diff --git a/quest/m0/README.md b/quest/m0/README.md index bd3d9b90d05..ddcf80111f1 100644 --- a/quest/m0/README.md +++ b/quest/m0/README.md @@ -38,4 +38,5 @@ since cutting segments from first-frame pts waits on nothing in m1. - [More tests under load](/quest/m0/test-flakes-2/README.md) - load-only and queue failures, one quest per flake, fixed at the cause, since each one fails a whole batch - [Backport the dot-segment and auth-redirect refusals](/quest/m0/backport-release-hardening.md) - `release`'s relay refuses dot-segment paths and its auth client refuses redirects - [HLS export on a declared grid](/quest/m0/hls-media-grid.md) - the catalog's `hls` field and record tracks go away, and the export numbers segments on `media.join.interval` from media tracks +- [Agents use quest's worktree hooks](/quest/m0/quest-worktree-hooks.md) - adopt `quest worktree hook` once it's shown to coexist with T3 Code's worktrees - [HLS lists contributions](/quest/m0/hls-contributions.md) - an HLS master lists each catalog contribution's renditions, and a contribution's media is subscribed only while players fetch it diff --git a/quest/m0/ci/mbx-r2-verify.md b/quest/m0/ci/mbx-r2-verify.md index 2ee49c8852d..8efed88add4 100644 --- a/quest/m0/ci/mbx-r2-verify.md +++ b/quest/m0/ci/mbx-r2-verify.md @@ -66,8 +66,8 @@ compilation is "not looked up". earlier lockfile state ever matched. The build then fell back to the newest local manifest, which was the install's. Since 1.28.0 an install builds under its own workspace, so the install order does not matter; it stays - where `main` has it. Windows waits on an mbx release that includes #870 and - a bump of the `version` pin in `.github/actions/rust-cache/action.yml`. + where `main` has it. mbx v1.29.1 (2026-11-01) includes #870, so bump the + `version` pin in `.github/actions/rust-cache/action.yml` from 1.28.0 to it. Then check that the first scheduled Nightly on it loads a manifest of more than 124 predictions and hits on Windows. - macOS: on 2026-10-28 the build did inherit, prefetching 732 predictions, diff --git a/quest/m0/hls-media-grid.md b/quest/m0/hls-media-grid.md index 08c6ea5882f..1f76219187d 100644 --- a/quest/m0/hls-media-grid.md +++ b/quest/m0/hls-media-grid.md @@ -82,6 +82,14 @@ deleted, `clock` moves to `media.clock`, and `moq_mux::hls` shrinks to declaring the interval. Wire: hang drops the `hls` field and the record tracks, adds `media.join.interval`, and moves `clock` to `media.clock`. +Decided 2026-11-02 (maintainer, /quest-unblock), a follow-up of +[moq-dev/moq#7387](https://github.com/moq-dev/moq/pull/7387): the grid export's +media playlists advertise `EXT-X-INDEPENDENT-SEGMENTS` whenever every listed +segment opens on a sync sample. No renderer emits it today, and +[export-sync-flags](/quest/m2/intra-refresh/export-sync-flags.md) assumes it, so +that quest only has to omit it for warmup renditions. #7387's conformance probe +then checks the tag against each segment's first sample. + ## Related - [Consumers by start](/quest/m1/time-addressing/consumers.md) - later swaps the first-frame pts for group starts diff --git a/quest/m0/quest-worktree-hooks.md b/quest/m0/quest-worktree-hooks.md new file mode 100644 index 00000000000..59c519d7eb8 --- /dev/null +++ b/quest/m0/quest-worktree-hooks.md @@ -0,0 +1,20 @@ +# [S] Agents use quest's worktree hooks + +## Goal + +moq's `.claude/settings.json` carries the `WorktreeCreate`, `WorktreeRemove`, +`SubagentStart` and `SubagentStop` hooks that `quest init` installs (running +`quest worktree hook`), so agent worktrees are created and cleaned up the way +quest expects, without breaking T3 Code's own worktree management. + +## Plan + +Planned 2026-11-02 (maintainer, /quest-unblock), from +[moq-dev/moq#7644](https://github.com/moq-dev/moq/pull/7644): the quest pin bump +left the hooks out because they change how agent worktrees are made here, and +moq never adopted them after moq-dev/quest#350. + +- First confirm the hooks coexist with T3 Code's worktree threads and its + isolated delegate checkouts (create, hand off, and clean up one of each with + the hooks on). If they conflict, stop and report rather than adopting. +- Then add them with `quest init` and document the behavior in AGENTS.md. diff --git a/quest/m0/test-flakes-2/ietf22-fallback.md b/quest/m0/test-flakes-2/ietf22-fallback.md index bd9d3aac3d0..343654e621c 100644 --- a/quest/m0/test-flakes-2/ietf22-fallback.md +++ b/quest/m0/test-flakes-2/ietf22-fallback.md @@ -8,14 +8,18 @@ lane's derived 1039 ms bound. It keeps its session, and its audio resumes with its video. Both failures are fixed at their cause once one recurs with enough evidence to reproduce it deterministically. +## Waiting + +- moq-dev/moq: either failure recurs with enough evidence to isolate a + deterministic cause. Advance it by running `just test media --media-cases + replacement --media-engines chromium` with + `RUST_LOG=moq_net=debug,moq_relay=debug`, and read any failure here with its + relay and browser logs. + ## Plan -Waiting on a recurrence: run `just test media --media-cases replacement ---media-engines chromium` with `RUST_LOG=moq_net=debug,moq_relay=debug`, and -read any failure here with its relay and browser logs. Neither failure has a -known cause or a reproduction yet, so no production or test change is -justified until one recurs with enough evidence to isolate a deterministic -cause. +Neither failure has a known cause or a reproduction yet, so no production or +test change is justified until one recurs. Two failures were seen while landing [moq-dev/moq#6548](https://github.com/moq-dev/moq/pull/6548), on a host at diff --git a/quest/m1/audio-jitter-target/capped-budgets.md b/quest/m1/audio-jitter-target/capped-budgets.md index 93e48938308..67402e58e40 100644 --- a/quest/m1/audio-jitter-target/capped-budgets.md +++ b/quest/m1/audio-jitter-target/capped-budgets.md @@ -9,13 +9,15 @@ about as much audio a minute as the Opus `bursty` rows, and the Opus `near-zero` row holding a 60 ms target with no underrun episodes. `test/audio-quality/budgets.json` carries every browser row's ceilings re-measured from them. -## Plan +## Waiting + +- moq-dev/moq: seven scheduled Nightly runs on `main` from 2026-10-29 on, the + seventh due about 2026-11-04. Check with `gh run download -n + audio-quality-run` on each; never dispatch the nightly by hand. The artifacts + expire after 7 days, so download each run's as it lands rather than waiting + for the seventh. -Waits on the nightlies: check the `audio-quality-run` artifacts of the -scheduled Nightly runs on `main` (`gh run download -n audio-quality-run`) -once seven have run. Never dispatch the nightly by hand. The artifacts expire -after 7 days, so download each run's as it lands rather than waiting for the -seventh. +## Plan Decided 2026-10-30 (maintainer): the count restarts at the 2026-10-29 nightly, the first after diff --git a/quest/m1/bench-ci.md b/quest/m1/bench-ci.md index fd50a8e0477..9ba42a82795 100644 --- a/quest/m1/bench-ci.md +++ b/quest/m1/bench-ci.md @@ -7,10 +7,12 @@ The paired Rust Criterion comparison each Nightly runs (see regressions: its threshold, confirmation budget, and round count come from A/A runs on the hosted ARM runners, and a deliberately slowed bench fails the job. -## Plan +## Waiting + +- Maintainer: dispatch the Nightly runs on `main` below (A/A and slowed + bench); agents never dispatch Nightly. Check each run's `paired.json`. -Waits on maintainer-dispatched Nightly runs on `main` after the paired mode -lands; agents never dispatch Nightly. +## Plan - A/A: dispatch Nightly on `main` with `bench-base` set to the run's own commit, at least twice. Read each paired group's `paired.json` (the diff --git a/quest/m1/cluster-routing/README.md b/quest/m1/cluster-routing/README.md index 39db5927b8e..a45d387a821 100644 --- a/quest/m1/cluster-routing/README.md +++ b/quest/m1/cluster-routing/README.md @@ -39,9 +39,9 @@ Decided with the maintainer, 2026-10-13: refresh from the origin; instant failover is not required. Measure failure detection, additional routing recovery, and total interruption separately. - An origin declares its own routing identity, fresh per process or tab - restart and shared across its live sessions. It is unverified: recognizing - a routing identity needs key proof, which - [Peer grants](/quest/m2/p2p/peer-grant.md) owns. Receiver-stamped + restart and shared across its live sessions. It is unverified. Decided 2026-11-02 (maintainer, /quest-unblock): no + key proof is planned; if a use case needs a provable identity, moq-lite-07's + route identity becomes a public key, independent of