Dependency bumps are not ported. The `0.11.x` accept refactors are already covered by this fork's split acceptance and CID routing. -[`2697de1e`](https://github.com/quinn-rs/quinn/commit/2697de1e1314cb9f98502b7323acd5bdd836b582)'s batched datagram API is not carried: the current WebTransport reader consumes one at a time, and the [planned driver lock profile](../../quest/m4/perf/quic-lock-contention.md) measures contention before adding batching. +[`2697de1e`](https://github.com/quinn-rs/quinn/commit/2697de1e1314cb9f98502b7323acd5bdd836b582)'s batched datagram API is not carried: the current WebTransport reader consumes one at a time, and the [planned driver lock profile](../../quest/m9/perf/quic-lock-contention.md) measures contention before adding batching. Checked through quinn `main` at [`aea1e538`](https://github.com/quinn-rs/quinn/commit/aea1e5384927258887119fa1dc4fe3f23f1182c7), and the `0.11.x` branch through [`77a51bdc`](https://github.com/quinn-rs/quinn/commit/77a51bdc5dc8b63bfba30429ab79fbb142bc8e2a) (2026-11-03); start the next sync there. diff --git a/sh/gh/test-inventory.allow b/sh/gh/test-inventory.allow index 27e36737d82..2617f581584 100644 --- a/sh/gh/test-inventory.allow +++ b/sh/gh/test-inventory.allow @@ -6,7 +6,7 @@ ignore moq-carla-bridge::gpu_video_reaches_the_worker /quest/m5/carla-bridge.md recipe carla::bridge-gpu /quest/m5/carla-bridge.md recipe rs::gpu /quest/m3/gpu-ci.md -ignore moq-video::encode::backend::vaapi::hardware_tests::vaapi_1080p_ladder_stays_on_the_gpu /quest/m4/vaapi-render-tests.md +ignore moq-video::encode::backend::vaapi::hardware_tests::vaapi_1080p_ladder_stays_on_the_gpu /quest/m3/vaapi-render-tests.md ignore moq-video::frame::vulkan::tests::vulkan_gpu_device /quest/m3/gpu-ci.md ignore moq-video::frame::vulkan::tests::vulkan_cuda_slot_reuse_and_teardown /quest/m3/gpu-ci.md ignore moq-video::frame::cuda::tests::vulkan_cuda_auto_external_encode /quest/m3/gpu-ci.md @@ -14,14 +14,14 @@ ignore moq-video::frame::cuda::tests::vulkan_cuda_convert_resize_encode /quest/m ignore moq-video::frame::cuda::tests::vulkan_cuda_three_view_workload /quest/m3/gpu-ci.md ignore moq-video::render::export::vulkan::tests::nvenc_encodes_the_export_in_place /quest/m3/gpu-ci.md ignore moq-video::render::export::vulkan::tests::vulkan_export_round_trips /quest/m3/gpu-ci.md -ignore moq-video::render::renderer::tests::decoded_frames_reach_the_gpu_without_a_download /quest/m4/vaapi-render-tests.md -ignore moq-video::render::renderer::tests::packed_dmabuf_renders_through_vulkan /quest/m4/vaapi-render-tests.md -ignore moq-video::render::renderer::tests::recycled_decoder_surfaces_draw_the_latest_picture /quest/m4/vaapi-render-tests.md -ignore moq-video::render::renderer::tests::the_i420_dmabuf_import_matches_the_cpu_path /quest/m4/vaapi-render-tests.md -ignore moq-video::render::renderer::tests::the_nv12_dmabuf_import_matches_the_cpu_path /quest/m4/vaapi-render-tests.md -ignore moq-video::capture::wgc::native::tests::wgc_displays_reopen_with_cursor_on_and_off /quest/m4/capture-wgc.md -ignore moq-video::capture::wgc::native::tests::wgc_minimized_window_is_rejected_before_the_first_frame /quest/m4/capture-wgc.md -ignore moq-video::capture::wgc::native::tests::wgc_window_reopens_with_cursor_on_and_off /quest/m4/capture-wgc.md +ignore moq-video::render::renderer::tests::decoded_frames_reach_the_gpu_without_a_download /quest/m3/vaapi-render-tests.md +ignore moq-video::render::renderer::tests::packed_dmabuf_renders_through_vulkan /quest/m3/vaapi-render-tests.md +ignore moq-video::render::renderer::tests::recycled_decoder_surfaces_draw_the_latest_picture /quest/m3/vaapi-render-tests.md +ignore moq-video::render::renderer::tests::the_i420_dmabuf_import_matches_the_cpu_path /quest/m3/vaapi-render-tests.md +ignore moq-video::render::renderer::tests::the_nv12_dmabuf_import_matches_the_cpu_path /quest/m3/vaapi-render-tests.md +ignore moq-video::capture::wgc::native::tests::wgc_displays_reopen_with_cursor_on_and_off /quest/m3/capture-wgc.md +ignore moq-video::capture::wgc::native::tests::wgc_minimized_window_is_rejected_before_the_first_frame /quest/m3/capture-wgc.md +ignore moq-video::capture::wgc::native::tests::wgc_window_reopens_with_cursor_on_and_off /quest/m3/capture-wgc.md ignore moq-video::capture::pipewire::camera::tests::pipewire_camera_captures_frames /quest/m5/video-hardware.md ignore moq-video::capture::pipewire::camera::tests::pipewire_camera_shares_a_running_stream /quest/m5/video-hardware.md ignore moq-video::capture::pipewire::tests::portal_captures_frames /quest/m5/video-hardware.md