diff --git a/quest/m0/ci/jsr-retire.md b/quest/m0/ci/jsr-retire.md index cad8df5c85f..8194d2c98a3 100644 --- a/quest/m0/ci/jsr-retire.md +++ b/quest/m0/ci/jsr-retire.md @@ -22,13 +22,15 @@ Checked 2026-11-07: no `JSR_TOKEN` exists as a repository secret or in the Archiving needs the maintainer's JSR login. Archiving also ends the repository's OIDC link, so no separate unlinking is needed. Archive only after -the backport lands, or a `release` push can still try to publish. +the backport lands, or a `release` push can still try to publish. It landed +in [moq-dev/moq#8119](https://github.com/moq-dev/moq/pull/8119); checked +2026-11-07, `git grep -n deno origin/release -- js/common +.github/workflows/release-js.yml` prints nothing. Public API: none. Wire: none. ## Waiting -- moq-dev/moq: the `release` backport that drops JSR publishing merges; check `git grep -n deno origin/release -- js/common .github/workflows/release-js.yml` prints nothing - Maintainer: archive each `@moq` package on jsr.io (agents have no JSR login, and JSR offers no archive API for tokenless callers). For each of `auth`, `binary`, `boy`, `flate`, `hang`, `json`, `loc`, `msf`, `net`, `pattern`, `publish`, `room`, `signals`, `token`, `watch`: open `https://jsr.io/@moq//settings`, set the description to `Moved to npm: https://www.npmjs.com/package/@moq/`, then click "Archive package". `boy`, `publish`, and `watch` never published a version, so deleting them from the same page also works (recommended: archive all 15 for one uniform state). Check with `curl -s https://api.jsr.io/scopes/moq/packages | jq '[.items[] | select(.isArchived | not) | .name]'` printing `[]` ## Related diff --git a/quest/m0/release-audio-group-cap.md b/quest/m0/release-audio-group-cap.md index fc7a001f42e..c8735ae67ba 100644 --- a/quest/m0/release-audio-group-cap.md +++ b/quest/m0/release-audio-group-cap.md @@ -16,10 +16,17 @@ and rolls at the net caps, tested with 2.5 ms frames and 1 MiB frames under a ([unbounded groups](/quest/m2/unbounded-groups/README.md)), so it gets no change. +Decided 2026-11-07 (maintainer, /quest-unblock): ✅ cut it now as 0.5.6, a +patch (rejected: waiting for a bigger cut). +[moq-dev/moq#8118](https://github.com/moq-dev/moq/pull/8118) merged into +`release` as c4c4078293, and +[moq-dev/moq#8213](https://github.com/moq-dev/moq/pull/8213) bumps +`@moq/publish` to 0.5.6 there; merging it publishes through `release-js.yml`. + Once the patch publishes, delete this quest. Public API: none. Wire: none. ## Waiting -- Maintainer: merge [moq-dev/moq#8118](https://github.com/moq-dev/moq/pull/8118) into `release` and cut the next `@moq/publish` patch; check that `npm view @moq/publish version` is past 0.5.5 and that its release cut contains [moq-dev/moq#8118](https://github.com/moq-dev/moq/pull/8118)'s merge commit. A release cut is a maintainer action, and version bumps are manual. +- Maintainer: merge [moq-dev/moq#8213](https://github.com/moq-dev/moq/pull/8213), the 0.5.6 bump on `release`; merging is the maintainer's pick. Check that `npm view @moq/publish version` prints 0.5.6. diff --git a/quest/m1/fuzz-findings/README.md b/quest/m1/fuzz-findings/README.md index 3dfe6e06956..ee46951543d 100644 --- a/quest/m1/fuzz-findings/README.md +++ b/quest/m1/fuzz-findings/README.md @@ -50,13 +50,8 @@ the bug on its own side and the bug is reported upstream (decided 2026-10-30, maintainer): [austinleroy/webm-iterable#11](https://github.com/austinleroy/webm-iterable/issues/11), [therishidesai/h264-parser#9](https://github.com/therishidesai/h264-parser/issues/9), [ScuffleCloud/scuffle#653](https://github.com/ScuffleCloud/scuffle/issues/653) -(scuffle-h265), and [ScuffleCloud/scuffle#654](https://github.com/ScuffleCloud/scuffle/issues/654) -(scuffle-expgolomb). Posting anything further on those repos still needs the +(scuffle-h265), [ScuffleCloud/scuffle#654](https://github.com/ScuffleCloud/scuffle/issues/654) +(scuffle-expgolomb), and [ScuffleCloud/scuffle#655](https://github.com/ScuffleCloud/scuffle/issues/655) +(scuffle-av1 `read_uvlc`, approved 2026-11-07 by the maintainer in /quest-unblock). Posting anything further on those repos still needs the maintainer's approval. -## Waiting - -- Maintainer: approve reporting the scuffle-av1 `read_uvlc` overflow fixed by - [moq-dev/moq#8042](https://github.com/moq-dev/moq/pull/8042) to ScuffleCloud/scuffle, - which has no issue for it; an agent cannot post there without approval. Recommend - approving, matching the scuffle-h265 and scuffle-expgolomb reports above. Done once the issue is filed and linked above. diff --git a/quest/m1/vaapi-radeonsi.md b/quest/m1/vaapi-radeonsi.md index b5cae63f4a8..7fac16a0bb0 100644 --- a/quest/m1/vaapi-radeonsi.md +++ b/quest/m1/vaapi-radeonsi.md @@ -22,20 +22,16 @@ Mesa 26.2.5); on moq-vaapi 0.2.0, 7 fail. That PR's body has the evidence. - Chroma: not an offset. Mesa's VPP reads the BT.601 standard as SMPTE 170M primaries and gamut-maps; moq-vaapi now names each side explicitly on Mesa. -What remains here: bump `moq-vaapi` in the root `Cargo.toml` to the release -with #13, refresh `Cargo.lock` with it, and say in the module docs of +Decided 2026-11-07 (maintainer, /quest-unblock): an agent merged #13 and +published it as moq-vaapi 0.2.1 (a patch: `cargo semver-checks` against 0.2.0 +needs no bump), tagged `v0.2.1`. + +What remains here: bump `moq-vaapi` in the root `Cargo.toml` to 0.2.1, refresh `Cargo.lock` with it, and say in the module docs of `rs/moq-video/src/encode/backend/vaapi.rs` that the tests pass on radeonsi. Then delete this quest. Public API: none. Wire: none. -## Waiting - -- Maintainer: merge [moq-dev/vaapi#13](https://github.com/moq-dev/vaapi/pull/13) - and publish it as moq-vaapi 0.2.1 (a patch: no public API change). Version - bumps are a person's call, so an agent does not cut the release. Check that - `cargo info moq-vaapi@0.2.1` finds it. - ## Related - [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - AMD's long-term encode path diff --git a/quest/m2/cluster-routing/routes/browser-proof.md b/quest/m2/cluster-routing/routes/browser-proof.md index 81dd79d753b..37f7f541f55 100644 --- a/quest/m2/cluster-routing/routes/browser-proof.md +++ b/quest/m2/cluster-routing/routes/browser-proof.md @@ -2,9 +2,11 @@ ## Goal -Before the lite-07 cut, a lite-07 session from a current Firefox tab connects -to `moq-relay` with its route-key proof verified, as a current Chromium tab's -does, and the nightly route key lane keeps both that way. +The lite-07 cut does not wait on Firefox: the release gate passes a Firefox +tab that connects unproven, as it does Safari. Once a Firefox release follows +the draft's exporter, a lite-07 session from it connects to `moq-relay` with +its route-key proof verified, as a current Chromium tab's does, and the +nightly route key lane keeps both that way. The lane (`test/interop/clients/js/route-key.ts`, in `test/interop` nightly and the release gate) dials the relay from each browser on lite-07 and reads @@ -32,8 +34,25 @@ connects unproven rather than being refused. Decided 2026-11-06 (maintainer, /quest-complete while queueing [moq-dev/moq#8022](https://github.com/moq-dev/moq/pull/8022)): cover Firefox -and Chromium, treat a mismatch as a bug that blocks the cut, and never send a -proof the relay will reject. +and Chromium, and never send a proof the relay will reject. + +Decided 2026-11-07 (maintainer, /quest-unblock): ✅ cut lite-07 with Firefox +unproven (rejected: the cut waits for a Firefox that proves; the relay also +accepting neqo's varint derivation, refused because it departs from the draft +and tries a second path at runtime; the relay picking the derivation from the +CONNECT's User-Agent, which is server-side UA sniffing). This replaces +2026-11-06's "a mismatch blocks the cut". An unproven tab