P2P. Receiver-stamped per-session identities were dropped 2026-10-26 (maintainer): in a mesh, two relays stamping their sessions from one origin forward its routes as two identities a third relay cannot recognize as one. Keep administrative diff --git a/quest/m1/cluster-routing/routes.md b/quest/m1/cluster-routing/routes.md index cf31bd17727..b7192ab4c77 100644 --- a/quest/m1/cluster-routing/routes.md +++ b/quest/m1/cluster-routing/routes.md @@ -117,9 +117,10 @@ The handoff contract below, decided 2026-10-16, settles claim freshness: drain. Existing readers resume only across a matching explicit epoch; epochless changes preserve Restart. - Fresh routing identity per process/tab restart, declared by the origin and - shared across its live sessions, decided 2026-10-13. It is unverified: - recognizing a routing identity needs key proof, which - [Peer grants](/quest/m2/p2p/peer-grant.md) owns (decided 2026-10-25). + shared across its live sessions, decided 2026-10-13. It is unverified. + Decided 2026-11-02 (maintainer, /quest-unblock): peer grants are dropped and no key proof is + planned; a provable identity, if a use case needs one, makes this route + identity a public key, independent of P2P. Decided 2026-10-26 (maintainer): receivers do not stamp a fresh identity per session. In a mesh, two relays stamping their sessions from one origin forward its routes as two identities a third relay cannot recognize as diff --git a/quest/m1/crates-trusted-publisher.md b/quest/m1/crates-trusted-publisher.md index 4e792dc663c..25f65999fea 100644 --- a/quest/m1/crates-trusted-publisher.md +++ b/quest/m1/crates-trusted-publisher.md @@ -2,14 +2,17 @@ ## Goal -Condition: every crate that `release-rs.yml` publishes has a crates.io trusted -publisher for `moq-dev/moq`, workflow `release-rs.yml` (and its environment, -if the job uses one). This is a maintainer action in each crate's crates.io -settings and blocks -[Rust releases publish through OIDC](/quest/m1/release-rs-oidc.md). +Every crate that `release-rs.yml` publishes has a crates.io trusted publisher +for `moq-dev/moq`, workflow `release-rs.yml`, and environment `publish`, so +[Rust releases publish through OIDC](/quest/m1/release-rs-oidc.md) can start. +Delete this quest once every crate has one. -Check: for each crate that release-plz publishes, its crates.io settings page -lists that configuration. Delete this quest once every crate has one. +## Waiting + +- Maintainer: add the trusted publisher in each crate's crates.io settings, + after a first publish by token for the crates not yet on crates.io (listed + under Plan). Settings pages need a crates.io login, so check with a + `release-rs.yml` run that publishes through OIDC. ## Plan @@ -22,3 +25,16 @@ Decided 2026-10-22: the human step is split from the workflow change so it stays visible as ready work. A crate that has never been published still needs its first publish done by hand before it can have a trusted publisher; list any such crate here. + +Decided 2026-11-02 (maintainer, /quest-unblock): the maintainer is adding the publishers now, +each with repository `moq-dev/moq`, workflow `release-rs.yml`, and environment +`publish` (the job runs in it). Already on crates.io (31): hang, kio, libmoq, +moq-archive, moq-audio, moq-auth, moq-boy, moq-cli, moq-ffi, moq-gst, moq-hls, +moq-json, moq-loc, moq-msf, moq-mux, moq-native, moq-net, moq-nvenc, +moq-pattern, moq-relay, moq-rtc, moq-rtmp, moq-sock, moq-srt, moq-stats, +moq-tokio, moq-transcode, moq-uring, moq-v4l, moq-video, qmux. Not yet +published (6), so their first publish needs the token and they get a trusted +publisher after it: kio-macros, moq-c, moq-congestion, moq-ice, moq-quic, +moq-rtsp. `CARGO_REGISTRY_TOKEN` is revoked only after all of them have one. +Settings pages need a crates.io login, so the check is a `release-rs.yml` run +publishing through OIDC. diff --git a/quest/m1/lite07-finalize.md b/quest/m1/lite07-finalize.md index 3b78290acbc..81e60f8342f 100644 --- a/quest/m1/lite07-finalize.md +++ b/quest/m1/lite07-finalize.md @@ -35,8 +35,8 @@ owns the tier/partition correctness tests, so they run before the cut. Peer grants and metadata-only subscriptions do not gate it: neither is clustering stability. Both are additive and, per the freeze policy, target whichever lite wip is current when they land. Decided 2026-10-25: no routing identity proof -gates the cut; the key and its proof belong to Peer grants, and lite-07 -routing identities stay declared and unverified. Decided 2026-10-26 +gates the cut, and lite-07 routing identities stay declared and unverified +(2026-11-02: peer grants are dropped and no key proof is planned). Decided 2026-10-26 (maintainer): receivers do not stamp per-session identities either, since that splits one origin into several identities across a relay mesh. diff --git a/quest/m1/moxygen-pin-bump.md b/quest/m1/moxygen-pin-bump.md index 35f5a21e86a..0195ee07c71 100644 --- a/quest/m1/moxygen-pin-bump.md +++ b/quest/m1/moxygen-pin-bump.md @@ -2,16 +2,16 @@ ## Goal -Waiting on upstream Moxygen. Once -[facebookexperimental/moxygen#264](https://github.com/facebookexperimental/moxygen/pull/264) -merges, `nix/moxygen.nix` pins a Moxygen that includes it, +`nix/moxygen.nix` pins a Moxygen that includes +[facebookexperimental/moxygen#264](https://github.com/facebookexperimental/moxygen/pull/264), `nix/moxygen-track-status.patch` is deleted, and the Moxygen interop and -conformance lanes still pass. - -Check moxygen#264's state. Delete this quest once the patch is gone. +conformance lanes still pass. Delete this quest once the patch is gone. ## Plan +Cleared 2026-11-02 (/quest-unblock): Meta's bot merged moxygen#264 as `aaa50a77` on 2026-10-30 (GitHub shows the PR closed). Our pin `c27ae635` (2026-10-23) predates it, so the bump is ready to start. + + The patch is on `main`: [moq-dev/moq#6729](https://github.com/moq-dev/moq/pull/6729) (merged 2026-10-25) added it to `nix/moxygen.nix`'s `patches`, carried until upstream diff --git a/quest/m1/vendor-sync/uniffi.md b/quest/m1/vendor-sync/uniffi.md index d62a2185337..c10135da88c 100644 --- a/quest/m1/vendor-sync/uniffi.md +++ b/quest/m1/vendor-sync/uniffi.md @@ -19,7 +19,7 @@ check that moq's generated bindings need, and moq pins the updated forks: Recurring child of the [vendor sync](/quest/m1/vendor-sync/README.md), which says how each run works and recurs. Decided 2026-10-30 (maintainer): covered -only until [the rust2 switch](/quest/m2/uniffi-fork.md) lands. After that, +only until [the switch to rust2's exporter](/quest/m2/ffi-exporter.md) lands. After that, rust2 owns the generators, and the next run drops this child. Until then, a fix landed here goes upstream into rust2's fork too, or rust2's import loses it. diff --git a/quest/m2/README.md b/quest/m2/README.md index 2c708e54fd9..a4736d1651d 100644 --- a/quest/m2/README.md +++ b/quest/m2/README.md @@ -35,8 +35,9 @@ lite-07; bitrate enforcement is not part of this release. - [Weekly mutation testing on critical modules](/quest/m2/mutants.md) - cargo-mutants runs weekly on moq-auth and moq-net's permission, resume, and expiry code - [Video quality](/quest/m2/video-quality/README.md) - a nightly VMAF harness grades every encoder backend and on-demand GoP pops against floors, and moq-boy's encode is fixed under it - [P2P](/quest/m2/p2p/README.md) - application-selected peer sessions use shared priorities, routing, and same-epoch handover -- [rust2 publishes its UniFFI fork](/quest/m2/rust2-uniffi.md) - waiting on moq-dev/rust2 to import UniFFI and moq's generators and publish them -- [Switch to rust2's UniFFI](/quest/m2/uniffi-fork.md) - moq-ffi and every wrapper build from moq-dev/rust2's fork under one pin, and moq's generator code is deleted +- [moq-auth drops the unused peer grant API](/quest/m2/auth-drop-peer-grant.md) - P2P trusts its peers, so `PeerGrant` and its `KeySet` methods go +- [rust2's exporter is ready for moq](/quest/m2/rust2-exporter.md) - waiting on moq-dev/rust2's exporter to generate every language moq ships +- [Switch to rust2's exporter](/quest/m2/ffi-exporter.md) - moq's bindings come from rust2-ffi, moq-ffi shrinks to its remainder, and moq's UniFFI generator code is deleted - [Generated C bindings](/quest/m2/c/README.md) - C generated from moq-ffi ships as `moq-c` 0.8.0 and replaces the hand-written crate - [A C origin can discover eagerly, like a relay](/quest/m2/c-origin-relay.md) - the C API puts an origin in relay mode, so forwarding a client session's routes needs no local announce watcher - [A runtime-free QUIC driver](/quest/m2/quic-sansio-driver.md) - a caller's own loop moves UDP datagrams and polls