diff --git a/quest/m0/README.md b/quest/m0/README.md index a1453262e52..065f6551d4f 100644 --- a/quest/m0/README.md +++ b/quest/m0/README.md @@ -64,6 +64,7 @@ fidelity: a red Interop hides real regressions. ## Required +- [CI cache release](/quest/m0/ci-cache-release.md) - a stable cache server supports anonymous reads and OIDC writes restricted to the main warmer - [Self-hosted CI](/quest/m0/self-hosted-ci.md) - same-repo Check and Test run on a self-hosted NixOS runner with a main-written local cache, behind a `CI_RUNNER` kill switch - [CI host](/quest/m0/ci-host.md) - the maintainer brings up the spare desktop as the `moq-ci` and `moq-gpu` runner host - [Draft-22 media on 0.17](/quest/m0/release-22/README.md) - a 0.17.x with the LOCATION_FILTER and FIRST_OBJECT fixes and moq-noq 1.3.5, before Seattle, then draft-20+ FETCH diff --git a/quest/m0/ci-cache-release.md b/quest/m0/ci-cache-release.md new file mode 100644 index 00000000000..58f6f7452fc --- /dev/null +++ b/quest/m0/ci-cache-release.md @@ -0,0 +1,25 @@ +# [XS] A stable cache server enforces the CI write policy + +## Goal + +Condition: a stable `jdx/mr-boxington-cache` release supports anonymous reads +and server-enforced GitHub OIDC writes restricted to the main cache warmer, +as [Self-hosted CI](/quest/m0/self-hosted-ci.md) requires. + +## Plan + +Decided in the quest audit: wait for a stable release rather than select an +unreleased pin or weaken the write boundary. The latest stable release, +v0.1.1, allows writes in anonymous mode; [#5218](https://github.com/moq-dev/moq/pull/5218) +therefore keeps workflow routing hosted while preparing the runner module. + +Check the upstream releases and their authentication policy. Advance this +condition by identifying a stable release that supports the required split; +record its version and policy evidence when deleting this quest. The +self-hosted CI quest owns packaging it and testing that anonymous and PR-ref +writes fail while the main warmer can write. Keep `CI_RUNNER` unset until +that implementation and its validation are complete. + +## Related + +- [Self-hosted CI](/quest/m0/self-hosted-ci.md) - packages and proves the cache policy before enabling workflow routing diff --git a/quest/m0/self-hosted-ci.md b/quest/m0/self-hosted-ci.md index 58123199563..3dcc0b4b634 100644 --- a/quest/m0/self-hosted-ci.md +++ b/quest/m0/self-hosted-ci.md @@ -106,7 +106,11 @@ group) with `CI_RUNNER` set and unset, and that the module evaluates (`nix eval` and with a token from a PR ref). End-to-end proof happens in [CI host](/quest/m0/ci-host.md). -Public API: none. Wire: none. +Public API: the NixOS runner module and its configuration options. Wire: none. + +## Required + +- [CI cache release](/quest/m0/ci-cache-release.md) - a stable server supports the enforced read/write split; keep routing hosted until its policy is proven ## Related diff --git a/quest/m1/README.md b/quest/m1/README.md index e7f6d811a19..2e3202bfe0c 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -76,7 +76,6 @@ disabled rung strands WHEP, HLS, and moq-mux export viewers. - [Controlled time](/quest/m1/time/README.md) - every crate and package reads time through moq-time or @moq/time, and every test runs on a controlled clock; ranks above the fork, whose switch adopts its instant - [Hard fork](/quest/m1/quic/fork/README.md) - quinn hard-forked in-tree as `moq-quic`, ranked ahead of perf; the rest of the QUIC line follows it - [Cluster routing](/quest/m1/cluster-routing/README.md) - any node routes toward a broadcast's origin over CDN and P2P links alike, with per-origin routes and path-less announces -- [Flat questlines](/quest/m1/quest-flat-lines.md) - moq pins the current quest CLI, lands its questline branches on main, and retires them - [Binding audio delay](/quest/m1/binding-surface.md) - moq-ffi and every wrapper configure and observe audio playout delay - [Per-frame arrivals in @moq/watch](/quest/m1/watch-arrivals.md) - the audio and video decoders expose a window of recent frame arrivals, late and skipped marks included, as a signal - [Bump web-transport-iroh for the capsule close](/quest/m1/iroh-capsule-bump.md) - condition: moq-dev/web-transport#419 ships in a release, then moq's iroh HTTP/3 client reports a peer's