is a supported +state, the neqo fix is one line, and the lane still fails loud if the relay +refuses a Firefox tab. + +Now: make the lane pass a Firefox tab that sends no proof under `CI_STRICT`, +as it treats a build that is not yet covered elsewhere, rather than failing +the release gate (`uncovered` in `test/interop/clients/js/route-key.ts`). +A Firefox tab that the relay refuses, or that sends a proof, still fails. + +Reported upstream on 2026-11-07 as +[mozilla/neqo#4319](https://github.com/mozilla/neqo/issues/4319), with the +one-line fix (approved by the maintainer in /quest-unblock). Once a Firefox release carries neqo's fix, gate `route_key.ts` on Firefox majors before it, set the lane's table (`uncovered` in `route-key.ts`) to @@ -44,20 +63,8 @@ does for Chromium. Verify the fixed build against the relay the way the ## Waiting -- Maintainer: approve reporting the derivation to mozilla/neqo (outside the - moq-dev and kixelated orgs, so an agent may not post unasked), as an issue - with a one-line patch writing the session ID as 64 bits; the draft is in - [moq-dev/moq#8116](https://github.com/moq-dev/moq/pull/8116). Done when the - issue or PR exists. -- Maintainer: decide whether the lite-07 cut waits for a Firefox that proves. - (1) Wait, as decided: the release gate fails on Firefox until a fixed - Firefox ships and Playwright bundles it, on a timeline outside our control. - (2) Cut with Firefox unproven, as Safari tabs already are: the lane passes a - Firefox that sends no proof under `CI_STRICT`, and flips to proven later. - (3) Have the relay also accept neqo's derivation: refused, since it departs - from the draft and tries a second path at runtime. Recommend (2): the fix - is upstream, an unproven tab is a supported state, and the lane still - fails loud if a Firefox tab is refused. Done when this Plan records the pick. -- mozilla/neqo: the exporter context writes the session ID as 64 bits; check - `webtransport_export_keying_material` in `neqo-http3/src/webtransport.rs`, - then the first Firefox release vendoring it. +- mozilla/neqo: the exporter context writes the session ID as 64 bits, as + asked in [mozilla/neqo#4319](https://github.com/mozilla/neqo/issues/4319); + check `webtransport_export_keying_material` in + `neqo-http3/src/webtransport.rs`, then the first Firefox release vendoring + it. diff --git a/quest/m2/cluster-routing/routes/playwright-pin.md b/quest/m2/cluster-routing/routes/playwright-pin.md index 9ff6c62e336..ebbbc0a5e32 100644 --- a/quest/m2/cluster-routing/routes/playwright-pin.md +++ b/quest/m2/cluster-routing/routes/playwright-pin.md @@ -36,4 +36,10 @@ rebuild and pin browser binaries outside nixpkgs. ## Waiting - nixpkgs: `playwright-driver` at 1.65.0 or later on `nixpkgs-unstable`; - check `pkgs/development/web/playwright/driver.nix` there. + check `pkgs/development/web/playwright/driver.nix` there. nixpkgs' + `update.sh` takes the driver version from playwright-python's latest + release, and `python3Packages.playwright` is coupled to the driver, so the + bump waits on a playwright-python v1.65 release (none as of 2026-11-07). + Once one exists, an agent opens the nixpkgs bump PR with that script from + the kixelated/nixpkgs fork (approved by the maintainer 2026-11-07 in + /quest-unblock). diff --git a/quest/m2/cluster-routing/routes/session.md b/quest/m2/cluster-routing/routes/session.md index 905d2b61296..479b6d8cd87 100644 --- a/quest/m2/cluster-routing/routes/session.md +++ b/quest/m2/cluster-routing/routes/session.md @@ -27,10 +27,11 @@ node table, and the public API drops what the route layer replaced. change and per interest (`Router::advertise`), so a large origin costs linear work per update. Do it with the scoping change, which reshapes the same per-stream presentation. -- Public API: `Hops`, `RouteInfo::hops`, and the saturating `Cost` stay while - lite-06 and the moq-transport cluster extension still carry hop chains. - Remove them, or hide them behind those wires, when the rollout window - closes (see Waiting). +- Public API: remove `Hops`, `RouteInfo::hops`, and the saturating `Cost` + from the public API now, keeping a private chain type for lite-06 and the + moq-transport cluster extension, which still carry hop chains. Decided + 2026-11-07 (maintainer, /quest-unblock): ✅ now (rejected: when lite-06 + retires). It is a breaking change; the PR marks its title with `!