moq sessions over moq-quic, for rust2's bindings diff --git a/quest/m2/ashpd-maintained.md b/quest/m2/ashpd-maintained.md index fa41f944fca..bd40d5d8276 100644 --- a/quest/m2/ashpd-maintained.md +++ b/quest/m2/ashpd-maintained.md @@ -7,12 +7,16 @@ pulls the retired zbus_names, zvariant, or zvariant_derive crates. Upgrade ashpd and remove RUSTSEC-2026-0370, RUSTSEC-2026-0371, and RUSTSEC-2026-0372 from `deny.toml`; `just rs audit` passes without those exceptions. +## Waiting + +- bilelmoussaoui/ashpd: a stable release carries the + [zbus 6 port](https://github.com/bilelmoussaoui/ashpd/pull/491) (merged + 2026-10-25; 0.13.13 was still the latest stable). Check crates.io. + ## Plan -The upstream [zbus 6 port](https://github.com/bilelmoussaoui/ashpd/pull/491) -merged on 2026-10-25, but ashpd 0.13.13 is still the latest stable release. -Check crates.io for one that carries it. Until a stable port is published, report that outside -condition without disabling portal capture or adding a local fork. +Until a stable port is published, wait without disabling portal capture or +adding a local fork. ## Related diff --git a/quest/m2/askama-maintained.md b/quest/m2/askama-maintained.md index aba8060cfc7..3b095495bf9 100644 --- a/quest/m2/askama-maintained.md +++ b/quest/m2/askama-maintained.md @@ -17,4 +17,4 @@ its template syntax. Verify binding generation after the upgrade. ## Related - [Dependency audit](/quest/m2/audit-deps.md) - recheck the rest of the dependency graph -- [Switch to rust2's UniFFI](/quest/m2/uniffi-fork.md) - askama reaches the lock only through `uniffi_bindgen`, which that switch moves to rust2 +- [Switch to rust2's exporter](/quest/m2/ffi-exporter.md) - askama reaches the lock only through `uniffi_bindgen`, which that switch deletes diff --git a/quest/m2/auth-drop-peer-grant.md b/quest/m2/auth-drop-peer-grant.md new file mode 100644 index 00000000000..63e971e1372 --- /dev/null +++ b/quest/m2/auth-drop-peer-grant.md @@ -0,0 +1,20 @@ +# [XS] moq-auth drops the unused peer grant API + +## Goal + +`moq-auth` no longer carries the P2P peer grant: `PeerGrant` +(`rs/moq-auth/src/peer.rs`), `KeySet::sign_peer`/`verify_peer`, +`KeySet::to_peer_set`, `PeerGrant::binds`, and their wire type are deleted, +along with the "Peer grants" section of `doc/bin/relay/auth.md` and the crate +docs that mention them. No other crate uses them. + +## Plan + +Planned 2026-11-02 (maintainer, /quest-unblock): P2P trusts its peers, so +relay-issued, key-bound peer grants are dropped until a real use case asks for +per-peer scopes ([Signaling and policy](/quest/m2/p2p/signal.md) records why). +The landed API has no consumer left. git history keeps it if a use case +returns. + +Public API: removes `moq_auth::PeerGrant` and the `KeySet` peer methods +(breaking for moq-auth); note it in its MIGRATE. Wire: none. diff --git a/quest/m2/browser-receive-credit.md b/quest/m2/browser-receive-credit.md index 88d01ab5ff0..2f2c2a1fb06 100644 --- a/quest/m2/browser-receive-credit.md +++ b/quest/m2/browser-receive-credit.md @@ -2,10 +2,15 @@ ## Goal -Condition: WebTransport tells the page how much receive credit a stream it has -stopped reading still holds, in the spec and in shipping Chrome and Firefox. -Then [JS fastest-reader pacing](/quest/m2/js-fastest-reader.md) can pause -groups safely. Delete this quest once the condition holds. +WebTransport tells the page how much receive credit a stream it has stopped +reading still holds, so [JS fastest-reader pacing](/quest/m2/js-fastest-reader.md) +can pause groups safely. Delete this quest once the condition holds. + +## Waiting + +- w3c/webtransport: the spec, and shipping Chrome and Firefox, expose that + credit. Check with the steps under Plan. Decided 2026-11-02 (maintainer, + /quest-unblock): no W3C issue is filed for now. ## Plan diff --git a/quest/m2/bun-check.md b/quest/m2/bun-check.md index ec6752d8b4f..004bb835d34 100644 --- a/quest/m2/bun-check.md +++ b/quest/m2/bun-check.md @@ -8,10 +8,11 @@ Every workspace package's `check` script, and the bench typechecks in ## Plan -Waits on nixpkgs shipping bun 1.4.4 or later (nixpkgs-unstable carried 1.4.2 -on 2026-10-22). `bun check` arrived in 1.4.3, but 1.4.4 adds the -`--no-script` the `check` scripts below need and fixes 1.4.3's false errors. -Bump `flake.lock` then; no bun overlay before it (decided 2026-10-22). +nixpkgs-unstable carries bun 1.4.4 as of 2026-11-02 (checked in +/quest-unblock), so start with a `flake.lock` bump; `main`'s dev shell still +has 1.4.2. `bun check` arrived in 1.4.3, but 1.4.4 adds the `--no-script` the +`check` scripts below need and fixes 1.4.3's false errors. No bun overlay +(decided 2026-10-22). Measured 2026-10-22, bun 1.4.4 against TypeScript 7.0.2 on `main`: identical errors and exit codes on all 33 `tsc` invocations in the `check` scripts and diff --git a/quest/m2/c/README.md b/quest/m2/c/README.md index 1b1ec4a442f..2979fde4076 100644 --- a/quest/m2/c/README.md +++ b/quest/m2/c/README.md @@ -40,7 +40,7 @@ The `moq-c` package (`c/moq`, `just c check`) builds the generated header and its probe in CI, so drift fails loud. moq-ffi's record `MoqAudioEncoderFinished` was renamed from `MoqAudioEncoderFinish` so moq's current generator emits the header; -[the switch](/quest/m2/uniffi-fork.md) reverts that rename. No `c-v*` tag ships +[the switch](/quest/m2/ffi-exporter.md) reverts that rename. No `c-v*` tag ships before the cut. The line owns the pre-publication check: every `doc/lib/c` sample and the C @@ -50,6 +50,6 @@ publication; [old-crate retirement](/quest/m2/release/c-retire.md) follows. ## Required -- [Switch to rust2's UniFFI](/quest/m2/uniffi-fork.md) - the C generator comes from rust2's fork +- [Switch to rust2's exporter](/quest/m2/ffi-exporter.md) - the C ABI comes from rust2's exporter instead of a UniFFI generator - [C consumers](/quest/m2/c/consumers.md) - the C interop client and `doc/lib/c` samples move onto the generated API - [Async failure contract](/quest/m2/c/async-failure-contract.md) - every generated async call returns a task and calls back exactly once, cancelled when its task was freed first diff --git a/quest/m2/c/async-failure-contract.md b/quest/m2/c/async-failure-contract.md index 71414022b28..6b12b14f901 100644 --- a/quest/m2/c/async-failure-contract.md +++ b/quest/m2/c/async-failure-contract.md @@ -4,7 +4,9 @@ Every async call in the generated C API has a callback contract a C caller can rely on for freeing `user_data`: the call returns a task, never NULL, and its callback runs exactly once, with the result or, when the task was freed first, with `cancelled` set. -[moq-dev/rust2#169](https://github.com/moq-dev/rust2/pull/169) implements it in rust2's C generator. Waiting on [Switch to rust2's UniFFI](/quest/m2/uniffi-fork.md) to move the C generator pin: `c/moq` adopts the new callback shape then, along with the CamelCase names, and `c/moq/test/probe.c`'s callbacks take `cancelled` and release `user_data` there. Delete this quest with that PR. +[moq-dev/rust2#169](https://github.com/moq-dev/rust2/pull/169) implements it in rust2's C generator. Waiting on [Switch to rust2's exporter](/quest/m2/ffi-exporter.md) to move the C generator pin: `c/moq` adopts the new callback shape then, along with the CamelCase names, and `c/moq/test/probe.c`'s callbacks take `cancelled` and release `user_data` there. Delete this quest with that PR. + +Decided 2026-11-02 (maintainer, /quest-unblock): moq switches to rust2's exporter, not the UniFFI fork, so this contract must hold for the exporter's C ABI futures; rust2#169's fork implementation is the reference. ## Plan diff --git a/quest/m2/c/consumers.md b/quest/m2/c/consumers.md index 88e4c96d55f..781c3f25577 100644 --- a/quest/m2/c/consumers.md +++ b/quest/m2/c/consumers.md @@ -40,4 +40,4 @@ so it can move only once the generated archive publishes, after this quest; ## Required - [Breaking release cut](/quest/m2/release/cut.md) - the hand-written 0.7 ships with this client and these docs as its proof -- [Switch to rust2's UniFFI](/quest/m2/uniffi-fork.md) - the generated C type names settle before consumers are written against them +- [Switch to rust2's exporter](/quest/m2/ffi-exporter.md) - the generated C type names settle before consumers are written against them diff --git a/quest/m2/cpal-alsa-runtime.md b/quest/m2/cpal-alsa-runtime.md index 256db13f380..b0869e69b28 100644 --- a/quest/m2/cpal-alsa-runtime.md +++ b/quest/m2/cpal-alsa-runtime.md @@ -2,22 +2,24 @@ ## Goal -Condition: a cpal release on crates.io whose Linux build no longer makes -libasound a load-time requirement, either by loading it at runtime (dlopen) -or by