close capsule @@ -161,7 +160,7 @@ disabled rung strands WHEP, HLS, and moq-mux export viewers. - [BBR ACK cleanup](/quest/m1/bbr-ack-cleanup.md) - packet bookkeeping scales with completed entries instead of scanning the flight on every ACK - [Perf](/quest/m1/perf/README.md) - eliminate measured hot-path costs across moq-uring, kio, and the moq-net model - [Benchmark regressions in CI](/quest/m1/bench-ci.md) - PRs get a non-blocking comparison of the Criterion benches they affect, and a nightly trend on main alerts on regressions -- [Mergeable bench buckets](/quest/m1/bench-buckets.md) - moq-bench emits per-interval latency buckets that sum across processes and hosts +- [Backport mergeable bench buckets](/quest/m1/bench-buckets.md) - release receives main's per-interval latency buckets and their tests - [Relay session bench](/quest/m1/bench-relay.md) - the same scenario through moq-relay's own connection handling - [Read-only lookup](/quest/m1/read-only-lookup.md) - a live track lookup stops waking the front and every demand watcher - [Rust-backed @moq/net](/quest/m1/rs2ts/README.md) - measure browser baselines, complete WASM bindings, and adopt Rust behind @moq/net before the later JS-only stage diff --git a/quest/m1/archive/replay-catalog.md b/quest/m1/archive/replay-catalog.md index c61f9d820c3..5659c547d06 100644 --- a/quest/m1/archive/replay-catalog.md +++ b/quest/m1/archive/replay-catalog.md @@ -64,10 +64,10 @@ this by hand-building a catalog. A live playlist stays capped even for a durable timeline; listing past the window is the explicit `moq_hls::export::Config::history` mode, which `moq export hls` does not expose yet. Give it a flag here. History mode lists from -the records the timeline restates when the exporter joins (at most 256 from a -`moq-mux` publisher), so "the whole recording" also needs the exporter to -read a longer recording's stored history rather than only its live -restatement. +stored timeline groups, including records older than the live restatement +(#5236). Reuse that recovery in the end-to-end proof rather than implement it +again. [Retained-floor history](/quest/m1/archive/history-floor.md) and +[HLS failure signaling](/quest/m1/hls-failure-signal.md) remain separate work. Add a CI test that records a broadcast longer than the default window, replays it, and asserts the stock exporter lists segment 0 with a durable diff --git a/quest/m1/bench-buckets.md b/quest/m1/bench-buckets.md index 780c05b77bd..bb1aa083c6d 100644 --- a/quest/m1/bench-buckets.md +++ b/quest/m1/bench-buckets.md @@ -1,51 +1,24 @@ -# [S] moq-bench: mergeable interval buckets +# [S] Backport moq-bench's mergeable interval buckets ## Goal -Every `moq-bench` JSONL line carries that interval's latency histogram in a -versioned, documented bucket layout, so the buckets from several bench -processes on several hosts sum element-wise over any window of whole report -intervals into one distribution; a window cutting through an interval cannot -separate that interval's samples. A percentile computed once from the summed buckets matches one -computed from the concatenated samples binned by the same layout and overflow -rule: the buckets are lossy, so the match is exact at bucket resolution, not -at raw-sample precision. +The `release` line carries main's mergeable interval latency histograms, so +load harnesses can combine several bench processes over whole report +intervals. Preserve the version-1 sparse millisecond buckets, interval +bounds, documentation, and regression tests already landed on `main`. ## Plan -Main-side draft implements the sparse version 1 histogram and interval bounds. -Remaining: after the main PR merges, backport the landed commit and any missing -#5021 interval delta to release in a separate draft. Do not cherry-pick an -unlanded main draft; keep this quest until the backport exists. - - -Decided 2026-10-05 by the maintainer: percentiles neither window nor merge, -and a load harness sums several generators against one relay, so the bench -emits mergeable buckets. - -- `Latency` (`rs/moq-bench/src/stats.rs`) counts 1 ms buckets for the whole - run and diffs them each interval for the `latency_interval_*` percentiles - ([#5021](https://github.com/moq-dev/moq/pull/5021)); this quest