putting the ALSA host behind a feature that the PulseAudio host can -replace. +A cpal release on crates.io whose Linux build no longer makes libasound a +load-time requirement, either by loading it at runtime (dlopen) or by putting +the ALSA host behind a feature that the PulseAudio host can replace. Once the +release is out, delete this quest. -Check: build a capture and playback `moq-cli` against that release (with -runtime loading, `--features capture`; with an optional host, cpal's ALSA -host off and `pulseaudio` on) and confirm `readelf -d` lists no -`libasound.so.2` under `NEEDED`. A clean `readelf` is not enough on its own: -the binary must also start, list devices, and play on a host without -libasound, as [capture-alsa-link](/quest/m2/capture-alsa-link.md) verifies. +## Waiting -Advance it by proposing the change upstream (RustAudio/cpal, and -diwic/alsa-sys if the loading lives there). Posting there needs maintainer -approval, which is pending: on 2026-10-06 the maintainer chose not to post -yet. Once the release is out, delete this quest. +- RustAudio/cpal: ships that release, as asked in + [RustAudio/cpal#1421](https://github.com/RustAudio/cpal/issues/1421) (or + diwic/alsa-sys, if the loading lives there). Check: build a capture and + playback `moq-cli` against that release (with runtime loading, + `--features capture`; with an optional host, cpal's ALSA host off and + `pulseaudio` on) and confirm `readelf -d` lists no `libasound.so.2` under + `NEEDED`. A clean `readelf` is not enough on its own: the binary must also + start, list devices, and play on a host without libasound, as + [capture-alsa-link](/quest/m2/capture-alsa-link.md) verifies. + +## Plan Decided 2026-10-08: the capture chain (this quest, [capture-alsa-link](/quest/m2/capture-alsa-link.md), and @@ -25,7 +27,7 @@ Decided 2026-10-08: the capture chain (this quest, until cpal ships runtime ALSA loading. Nothing is posted upstream and no fork is taken meanwhile. -## Plan +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) both depend on `alsa` unconditionally on Linux and the BSDs. `alsa` 0.12.1 diff --git a/quest/m2/crossbeam-pin-fence.md b/quest/m2/crossbeam-pin-fence.md index 47000acfaf8..d9839694653 100644 --- a/quest/m2/crossbeam-pin-fence.md +++ b/quest/m2/crossbeam-pin-fence.md @@ -2,14 +2,18 @@ ## Goal -Waiting on upstream crossbeam. Once -[crossbeam-rs/crossbeam#1352](https://github.com/crossbeam-rs/crossbeam/issues/1352) -is fixed in a released crossbeam-epoch, the workspace takes that release and +The workspace takes a crossbeam-epoch release that fixes `Local::pin`, and kio's `fence(Release)` in `Read::drop` (`rs/kio/src/log.rs`) is deleted, with kio's log tests still passing on ARM. -Check #1352's state and crossbeam-epoch's latest release. Delete this quest if -upstream declines the fix. +## Waiting + +- crossbeam-rs/crossbeam: a released crossbeam-epoch fixes + [crossbeam-rs/crossbeam#1352](https://github.com/crossbeam-rs/crossbeam/issues/1352) + (the fix is proposed in + [crossbeam-rs/crossbeam#1355](https://github.com/crossbeam-rs/crossbeam/pull/1355)). + Check #1352's state and crossbeam-epoch's latest release. Delete this quest + if upstream declines the fix. ## Plan diff --git a/quest/m2/uniffi-fork.md b/quest/m2/ffi-exporter.md similarity index 75% rename from quest/m2/uniffi-fork.md rename to quest/m2/ffi-exporter.md index 6f784be3000..2b07ba685be 100644 --- a/quest/m2/uniffi-fork.md +++ b/quest/m2/ffi-exporter.md @@ -1,18 +1,32 @@ -# [M] Switch moq's bindings to rust2's UniFFI fork +# [L] Switch moq's bindings to rust2's exporter ## Goal -moq-ffi depends on the UniFFI fork's renamed crates from crates.io, and every -generator moq uses (Kotlin, Swift, Python, Go, Dart, C++, and C) comes from -[moq-dev/rust2](https://github.com/moq-dev/rust2) through one pinned flake -input. `rs/uniffi-bindgen`, `nix/uniffi-dart-*`, and the per-generator pins are -deleted, and the three `kixelated/uniffi-*` generator forks are archived. -Generated bindings are unchanged by the switch -itself, except C's type names and the one record rename the C name rule -undoes. +moq's bindings (Kotlin, Swift, Python, Go, Dart, C++, and C) come from +[rust2's exporter](https://github.com/moq-dev/rust2/blob/main/quest/m1/ffi-export/README.md) +(`rust2-ffi`), which exports moq's crates' `pub` API as a C ABI with no +attributes. moq-ffi shrinks to the remainder that is genuinely new API. +`rs/uniffi-bindgen`, `nix/uniffi-dart-*`, the per-generator pins, and the +UniFFI dependency are deleted, and the three `kixelated/uniffi-*` generator +forks are archived. Wrapper APIs do not change except where a decision below +says so. ## Plan +Decided 2026-11-02 (maintainer, /quest-unblock): moq skips rust2's UniFFI fork and switches straight to the +exporter. rust2 decided on 2026-10-25, from +[moq-dev/rust2#41](https://github.com/moq-dev/rust2/pull/41), that the exporter +replaces the fork, which stays in maintenance; publishing `rust2-uniffi` to +crates.io would commit permanent names to code already slated for +retirement, and switching moq to it would be churn on the way to the +exporter. Until the switch, moq keeps building from its pinned +`kixelated/uniffi-*` forks ([UniFFI generator forks are current](/quest/m1/vendor-sync/uniffi.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): + + Decided with the maintainer on 2026-10-13, after a review of how moq uses UniFFI: @@ -83,7 +97,7 @@ none. ## Required -- [rust2 publishes its UniFFI fork](/quest/m2/rust2-uniffi.md) - the crates and generators this switches to +- [rust2's exporter is ready for moq](/quest/m2/rust2-exporter.md) - the exporter generates every language moq ships and moq can pin it ## Related diff --git a/quest/m2/ffi-poll-migrate.md b/quest/m2/ffi-poll-migrate.md index 91ef76c574b..06277f985ec 100644 --- a/quest/m2/ffi-poll-migrate.md +++ b/quest/m2/ffi-poll-migrate.md @@ -18,7 +18,7 @@ once this lands. - **No UniFFI change on this path**: dropping a future at `rust_future_cancel` (mozilla/uniffi-rs#2771) is optional in - [the fork](/quest/m2/uniffi-fork.md). Shape A has no read lane for an + [the exporter switch](/quest/m2/ffi-exporter.md). Shape A has no read lane for an unfreed future to hold, so this quest must not wait for it. - **What the spike found**: in Go, Shape A cut a ready `read_frame` from 2.3 to 1.2 µs, and where every read parks it raised throughput 13 to 32 percent diff --git a/quest/m2/go-cancelled-result.md b/quest/m2/go-cancelled-result.md index d51bcf5f84b..96986f0f84a 100644 --- a/quest/m2/go-cancelled-result.md +++ b/quest/m2/go-cancelled-result.md @@ -36,7 +36,7 @@ they are. A binding that destroys the result returns `ctx.Err()`, which `run` never releases, so nothing is destroyed twice. Check whether `bridge.CallHandle` still covers anything once the binding releases these results itself. If the -[generator hard fork](/quest/m2/uniffi-fork.md) has landed, the change goes +[exporter switch](/quest/m2/ffi-exporter.md) has landed, the change goes there; otherwise into the current fork, with a pin bump. Test it with a cancelled context racing a call that completes at once, and diff --git a/quest/m2/mp4-repair-lock.md b/quest/m2/mp4-repair-lock.md index ad067eb8862..09dd2babab7 100644 --- a/quest/m2/mp4-repair-lock.md +++ b/quest/m2/mp4-repair-lock.md @@ -6,7 +6,7 @@ ## Plan -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. This waits on #7560 landing; `moq repair mp4` and `moq_mux::container::mp4::repair` don't exist on `main` until then. +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. diff --git a/quest/m2/p2p/README.md b/quest/m2/p2p/README.md index 3ba0af6db5c..9df550b173c 100644 --- a/quest/m2/p2p/README.md +++ b/quest/m2/p2p/README.md @@ -71,8 +71,7 @@ off by default, and used only when the application lists it. ### Policy is the application's -`Peers` takes the shared origin, a reconnect-safe relay auth handle for -[peer grants](/quest/m2/p2p/peer-grant.md), plus: +`Peers` takes the shared origin, plus: - `enabled`: an explicit on/off signal, off by default; - `prefix`: the reserved prefix, `.p2p/` under the session root by default; @@ -119,12 +118,19 @@ Routing is the [cluster routing line](/quest/m1/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. -A routing node id is declared by its origin and unverified; recognizing it -needs the key proof [Peer grants](/quest/m2/p2p/peer-grant.md) owns -(decided 2026-10-26, maintainer, dropping receiver-stamped per-session -identities). Switching between a CDN and a direct path resumes only across -a matching explicit epoch. [Signaling and policy](/quest/m2/p2p/signal.md) -owns the open question of what key maps announces to roster peers. +A routing node id is declared by its origin and unverified (decided +2026-10-26, maintainer, dropping receiver-stamped per-session identities). +Switching between a CDN and a direct path resumes only across a matching +explicit epoch. + +Decided 2026-11-02 (maintainer, /quest-unblock): P2P trusts its peers. Joining the prefix takes the +relay's token, so a peer is as trusted as any publisher that token admits, +and a direct session serves without a peer grant or key proof. Peer grants and +P2P identity planning are dropped until a real use case asks for proof. Each +peer's roster `` is its public key, like an iroh endpoint id; nothing proves +it yet. Which roster peer serves an announce stays the application's explicit +selection. If identity or signing is wanted later, moq-lite-07's route +identity becomes a public key, independent of P2P. Decided 2026-10-13: priority then total cost selects among authorized routes; a direct connection does not always win. A watching tab never re-serves what @@ -155,14 +161,12 @@ Extractor declaration checks, which analyze only published packages since ## Required -- [Peer grants](/quest/m2/p2p/peer-grant.md) - the relay issues a key-bound, asymmetrically signed grant a browser can verify; HS256 keys issue none - [Signaling and policy](/quest/m2/p2p/signal.md) - opted-in peers find each other under the prefix, the application picks who to dial, and the roster-size gate decides whether STUN is used - [Native pairs over ICE and QUIC](/quest/m2/p2p/native-quic.md) - two native peers connect through ICE, then `moq-quic` over the nominated path with fingerprints pinned from the offer - [moq-cli joins](/quest/m2/p2p/cli.md) - `--p2p` publishes a roster entry, dials native peers over ICE and QUIC, and serves browsers as a transit hop - [Peer route policy](/quest/m2/p2p/cost-scopes.md) - application priorities and total cost choose eligible routes, with epoch-aware handover - [Watch opts in](/quest/m2/p2p/watch.md) - one attribute turns it on in the demo and the watcher migrates to the best eligible route - [Harness](/quest/m2/p2p/harness.md) - the Playwright harness and the numbers behind every mapping decision -- [Plan P2P identity](/quest/m2/p2p/identity-plan.md) - condition: the maintainer settles the roster key, exploring moq-e2ee signing first ## Related diff --git a/quest/m2/p2p/cost-scopes.md b/quest/m2/p2p/cost-scopes.md index ef8f0764665..04a75e30bd8 100644 --- a/quest/m2/p2p/cost-scopes.md +++ b/quest/m2/p2p/cost-scopes.md @@ -24,10 +24,10 @@ Decided 2026-10-13, replacing the unconditional direct-first policy: the CDN through a peer whose uplink died, off for conference offload. Browser tabs never relay, whatever the backend (decided 2026-10-22, maintainer). A neighboring watcher is not a source of its received - media. A publisher's routing identity is declared and unverified until - [Peer grants](/quest/m2/p2p/peer-grant.md) key proof; switching between + media. A publisher's routing identity is declared and unverified (peers are + trusted, decided 2026-11-02 (maintainer, /quest-unblock)); switching between its CDN and direct copies resumes only across a matching explicit epoch. - What maps announcements to roster peers is open in + Which roster peer serves an announcement is the application's selection in [Signaling and policy](/quest/m2/p2p/signal.md). - Matching explicit epochs permit shared logical-track resume. Different or missing epochs require Restart and fresh resolution. Keep CDN control and diff --git a/quest/m2/p2p/identity-plan.md b/quest/m2/p2p/identity-plan.md deleted file mode 100644 index 71fd144dc73..00000000000 --- a/quest/m2/p2p/identity-plan.md +++ /dev/null @@ -1,23 +0,0 @@ -# [S] Plan P2P identity - -## Goal - -Condition: the maintainer runs `/quest-plan` on P2P identity and the -[Signaling and policy](/quest/m2/p2p/signal.md) and -[Peer grants](/quest/m2/p2p/peer-grant.md) quests reflect the result. Delete -this quest then. - -## Plan - -- Parked 2026-10-25 (maintainer): the roster-key question that - [moq-dev/moq#6310](https://github.com/moq-dev/moq/pull/6310) left open in - `signal.md` (what names a roster entry and maps an announce to a peer, - while routing identities are declared and unverified) needs a walkthrough - before any choice. -- Maintainer direction to explore first: when peers don't trust each other - and need proof, extend moq-e2ee to sign content (not only encrypt it), so - proof rides the media rather than a routing identity. - -## Related - -- [E2EE](/quest/m2/e2ee/README.md) - where signing would live diff --git a/quest/m2/p2p/native-quic.md b/quest/m2/p2p/native-quic.md index 7ad753f0edd..fa85ec93ede 100644 --- a/quest/m2/p2p/native-quic.md +++ b/quest/m2/p2p/native-quic.md @@ -27,8 +27,8 @@ Decided with the maintainer on 2026-10-13: - QUIC is `moq-quic`, fed datagrams from the nominated pair. - Authentication: each side generates a self-signed certificate and sends its SHA-256 fingerprint in its offer or answer, as DTLS fingerprints ride - in SDP. Both sides pin it with the existing `tls.fingerprint` check. MoQ - AUTH with a [peer grant](/quest/m2/p2p/peer-grant.md) still runs on top. + in SDP. Both sides pin it with the existing `tls.fingerprint` check. Peers + are trusted (decided 2026-11-02 (maintainer, /quest-unblock)), so no peer grant runs on top. - After nomination the session's socket carries ICE consent checks and QUIC, split by `moq_sock::demux::classify`: `Ice` and `Stun` go to the ICE agent and `Quic` to the endpoint. The tokio listener's `quic::demux` diff --git a/quest/m2/p2p/peer-grant.md b/quest/m2/p2p/peer-grant.md deleted file mode 100644 index 52c560da98a..00000000000 --- a/quest/m2/p2p/peer-grant.md +++ /dev/null @@ -1,88 +0,0 @@ -# [M] Peer grants - -## Goal - -The relay issues each session a short-lived, key-bound grant that another -client can verify without learning a signing key. A browser presents that -grant on a direct session and serves only the paths it names. HMAC keys -never leave the relay: a peer grant is signed with an asymmetric key -(ES256 or EdDSA), and only the public JWK is distributed. An operator with -only HS256 keys issues no peer grant, and a session with none serves -nothing on a direct connection. - -## Plan - -`moq_auth::Claims` stays the bearer token the relay verifies. A peer grant -is a separate type (`moq_auth::PeerGrant`) so presenting one as `?jwt=` -fails verification rather than widening a relay session. It carries the -session's publish and subscribe patterns, a presenter public key (`cnf`), -`iat`, and a short `exp`, and no routing identity. Landed in `moq-auth`: -the type (a `root` plus patterns, `cnf.jwk` per RFC 7800), ES256/EdDSA-only -`KeySet::sign_peer`/`verify_peer`, the public `KeySet::to_peer_set`, and -`PeerGrant::binds` for the roster and proven-key checks, with the docs. -What remains is the relay issuance, the wire, the possession proof, and -`@moq/auth`. The presenter key is -public on the roster, so a copied JWT is not enough: the presenter proves -possession of the matching private key on the direct AUTH stream. A roster peer that -received someone else's grant cannot replay it. - -Issuance: the relay owns AUTH (`rs/moq-relay/src/auth/refresh.rs`). The -client mints one key per process or tab and announces its public half on the -relay's AUTH stream (AUTH_CNF); it is the `cnf`. Then, after AUTH_OK: - -- `AUTH_KEYS { Jwks (b) }`, the relay's kid-indexed public JWKS, - asymmetric keys only, HMAC material never included; -- `AUTH_PEER { Grant (b) }`, the peer-grant JWT. - -AUTH_KEYS and AUTH_PEER may repeat for refresh and key rotation. The lite -draft and both language codecs gain AUTH_KEYS and AUTH_PEER in whichever lite -wip is current when this lands; they do not gate the lite-07 cut (decided -2026-10-13). Decided 2026-10-25 (maintainer): this quest owns AUTH_CNF and -the possession proof, which do not gate the cut either, since lite-07 -routing identities stay declared and unverified; `just drafts check` passes. Reissue before expiry; the client -replaces the grant on each live peer session with `auth().add` and drops -the old one. A key change mints a new grant; the old key is dead. - -Verification on a direct session: the opener writes `AUTH { Token (b) }` -with the JWT and proves possession of the `cnf` key bound to this session. -The proof prototype ([#6310](https://github.com/moq-dev/moq/pull/6310)) found -that mutual fresh nonces alone let a keyless intermediary forward both honest -exchanges. ✅ Decided 2026-10-30 (maintainer): the presenter signs each -routing control the