emits that - same delta in a mergeable layout, so the two share one representation. -- 60,001 dense buckets per line is too much JSON. Pick a compact layout (sparse - non-zero buckets, or a log-linear layout with bounded relative error) and - record why. -- The output documents: the layout version, units, the interval's start and - end, the sample count, and what an empty or partial interval (the first and - the last) looks like. -- Harness rule: publisher and subscriber run on one host, so latency closes - on one clock and buckets from several hosts merge without skew; - `latency_clock_skew` keeps flagging a violation. Replace the README's - advice to NTP-sync separate publisher and subscriber hosts - (`rs/moq-bench/README.md`, "Machine-readable output") with this rule. -- A test sums fixture buckets from two runs and checks the percentiles - against the concatenated samples, binned by the same layout and the same - overflow bucket (60,000 ms and above today) before the percentile is - taken. Update `rs/moq-bench/README.md`. - -Lands on main, then is backported to `release` as an additive cherry-pick -PR, like [#4882](https://github.com/moq-dev/moq/pull/4882), since downstream load -harnesses track `release` (decided 2026-10-05). The backport carries #5021's -delta along if `release` does not have it yet. +[#5268](https://github.com/moq-dev/moq/pull/5268) landed as `0a47269be`. +Decided in the quest audit: only its release backport remains; do not reopen +the histogram layout decision. Backport that commit in a separate PR to +`release`, including [#5021](https://github.com/moq-dev/moq/pull/5021)'s interval +delta if it is still absent there. Downstream load harnesses track `release`, +which is why the maintainer requested this backport on 2026-10-05. + +Run the landed histogram merge and interval tests on the backport and keep +the same JSONL contract: merging is exact at bucket resolution, a partial +interval cannot be split, and publisher/subscriber measurements share one +host clock. Preserve the README's layout and clock documentation. Public API: none (CLI output only). Wire: none. diff --git a/quest/m1/c/README.md b/quest/m1/c/README.md index 05be1c6b0a3..0d9993169cc 100644 --- a/quest/m1/c/README.md +++ b/quest/m1/c/README.md @@ -43,6 +43,10 @@ client build and run against the released 0.8.0 archive, not only in-tree. The hand-written crate gets no more feature work: its shutdown, CMake library, and fetch quests were abandoned for this line, and hidden is done. +Decided in the quest audit: [claim epochs](/quest/m1/claim-answer-epochs.md) +is a narrow correctness exception. Its existing claim API migrates with +Rust and JS so C cannot retain the unsafe accept path while this line waits. +That change owns its C tests and upgrade notes; other API reshaping stays here. ## Required diff --git a/quest/m1/claim-answer-epochs.md b/quest/m1/claim-answer-epochs.md index 4a28eb1d08b..6997edad0ba 100644 --- a/quest/m1/claim-answer-epochs.md +++ b/quest/m1/claim-answer-epochs.md @@ -95,6 +95,13 @@ Public API: breaking. A claim request is answered through `prepare` and `publish`, and a public claim takes no epoch, in Rust, JS, moq-ffi, moq-c and every binding wrapper, plus the claim docs. Wire: none. +Decided in the quest audit: migrate the hand-written C claim API in this +change as a narrow correctness exception to the +[C freeze](/quest/m1/c/README.md). Leaving its existing accept path would +preserve the identity bug, so claim safety does not wait for generated C. +This permits only the claim migration and its tests and upgrade notes; +the broader C API still waits for the generated replacement. + ## Related - [Announcement takeover](/quest/m1/announce-takeover.md) - its claim-to-route operation is `Request::prepare` then `publish`; TAKEOVER then removes the cold-start reset diff --git a/quest/m1/dropped-sources.md b/quest/m1/dropped-sources.md index a63efbfd815..82c189daed0 100644 --- a/quest/m1/dropped-sources.md +++ b/quest/m1/dropped-sources.md @@ -12,8 +12,8 @@ Consumers never decide behavior by matching an error; errors are diagnostics. - Rust already maps IETF `PUBLISH_DONE` Unauthorized to `Error::Unauthorized`, and a closed source's standing route no longer re-requests it. -- #4179 