session sends with the `cnf` key, over a session -transcript. This works over WebRTC data channels, which expose no DTLS -exporter, and keeps one code path and one set of replay tests. Binding a -transport exporter was rejected: browsers expose none, so it covered only -native QUIC pairs, and a hybrid would need both. The wire work defines the -transcript and adds the replay tests before the wire is fixed. The -receiver verifies the relay signature against the JWKS from its own -AUTH_KEYS, checks `exp`, matches `cnf` against the presenter key in the -peer's roster entry, and requires `cnf` to equal the key the session proved. -The origin then serves only the granted paths. A missing, expired, -HMAC-signed, or unproven grant, or one whose `cnf` is not the session's -proven key, is not a grant. - -The grant authorizes paths and binds a key; it carries no routing identity. -A routing identity is declared by its origin and unverified -([Routes and announces](/quest/m1/cluster-routing/routes.md)). Decided -2026-10-26 (maintainer): recognizing a routing identity needs key proof, and -this quest owns the key and its proof; receiver-stamped per-session -identities were dropped because they split one origin into several -identities across a relay mesh. Binding a routing identity to the key is -not planned yet. Switching between a CDN and a direct path resumes only -across a matching explicit epoch. - -[Signaling and policy](/quest/m2/p2p/signal.md) presents the grant in band -on each direct session. P2P is its only consumer, so it lives on this line -rather than the auth line (decided in the 2026-09-30 audit). - -Docs: `doc/bin/relay/auth.md` states that peer grants need an asymmetric -key, that HS256 operators get none, and that the public JWKS is not a -secret; update its last paragraph once the relay issues them. Tests -remaining, on the wire: a key that does not match the -peer's roster entry is refused; a -grant whose `cnf` is not the session's proven key is refused; a grant copied onto another -session without the private key is refused; copying both grant and a prior -valid proof onto another session is refused; a refreshed grant replaces -the old one without dropping the session. - -Additive. diff --git a/quest/m2/p2p/signal.md b/quest/m2/p2p/signal.md index e424101b536..713bdb4d49c 100644 --- a/quest/m2/p2p/signal.md +++ b/quest/m2/p2p/signal.md @@ -16,29 +16,23 @@ relay connection's auth, and the policy knobs from the `iceServers`, `max`, `meta`, plus application-owned selection and per-connection priority/cost. Decided 2026-10-13: only explicitly selected peers are dialed; there is no default dial-all. Prefer a selected set/handle to a callback, -settling the public shape with its consumer. The origin carries no auth state -and the pooled `Connection` hides `Established.auth`, yet peer grants need -the presenter key announced and AUTH_KEYS/AUTH_PEER received across relay reconnects, so -define that handle before fixing `Peers`. Publishing into the origin reaches +settling the public shape with its consumer. Publishing into the origin reaches the relay and its announce stream is the roster. Tests inject transport factories and cover pairing, application selection, and the roster gate. -Identity: the route-layer node id is declared by the origin and unverified -([routes](/quest/m1/cluster-routing/routes.md)); recognizing it needs the -key proof [peer grants](/quest/m2/p2p/peer-grant.md) owns, so it cannot be -trusted as the roster id on its own. Switching between a CDN and a direct -path resumes only across a matching explicit epoch. Open, to settle here: -what names a roster entry (`` below, also the perfect-negotiation -tiebreak) and what key maps an announce to a roster peer. An announce -reaching a client through the CDN carries the broadcast path, its epoch, -and the publisher's declared identity. Candidates the plans already hold are the -presenter key in the roster entry, which the peer grant binds, and the -grant's publish patterns, which are verified only after connecting. Since -the application selects peers, it may need no automatic mapping at all. +Identity, decided 2026-11-02 (maintainer, /quest-unblock): each peer generates a keypair, and its +roster `` (also the perfect-negotiation tiebreak) is its public key, like an +iroh endpoint id. Nothing proves the key yet: peers are trusted (see Trust). +Which roster peer serves an announce is the application's explicit selection; +moq maps nothing automatically. The route-layer node id stays declared and +unverified ([routes](/quest/m1/cluster-routing/routes.md)); if a use case +needs provable identity or signing, moq-lite-07's route identity becomes a +public key, independent of P2P. P2P identity planning and peer grants were +dropped until a real use case asks for them. Switching between a CDN and a +direct path resumes only across a matching explicit epoch. Roster: each peer publishes `` with an `info.json` snapshot track: -the moq ALPNs it accepts, `webrtc: true`, the presenter public key that -[peer grants](/quest/m2/p2p/peer-grant.md) bind, the application's `meta`, +the moq ALPNs it accepts, `webrtc: true`, the application's `meta`, and for native peers an optional WebTransport entry with an ordered list of URLs and the listener's certificate fingerprint, plus the field [native pairs](/quest/m2/p2p/native-quic.md) add for ICE+QUIC. @@ -83,23 +77,11 @@ transport on the dialing side and `accept` on the answering side, both with `publish: origin.consume()` and `consume: origin`, so the tab serves what it publishes, never what it receives (a tab is not transit). -Trust: publishing under the prefix proves only that the relay admitted the -peer to the prefix, not that it may read everything this tab can. Tokens -within one project carry different subscribe scopes, so a peer session is -scoped like a relay session, but the relay token itself never crosses it: a -`moq_auth` JWT is a bearer credential with no audience or proof of -possession, so a peer that received one could replay it against the relay. -Instead the relay issues each session a peer grant, a relay-signed statement -of that session's path scopes bound to its presenter public key and -short-lived, which the peer presents in band with a proof of possession; -the other side verifies the relay's signature, the key against the -roster entry, and that the key is the one the session proved, then serves only the granted paths. Issuance, -asymmetric keys, JWKS, and refresh live in -[Peer grants](/quest/m2/p2p/peer-grant.md), along with the key's proof: HMAC keys cannot be given to -browsers without also letting them forge grants, so an HS256-only relay -issues nothing. A peer session with no verifiable grant serves nothing; -there is no equal-scope shortcut. - -## Required - -- [Peer grants](/quest/m2/p2p/peer-grant.md) - the key-bound, asymmetrically signed credential a direct session presents; HS256 keys issue none +Trust, decided 2026-11-02 (maintainer, /quest-unblock): P2P trusts its peers. Joining the prefix +takes the relay's token, so a peer admitted to the prefix is as trusted as any +publisher that token admits, and a direct session serves what this tab +publishes without a peer grant or key proof. The relay token itself still +never crosses a peer session: a `moq_auth` JWT is a bearer credential, so a +peer that received one could replay it against the relay. Rejected: relay-issued, +key-bound peer grants with a possession proof (the earlier plan), until a use +case needs per-peer scopes. diff --git a/quest/m2/release/relay-loader-adoption.md b/quest/m2/release/relay-loader-adoption.md index fb662c343e6..3976e3ca232 100644 --- a/quest/m2/release/relay-loader-adoption.md +++ b/quest/m2/release/relay-loader-adoption.md @@ -2,19 +2,16 @@ ## Goal -Condition: an embedding consumer's moq-relay pin includes -[moq-dev/moq#6535](https://github.com/moq-dev/moq/pull/6535), which adds -`Config::merge(&CliLayer)`. Check each consumer's lockfile for a moq-relay -version or revision that carries it, and advance a pin that lacks it through -that consumer's normal upstream-first process. - -Then every embedding consumer loads its relay config through +Every embedding consumer loads its relay config through `Config::merge`, an explicit obsolete file or environment spelling refuses its startup, and its embedded CLI-only fields survive the merge. Delete this quest once every embedder is migrated. ## Plan +Cleared 2026-11-02 (/quest-unblock): every embedder's moq pin now carries [moq-dev/moq#6535](https://github.com/moq-dev/moq/pull/6535)'s `Config::merge(&CliLayer)` (moq.pro at `80cafb3a`, studio at `96244dd8`). studio only calls `moq_relay::test_relay()`, so moq.pro is the one embedder to migrate; its `rs/edge/src/config.rs` still does `toml::from_str` and `update_from`. + + Split from the relay config API audit (maintainer, 2026-10-23): the upstream loader landed first, because migrating a consumer before its pin carries the API would break it. diff --git a/quest/m2/release/runner-client-bump.md b/quest/m2/release/runner-client-bump.md index aa7deb9a32e..8924989598e 100644 --- a/quest/m2/release/runner-client-bump.md +++ b/quest/m2/release/runner-client-bump.md @@ -2,15 +2,17 @@ ## Goal -Condition: a moq-tokio release carries -[moq-dev/moq#5083](https://github.com/moq-dev/moq/pull/5083), a breaking -change on `main` that publishes only with the next cut. Check with -`cargo info moq-tokio` and that release's changelog. - -Then bump the moq-dev-rs client in +The moq-dev-rs client in [englishm/moq-interop-runner](https://github.com/englishm/moq-interop-runner) -to it, so it stops attaching a Timestamp property to every object, and delete -this quest once that PR merges. +runs a moq-tokio that stops attaching a Timestamp property to every object. +Delete this quest once that bump merges. + +## Waiting + +- moq-dev/moq: a moq-tokio release carries + [moq-dev/moq#5083](https://github.com/moq-dev/moq/pull/5083), a breaking + change on `main` that publishes only with the next cut. Check with + `cargo info moq-tokio` and that release's changelog. ## Plan diff --git a/quest/m2/release/smoke-bindings-rename.md b/quest/m2/release/smoke-bindings-rename.md index 0c21301a6fb..0f787d2194e 100644 --- a/quest/m2/release/smoke-bindings-rename.md +++ b/quest/m2/release/smoke-bindings-rename.md @@ -2,11 +2,15 @@ ## Goal -Condition: a binding release ships the `request` and `follow` names from -[moq-dev/moq#7193](https://github.com/moq-dev/moq/pull/7193). Check each -registry's latest moq-ffi-based package (Go, Python, Kotlin, Swift) for -`request`/`follow`. Then move `test/smoke/clients/{go,python,kotlin,swift}` -off `request_broadcast` and `announced_broadcast`, and delete this quest. +`test/smoke/clients/{go,python,kotlin,swift}` call `request` and `follow` +instead of `request_broadcast` and `announced_broadcast`. + +## Waiting + +- moq-dev/moq: a binding release ships the `request` and `follow` names from + [moq-dev/moq#7193](https://github.com/moq-dev/moq/pull/7193). Check each + registry's latest moq-ffi-based package (Go, Python, Kotlin, Swift) for + `request`/`follow`. ## Plan diff --git a/quest/m2/rust2-exporter.md b/quest/m2/rust2-exporter.md new file mode 100644 index 00000000000..fe250c569b9 --- /dev/null +++ b/quest/m2/rust2-exporter.md @@ -0,0 +1,21 @@ +# [XS] Waiting: rust2's exporter is ready for moq + +## Goal + +rust2's exporter generates every language moq ships from moq's crates, so +[the switch](/quest/m2/ffi-exporter.md) can start. Delete this quest once the +condition holds, so the switch resurfaces. + +## Waiting + +- moq-dev/rust2: [its exporter epic](https://github.com/moq-dev/rust2/blob/main/quest/m1/ffi-export/README.md) + finishes with a tag or flake revision moq can pin, and moq-ffi built on it + can still publish to crates.io (no git dependency). Check that rust2's + `quest/m1/ffi-export/README.md` is gone from its `main` and its README lists + the generators. + +## Plan + +Decided 2026-11-02 (maintainer, /quest-unblock): this replaces the earlier wait on rust2 publishing its UniFFI +fork. rust2's `quest/m1/uniffi-release.md` is dropped with it: the fork is not +published. diff --git a/quest/m2/rust2-uniffi.md b/quest/m2/rust2-uniffi.md deleted file mode 100644 index 1925b3376e9..00000000000 --- a/quest/m2/rust2-uniffi.md +++ /dev/null @@ -1,16 +0,0 @@ -# [XS] Waiting: rust2 publishes its UniFFI fork - -## Goal - -[moq-dev/rust2](https://github.com/moq-dev/rust2) has merged its -[UniFFI import quest](https://github.com/moq-dev/rust2/blob/main/quest/m1/uniffi.md) -and published `rust2-uniffi` and `rust2-uniffi-bindgen` to crates.io, with a -tag moq can pin as a flake input. Check: `cargo search rust2-uniffi` lists a -release, and rust2's `quest/m1/uniffi.md` is gone from its `main`. Delete this -quest once both hold. - -## Plan - -Decided with the maintainer on 2026-10-23: the UniFFI hard fork and moq's -generator code live in rust2. This quest stands in for that cross-repository -blocker so [the switch](/quest/m2/uniffi-fork.md) resurfaces when it clears. diff --git a/quest/m2/uring-bench-followups.md b/quest/m2/uring-bench-followups.md index f0ecf486149..4f56b2d56c6 100644 --- a/quest/m2/uring-bench-followups.md +++ b/quest/m2/uring-bench-followups.md @@ -6,7 +6,7 @@ Run-to-quiescence on the io_uring workers has acceptance numbers from a quiet ho ## Plan -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%. This waits on [moq-dev/moq#7530](https://github.com/moq-dev/moq/pull/7530) landing. +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. diff --git a/quest/m2/usage-settings-once.md b/quest/m2/usage-settings-once.md index 97fd021a541..e9697761fe4 100644 --- a/quest/m2/usage-settings-once.md +++ b/quest/m2/usage-settings-once.md @@ -9,13 +9,17 @@ 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 +## Waiting + +- jdx/usage: a usage-rs release ships + [jdx/usage#1608](https://github.com/jdx/usage/issues/1608). Check the issue + and usage-rs releases. If upstream declines, delete this quest and keep the + drift test. -Waits on two outside conditions, in order: +## Plan -1. The maintainer approves posting the proposal on jdx/usage. Decided - 2026-10-27 (maintainer interview): ask before posting; no approval yet. -2. Upstream ships it. Check the issue and usage-rs releases. +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 diff --git a/quest/m2/usage-unknown-env.md b/quest/m2/usage-unknown-env.md index 79d73af0648..1a11a62bee1 100644 --- a/quest/m2/usage-unknown-env.md +++ b/quest/m2/usage-unknown-env.md @@ -8,9 +8,15 @@ 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`. +## Waiting + +- jdx/usage: a usage-rs release ships + [jdx/usage#1607](https://github.com/jdx/usage/issues/1607). Check the issue + and usage-rs releases. If upstream declines, delete this quest and keep the + `check_env` helper. + ## Plan -Decided 2026-10-23 (maintainer interview): the maintainer approved posting -an issue or PR on jdx/usage for this. Waiting on jdx/usage is the condition: -check the upstream issue or PR and its releases; delete this quest if -upstream declines, keeping the helper. +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. diff --git a/quest/m2/video-quality/gpu-runners.md b/quest/m2/video-quality/gpu-runners.md index 87f8fb243d4..69d2f728691 100644 --- a/quest/m2/video-quality/gpu-runners.md +++ b/quest/m2/video-quality/gpu-runners.md @@ -25,5 +25,4 @@ online with its labels. Delete this quest once they are listed. cannot protect it. - Follow `nixos`'s sandboxed runner containers where the host allows it, binding in only the GPU device nodes and the host driver's libraries. -- The [rented gate host](/quest/m3/ci-host.md) does not change this. The - quality job is nightly, so its runners can stay at home. +- The quality job is nightly, so its runners can stay at home. diff --git a/quest/m3/README.md b/quest/m3/README.md index c75070d757e..a0a8c6d4795 100644 --- a/quest/m3/README.md +++ b/quest/m3/README.md @@ -77,8 +77,6 @@ m4's upstream waits in here and moved m2 work with no consumer. - [moq-audio in the browser](/quest/m3/audio-web.md) - AudioWorklet capture and playback and WebCodecs Opus and AAC, for the web studio's audio - [moq-audio filter chain](/quest/m3/audio-filters.md) - per-source suppression, gate, compressor, limiter, ducking, and plugins, which moq-dev/studio configures - [Mesh keys](/quest/m3/mesh-keys.md) - parked: encrypted and HMAC-authenticated mesh broadcasts through a wrapped `.keys` track -- [A gate host is rented](/quest/m3/ci-host-rented.md) - condition: the maintainer rents a host for the public merge gates -- [Gate on a rented host](/quest/m3/ci-host.md) - moq, quest, and rust2 gate on one rented box with a warm disk cache instead of the home host - [rust2-ts reaches moq-net's model](/quest/m3/rs2ts-model-ready.md) - rust2's translator is ready to translate the model - [Keep moq-net's model in rust2-ts's subset](/quest/m3/rs2ts-model-subset.md) - safe twins for the model's internal `unsafe`, `dyn Expiry` as an enum, once rust2-ts reaches the model - [Wait for a stable bun lint](/quest/m3/bun-lint-release.md) - condition: `bun lint` ships in a stable Bun that nixpkgs carries diff --git a/quest/m3/bun-lint-release.md