fixes revoked upstream subscriptions and #4120 preserves session death - and resumed-track errors. After #4179 reaches `main`, verify the remaining +- #4179 fixed revoked upstream subscriptions and #4120 preserves session death + and resumed-track errors. Both are on `main`; verify the remaining track paths (source close, route removal, broadcast withdrawal) locally and over a mock session, and map JS `PUBLISH_DONE` Unauthorized to #4179's shared error. Preserve causes at the source rather than remapping `Dropped` at @@ -79,10 +79,6 @@ Consumers never decide behavior by matching an error; errors are diagnostics. Public API: none expected; error values consumers observe change. Wire: none. -## Required - -- [Auth](/quest/m1/auth/README.md) - its Unauthorized quest (#4179, done on the line) supplies shared Unauthorized errors and revoked-stream handling - ## Closes - [#5347](https://github.com/moq-dev/moq/issues/5347) - `export ts --stitch --linger` exits on `internal error` when a catalog request lands on a dying publisher diff --git a/quest/m1/ffi-shape/README.md b/quest/m1/ffi-shape/README.md index c9d7c104951..a95780f56d7 100644 --- a/quest/m1/ffi-shape/README.md +++ b/quest/m1/ffi-shape/README.md @@ -14,9 +14,8 @@ The line landed on `main` in #4519 (2026-10-09) with its json/flate, net, media, and request-accept children, and the remaining children PR straight to `main`. Decided by the maintainer, 2026-10-09: -- Land now rather than keep collecting on the line branch, because questline - branches are retiring ([Flat questlines](/quest/m1/quest-flat-lines.md)) and - each `main` merge meant hand-porting moq-ffi changes onto moved code. +- Children PR directly to `main`, so moving a binding layer does not require + hand-porting concurrent moq-ffi changes onto a separate line branch. - The bindings still break once per release, not once per merge: binding releases (moq-ffi and the Python, Go, Swift, Kotlin, Dart, and C++ packages) wait until [Codecs](/quest/m1/ffi-shape/codec.md) merges @@ -59,7 +58,9 @@ Settled shape: hand-written moq-c is out of scope: the [generated C](/quest/m1/c/README.md) and C++ bindings inherit this shape from moq-ffi, so reshaping the hand-written C - ABI would break C users twice. + ABI would break C users twice. The narrow + [claim-epoch correctness exception](/quest/m1/claim-answer-epochs.md) does + not expand this layer reshaping to the hand-written ABI. - `rs/moq-ffi/AGENTS.md` records the per-language namespace pattern `json` set. diff --git a/quest/m1/quest-flat-lines.md b/quest/m1/quest-flat-lines.md deleted file mode 100644 index 4f839fa0ecf..00000000000 --- a/quest/m1/quest-flat-lines.md +++ /dev/null @@ -1,71 +0,0 @@ -# [M] moq adopts flat questlines from the current quest CLI - -## Goal - -moq pins a quest release where questlines are planning groups with no branch -of their own: every quest, including a questline README's remaining work, is -its own PR against `main`. The child work already merged into questline -branches reaches `main`, and those branches retire. - -## Plan - -kixelated/quest#44 dropped questline branches, `quest branch`, and -`quest ready --remote`; kixelated/quest#51 changed the import skill to treat a -report's fix as a claim and plan each issue fully. moq pins kixelated/quest -`677f8d1`, which includes both; `quest check` passed on the tree unchanged. - -Decided 2026-10-04: - -- A quest of its own, not a routine pin bump, because questline branches - carried merged child work that is not on `main`. -- Goal confirmed: questlines become planning groups whose children PR straight - to `main`. Finishing any line's remaining children is a non-goal. -- Bump the CLI now, not after the lines land, because the point is to get rid - of line branches. -- Each code-carrying line branch gets `main` merged in and its umbrella PR - marked ready; the maintainer lands them later with `/quest-complete`. Merging - `main` in is less churn than re-cutting each child, and finishing a line - first would keep its branch alive. -- Code-free lines (quic #3975, cluster-routing #4654, broadcast-epoch #4805) - are closed and their branches deleted; their children PR to `main`. Nothing - on them needs landing. -- Nested lines (rs2ts/sans-io #4438, archive/track-timeline #4255) fold into - their parent branch, then close, so each tree lands through one PR. -- The quest CLI's PR-adoption skill is installed as `quest-iterate` and - replaces the `takeover` skill, so one skill covers it. kixelated/quest#52 - renamed it upstream to match. -- `CONTRIBUTING.md` drops ", a questline" so the merge-commit rule names only - `release`. - -Interview paper trail (✅ marks the choice): - -- Goal confirm: ✅ Confirm / Also finish lines / Bump only. -- Landing default: Land partial now / Re-cut per child / Finish then land / - ✅ (user) "merge main into the branch, mark PR as ready for review, then I'll - do quest-complete later". -- Split: Per-line quests / One quest / Batch by size / ✅ (user) "spawn - sub-agents to merge main for all of them, then mark as ready". -- Order: Bump last / Bump first / Bump first, sync quest files / ✅ (user) "the - idea is to get rid of line branches", then confirmed ✅ Bump now. -- Empty lines: ✅ Close and retarget / Merge main and ready. -- Nested: ✅ Fold into parent / Retarget to main / Merge main, keep nesting. -- Takeover: Add stub, replace takeover / Add stub, keep both / Skip / ✅ (user) - "add stub, replace takeover, rename it to quest-iterate". -- CONTRIBUTING: ✅ Yes, edit it / Leave it. - -Remaining: - -- Land each line's umbrella PR with `/quest-complete`, then delete its branch. - Landed so far: #4403 wildcard, #4034 archive (with #4255 folded in), #4080 - obs-moq-video, #4162 audio-jitter-target, #4180 transport-upgrade, #4437 - rs2ts (with #4438 folded in), #4640 tstd, #4653 test-flakes-2, #4079 cpp, - #4519 ffi-shape, and #4039 auth (UNAUTHORIZED moved to 0x3B). No umbrella - PR remains open. -- #4133 qos closed without landing (2026-10-09). Its egress `lag` histogram - is owned by [Viewer lag histogram](/quest/m1/qos/lag-histogram.md), which - ports it once stats prefix tracks land. Do not reopen the umbrella PR. -- Check for leftover questline branches and retire them only once their work - is merged or assigned to an outstanding quest. - -Done when `flake.nix` pins the new quest, no `quest/*README` branch remains, -and `quest check` passes. diff --git a/quest/m1/stats/rust.md b/quest/m1/stats/rust.md index 450e7054be2..5f94bb403fb 100644 --- a/quest/m1/stats/rust.md +++ b/quest/m1/stats/rust.md @@ -26,9 +26,11 @@ `rs/moq-audio/src/playback` is already counted privately. - `moq-cli` folds the counters into `hang::stats::Snapshot` on the stats interval and writes it through `moq_json::snapshot`, with the `.z` sibling. - A TS import flattens `ts::stats::Snapshot` in as `mpegts`, and a TS - export flattens its egress counters into `.echo` the same way. `transport` comes from - the connection's `ConnectionStats`. + `transport` comes from the connection's `ConnectionStats`. + Decided in the quest audit: this quest owns generic per-rendition reporting; + [TS health stats](/quest/m2/ts-health-stats.md) owns serializing and flattening + TS container counters under `mpegts` in publisher stats and export `.echo`. + That integration stays in m2 so this quest does not duplicate its work. - Feedback: after reading a catalog with an `echo` section, the player resolves the echo path against the broadcast, appends `.echo`, and publishes there with the fixed feedback @@ -53,4 +55,5 @@ ## Related +- [TS health stats](/quest/m2/ts-health-stats.md) - adds TS-specific counters after this generic reporting plumbing lands - [Encrypted native CLI](/quest/m1/e2ee/cli.md) - refuses stats and feedback on an encrypted broadcast diff --git a/quest/m1/tstd/mux-overhead.md b/quest/m1/tstd/mux-overhead.md index f1b604c7197..6226dbb3c92 100644 --- a/quest/m1/tstd/mux-overhead.md +++ b/quest/m1/tstd/mux-overhead.md @@ -27,8 +27,9 @@ Decided 2026-10-12: into one PES while the PES fits the PID's B, timed by the first frame's PTS. This applies to every audio stream the export frames (AAC, MP2, AC-3, E-AC-3), not only DVB AC-3. -- Land after [#5363](https://github.com/moq-dev/moq/pull/5363), which reworks - the DVB AC-3 framing this packs. +- [#5363](https://github.com/moq-dev/moq/pull/5363) landed the DVB AC-3 framing + this packs. Video leak accounting must land first too, so the full-rate + proof covers every part of #5362 before this quest closes it. - Test: a