b/quest/m3/bun-lint-release.md index 933385f5e08..9f4d1cff6bc 100644 --- a/quest/m3/bun-lint-release.md +++ b/quest/m3/bun-lint-release.md @@ -2,13 +2,18 @@ ## Goal -Condition: a stable Bun release ships `bun lint`, and nixpkgs carries it. -Check the [Bun blog](https://bun.com/blog) or `bun lint --help` on the -nixpkgs-unstable bun; `bun fmt` is already stable since 1.4.4. Delete this -quest once both hold. +A stable Bun with `bun lint` is available from nixpkgs. Delete this quest once +it is. + +## Waiting + +- oven-sh/bun: a stable release ships `bun lint`, and NixOS/nixpkgs carries + it. Check the [Bun blog](https://bun.com/blog) or `bun lint --help` on the + nixpkgs-unstable bun; `bun fmt` is already stable since 1.4.4. ## Plan Announced on 2026-10-15 for "the next version" and only on canary (`bun upgrade --canary`) when written on 2026-10-29; nixpkgs-unstable carried -bun 1.4.2. +bun 1.4.2. On 2026-11-02, nixpkgs-unstable's bun 1.4.4 still answers +`bun lint` with `Script not found "lint"`. diff --git a/quest/m3/ci-host-rented.md b/quest/m3/ci-host-rented.md deleted file mode 100644 index 784f486c5bc..00000000000 --- a/quest/m3/ci-host-rented.md +++ /dev/null @@ -1,15 +0,0 @@ -# [XS] A gate host is rented - -## Goal - -Condition: the maintainer picks a host for the public repositories' merge -gates and rents it, with root SSH reachable for a NixOS install. Check with -`ssh root@ true`. Delete this quest once it passes. - -## Plan - -Shelved 2026-10-26 by the maintainer after the comparison in -[the gate host quest](/quest/m3/ci-host.md). Recommendation: netcup RS 4000 -G12.5, one month first, then a 12-month term once entry times match the home -host. Before ordering, re-measure the gate's busy hours if `quest land` has -shipped, since batching may change the answer (see that quest's SaaS row). diff --git a/quest/m3/ci-host.md b/quest/m3/ci-host.md deleted file mode 100644 index e1946405641..00000000000 --- a/quest/m3/ci-host.md +++ /dev/null @@ -1,97 +0,0 @@ -# [M] Gate on a rented host - -## Goal - -The merge gates for moq-dev/moq, moq-dev/quest, and moq-dev/rust2 run on -one rented always-on host with a warm on-disk cache, instead of the -maintainer's home `nixos` box. Gate entries are no slower than the home -runner's measured times, and the home box is off the merge path. moq.pro's -gate is out of scope: it stays home for security reasons. - -## Plan - -Decided 2026-10-26 (maintainer): - -- Motivation: uptime (home power, network, reboots), freeing the home - machine, and keeping PR code off the home network. -- Scope: the public repositories only. moq.pro keeps `container@moq-pro` at - home; this quest does not touch it. -- Each repository keeps its own runner process (kixelated is a user account, - so its runners register per repository). All three run concurrently in one - shared systemd-nspawn container, under one user and one mbx store, and mbx's - machine-wide budget splits the CPU. No lock serializes them. Rejected: a - container per repository (three cold caches) and moving the repositories - into moq-dev for one org runner. -- The cache stays on the host's disk: mbx's local store and the host Nix - store, substituting from `nix.moq.dev`. The host writes to neither R2 nor - `nix.moq.dev`. -- The host's NixOS config lives in this repository under `infra/`, - installed with nixos-anywhere. Runner registration tokens are placed by - hand, never committed. -- Once the rented host passes a day of gate runs, the home `moq-queue` - runner is deregistered. Rollback is re-registering it. -- `nix/runner` (the persistent-VM module, rejected 2026-10-25) is deleted, - with its `runner-vm` recipes and flake checks. It can land ahead as its own - PR. Then update `quest/m2/gpu-ci.md`, which assumes the VM module. -- Independent of `quest land`: the host only changes which machine answers - `vars.MOQ_QUEUE_RUNNER`. -- If entries are slower than home, step up the host (more cores, or an OVH - RISE-M), not the VPS tier. - -Measured 2026-10-26 on the home host. `container@moq-ci` is capped at -16 threads and 18 GiB. - -- Over 42 hours, the queue runner was busy 46% (about 330 runner-hours a - month), and more than half the time in 20 of 27 active hours. A Check job - takes a median 3.9 minutes, p90 14.3. -- One hour sampled every 2 s, 86% busy. Cores in use while busy: median 2, - mean 4.0, p90 9.4, p95 15.3. Only 16% of busy time is above 8 cores, and - capping at 8 would add about 7% to job time. So the critical path, and - therefore single-thread speed, sets job time. -- Anonymous memory peaked at 9.4 GiB (p95 3.9), with up to 15 GiB of page - cache. The container uses about 15 GB of disk. - -Hosts compared on 2026-10-26 (provider pages; Geekbench 6 from -vpsbenchmarks). The home 5900XT scores about 2285 single-thread. - -| Host | Cores / RAM / disk | Price/mo | Verdict | -|---|---|---|---| -| netcup RS 4000 G12.5 | 12 dedicated EPYC 9645 (GB6 2141 / 13016), 32 GB, 512 GB | €77.83 (12-month) | Recommended: about home speed, within the ~$80 budget | -| OVH VPS-4 2027 | 8 Haswell-class vCores (GB6 1021 / 5294, stability grade D), 24 GB, 200 GB | ~$27 | Rejected: about 2x slower jobs would saturate the runner; VPS-5/6 add cores, not per-core speed | -| OVH RISE-M | Ryzen 9 9900X 12C/24T, 64 GB, 2x512 GB | €99.99 | Fallback, about 1.34x home single-thread | -| OVH RISE-L, Hetzner AX102 | 16C/32T Zen 4/5, 128 GB | ~€150-160 | Most headroom, over budget | -| Dedicated cloud VMs (Hetzner CCX, OVH c3, AWS, GCP) | 16 vCPU | €250+ | Not competitive always-on | -| Per-minute SaaS (Ubicloud, Namespace, Blacksmith, Depot, GitHub larger runners) | ~16 vCPU | ~$200-950 at 330 h | Break-even with a ~$90 box is about 13% duty; revisit only if `quest land` batching cuts runner-hours several-fold | - -Remaining work, once [the host is rented](/quest/m3/ci-host-rented.md): - -- `infra/` host config: disko, the shared container, three runner services - with their own work directories and labels, and a deploy recipe. -- Size `NEXTEST_TEST_THREADS` and the container's CPU and memory to the host - rather than the home box's 24 threads. -- Make mbx collection safe for concurrent jobs before the runners share a - store. Each moq job runs `mbx gc` before it builds - (`.github/actions/rust-cache`), and under disk pressure that sweep can evict - generated source trees another job's next build reads. Verify with - overlapping jobs under cache pressure. Safe collection need not serialize - jobs. -- Give moq-dev/quest and moq-dev/rust2 a gate that runs on the rented - runner. Their Check workflows hardcode hosted runners (`ubuntu-latest`, - `ubuntu-24.04-arm`), and quest's has no `merge_group` trigger, so a runner - variable alone leaves them off the host. Route only trusted gate runs - (queue or `quest land`) to it, never PR pushes. That is a PR in each of - those repositories. -- Register the runners, point moq's `vars.MOQ_QUEUE_RUNNER` at its runner, - and compare a day of entry times against the numbers above. -- Retire the home `moq-queue` runner and its container definition. -- Update `infra/README.md` and the CI epic's runner decision. - -Public API: none. Wire: none. - -## Required - -- [A gate host is rented](/quest/m3/ci-host-rented.md) - condition: the maintainer rents the host; then this starts - -## Related - -- [CI cost and latency](/quest/m0/ci/README.md) - the epic whose queue runner this moves diff --git a/quest/m3/interop-runner-approval.md b/quest/m3/interop-runner-approval.md index 7b27f81424a..27ef8619179 100644 --- a/quest/m3/interop-runner-approval.md +++ b/quest/m3/interop-runner-approval.md @@ -2,15 +2,16 @@ ## Goal -Condition: a released moq-tokio and `@moq/net` carry the session `close()` -that withdraws this session's announcements before disconnecting. Then the -moq.dev clients in englishm/moq-interop-runner (`builds/moq-dev-rs`, -`builds/moq-dev-js`) bump to that release and end each publishing session -with `close()` instead of aborting. +The moq.dev clients in englishm/moq-interop-runner (`builds/moq-dev-rs`, +`builds/moq-dev-js`) end each publishing session with a `close()` that +withdraws its announcements, instead of aborting. Delete this quest once they +use that release's `close()`. -Check: read the moq-tokio and `@moq/net` changelogs for a release with the -withdrawing `close()`. Delete this quest once the runner's moq.dev clients use that release's -`close()`. +## Waiting + +- moq-dev/moq: a released moq-tokio and `@moq/net` carry the session `close()` + that withdraws this session's announcements before disconnecting. Check the + moq-tokio and `@moq/net` changelogs for such a release. ## Plan