generated CBR source with no nulls, at the source's own `--mux-rate`, exports for the whole clip and passes the strict T-STD check. @@ -36,6 +37,6 @@ Decided 2026-10-12: - [#5362](https://github.com/moq-dev/moq/issues/5362) - a full-rate CBR contribution feed exports at its own rate -## Related +## Required - [TS export admits video at its full leak](/quest/m1/tstd/video-leak.md) - #5362 point 2 diff --git a/quest/m1/viewer-upswitch.md b/quest/m1/viewer-upswitch.md index 864de954e7a..d12326950d5 100644 --- a/quest/m1/viewer-upswitch.md +++ b/quest/m1/viewer-upswitch.md @@ -10,12 +10,14 @@ a larger rendition, without pinning by name. `js/watch/src/video/source.ts` caps automatic selection at 0.8 times `estimatedRecvRate`. The estimate only follows what is being sent, so it stays -app-limited, and `js/net` never sends a PROBE target, so the relay never pads -above it. +app-limited, and `js/net` never sends a PROBE target to request headroom above +it. Decided (2026-10-04): its own quest, after the transport can validate -headroom. The viewer sends a PROBE target, the relay pads up to it, and -selection uses the validated estimate. If [quic-probe](/quest/m1/quic-probe.md) +headroom. The viewer sends a PROBE target and selection uses the validated +estimate. Decided in the quest audit: use the mechanism selected by the +capacity measurements, without requiring padding before that comparison. +If [quic-probe](/quest/m1/quic-probe.md) ends in a measured no-go, revisit this quest rather than leaving it blocked. Promoted from m2 on 2026-10-09 by maintainer priority, with the capacity diff --git a/quest/m2/rate-grant.md b/quest/m2/rate-grant.md index b43360fc938..1b2462ef066 100644 --- a/quest/m2/rate-grant.md +++ b/quest/m2/rate-grant.md @@ -15,17 +15,21 @@ expiry; add the two rates in the same nested shape as the and container overhead under the cap. A grant change (refresh, union) re-clamps a live encoder. -The grant's wire shape is specified in `drafts/draft-lcurley-moq-auth.md` -(which the in-band auth line adds); update it in the same PR and run -`just drafts check`. Wire work targets the wip lite version until it is cut -(decided 2026-10-05, see [In-band auth](/quest/m1/auth/README.md)), -and the Auth Stream exists only there, so the rates ride only the wip -version's grant. +Decided in the quest audit: this quest owns net's in-band grants, their lite +codecs, relay propagation, and publisher clamps. The +[bitrate claim](/quest/m2/rate-claim.md) owns token and HTTP grant types in +`moq-auth` and `js/auth`; reuse their admitted caps here. + +Update the AUTH grant in `drafts/draft-lcurley-moq-lite.md` and run +`just drafts check`. Rates ride only the wip lite version's grant. Leave the +negotiated IETF AUTH extension and `drafts/draft-lcurley-moq-auth.md` +unchanged: extending cap reporting to IETF is outside this quest. Keeping +that boundary avoids expanding the agreed lite-only wire change. Tests: a capped session's grant carries the rates; the publisher's video target never exceeds the cap minus overhead; a raised cap lets it climb back. -Public API: the grant types in `rs/moq-auth` and `js/auth` gain the two +Public API: the in-band grant types in `rs/moq-net` and `js/net` gain the two rates, and the publishers gain a clamp. Wire: the AUTH grant gains optional rates, on the wip lite version only. diff --git a/quest/m2/ts-health-stats.md b/quest/m2/ts-health-stats.md index 238754b71e4..ae278ef3435 100644 --- a/quest/m2/ts-health-stats.md +++ b/quest/m2/ts-health-stats.md @@ -12,6 +12,10 @@ counter rates and PID quiet times. Decided while planning [#1838](https://github.com/moq-dev/moq/issues/1838): +The quest audit confirmed this stays in m2 and owns TS counter serialization +and integration. Rust reporters in m1 supplies only the generic per-rendition +plumbing; it does not implement this quest's `mpegts` payloads. + - **Surface through the stats plumbing**, as the issue settled. The [media stats schema](/quest/m1/stats/schema.md) settles the shape: an import flattens `ts::stats::Snapshot` into the publisher's stats snapshot under the diff --git a/quest/m2/uring-demux.md b/quest/m2/uring-demux.md index 2cd45103597..138fff83e3e 100644 --- a/quest/m2/uring-demux.md +++ b/quest/m2/uring-demux.md @@ -39,15 +39,20 @@ Keep the cross-thread hop off the QUIC path: a QUIC datagram never leaves its worker. Bound the channel per class with a drop counter, as the tokio demux does. -Tests: the existing uring relay integration test gains a WebRTC-class -datagram echoed back through the remote-send path, on the QUIC port with a -QUIC session active. QUIC-bit greasing is -already off on the ring, from shard steering. +Decided in the quest audit: functional parity requires the RTC and SRT feed +implementations, not only packet classification. Reuse their integration +tests through the uring remote-send path with a QUIC session active on the +same port, proving real WebRTC and SRT traffic reaches each stack and its +replies return through the owning worker. An echoed packet can test the +channel but does not prove stack integration. QUIC-bit greasing is already +off on the ring, from shard steering. ## Required - [UDP demux](/quest/m2/one-port/udp-demux.md) - the classifier and flow table this runs on the ring - [Steer only QUIC by connection ID](/quest/m2/one-port/shard-steering.md) - keeps each non-QUIC flow on one worker +- [WebRTC on the shared socket](/quest/m2/one-port/rtc-feed.md) - the fed RTC socket and transmit contract the ring integrates +- [SRT on the shared socket](/quest/m2/one-port/srt-demux.md) - the SRT state machines fed through the shared demux ## Related diff --git a/quest/m3/p2p/cli.md b/quest/m3/p2p/cli.md index cad2429b9af..06b2aa235ae 100644 --- a/quest/m3/p2p/cli.md +++ b/quest/m3/p2p/cli.md @@ -18,7 +18,9 @@ than adding startup checks. `--p2p-ice-server`, The roster entry mirrors the `info.json` schema from [signaling](/quest/m3/p2p/signal.md): the listener's LAN addresses as WebTransport URLs, ordered the way `mdns::Peer::urls` orders them, with the -listener's certificate fingerprint; the iroh endpoint id when `--iroh` is on, +listener's certificate fingerprint. Use signaling's ordered-list shape and +dialing order, confirmed in the quest audit, so multi-interface listeners +advertise one shared schema. Include the iroh endpoint id when `--iroh` is on, which is the first place an endpoint id appears in any roster; `webrtc: true`; and `meta`. The generated certificate must satisfy the browser's `serverCertificateHashes` rules (ECDSA P-256, under fourteen days); verify the diff --git a/quest/m3/p2p/signal.md b/quest/m3/p2p/signal.md index 530c7b47351..907d0441f95 100644 --- a/quest/m3/p2p/signal.md +++ b/quest/m3/p2p/signal.md @@ -25,9 +25,11 @@ hop list, so no hop chain carries it. 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/m3/p2p/peer-grant.md) bind, the application's `meta`, -and for native peers an optional `webtransport: { url, fingerprint }` and -`iroh` endpoint id. The schema is shared with -[moq-cli](/quest/m3/p2p/cli.md). +and for native peers an optional WebTransport entry with an ordered list of +URLs and the listener's certificate fingerprint, plus an `iroh` endpoint id. +Decided in the quest audit: use a list so a native peer can advertise its +reachable interfaces without choosing one for every browser. This quest +owns the schema shared with [moq-cli](/quest/m3/p2p/cli.md). Pairing is sparse: broadcasts exist only for pairs some `select` chose, so the cost is the dialed pairs, not the square of the roster. Whoever selects a @@ -46,9 +48,11 @@ The gate: a peer connection is built with `iceServers` when the roster holds candidates are always gathered, so LAN pairs form at any roster size. Established pairs are never closed by the gate. -Native entries: when `info.webtransport` is present the dialer first tries -`connect` with the fingerprint pinned and a short timeout, then falls -through to WebRTC. Chrome 147 may prompt for local network access on that +Native entries: when `info.webtransport` is present the dialer tries its +URLs in advertised order, using `connect` with the listener's fingerprint +pinned and the bounded attempt policy, then falls through to WebRTC if none +connects. Mirror this order in the CLI and test that a later address is tried +after an earlier one fails. Chrome 147 may prompt for local network access on that dial; a denial falls through the same way. Lifecycle: a retracted roster entry closes that peer's session. An ICE