`. Added 2026-11-07 (maintainer, /quest-plan from the API shape audit), for the PR that lands this ([moq-dev/moq#8023](https://github.com/moq-dev/moq/pull/8023)): @@ -42,15 +43,6 @@ the PR that lands this ([moq-dev/moq#8023](https://github.com/moq-dev/moq/pull/8 the [epic](/quest/m2/cluster-routing/README.md)), so sessions stop stamping `Cost::DRAIN` on GOAWAY, including on moq-transport. -## Waiting - -- Maintainer: decide when `Hops`, `RouteInfo::hops`, and the saturating - `Cost` leave the public API. Every older wire still fills them, so an agent - cannot drop them without picking the rollout's end. Options: **when lite-06 - retires (recommended)**, since removing them now only moves them behind a - private type; or **now**, keeping a private chain type for the older wires. - Done when the Plan item above names the point. - ## Related - [Routes and announces](/quest/m2/cluster-routing/routes/README.md) - the contract diff --git a/quest/m2/subscribe-ranges/lite07-fetch.md b/quest/m2/subscribe-ranges/lite07-fetch.md index dda0805168b..b7c5ae30568 100644 --- a/quest/m2/subscribe-ranges/lite07-fetch.md +++ b/quest/m2/subscribe-ranges/lite07-fetch.md @@ -18,6 +18,18 @@ and does not support slicing within a group." Route recovery (`resume::Recover`), which fills from frame N, keeps its FETCH (`group::Fetch::frame_start`). +Decided 2026-11-07 (maintainer, /quest-unblock): ✅ lite-07 keeps +single-group FETCH (rejected: FETCH of several ranges delivered in order; +dropping FETCH from lite-07). Head-of-line blocking between groups is not a +feature: under backpressure the head is evicted and cancels every group +queued behind it, and out-of-order arrival helps even a 30 s jitter buffer, +while a relay should serve newer cached groups rather than wait on old ones +from cold storage. Dropping FETCH would move frame and byte offsets into every +SUBSCRIBE range operation instead. Since lite-07 names a group by its start +time, a reader that knows only a time reads it through a SUBSCRIBE range +bounded by times; FETCH stays for readers that know the group's start, such +as route recovery, flights, and byte holes. + Decided 2026-10-25 (maintainer, on [moq-dev/moq#6807](https://github.com/moq-dev/moq/pull/6807)) and still standing: Rust and JS change together, since both request and serve wire FETCH diff --git a/quest/m2/unbounded-groups/README.md b/quest/m2/unbounded-groups/README.md index 25122dceb96..aaf669e83fb 100644 --- a/quest/m2/unbounded-groups/README.md +++ b/quest/m2/unbounded-groups/README.md @@ -183,4 +183,4 @@ the model layer (maintainer, 2026-10-30). - [Byte holes](/quest/m2/unbounded-groups/byte-holes.md) - segments inside a frame shed like front frames and flights start at `(frame, byte)` on the current lite wip; the frame cap goes - [JS byte holes](/quest/m2/unbounded-groups/js-byte-holes.md) - `@moq/net` mirrors the byte-offset wire and continues a frame cut off mid-way - [Compressed frames have no size cap and inflate as a bounded stream](/quest/m2/unbounded-groups/compressed-stream.md) - A compressed frame of any size publishes and is received in Rust and `@moq/net`, like a plain one -- [Lite-07 history runs are paced](/quest/m2/unbounded-groups/run-fill-pacing.md) - range fills, recoveries, and a reader's own history range backpressure upstream once a reader loses a group to eviction +- [Remove FETCH-run pacing](/quest/m2/unbounded-groups/fetch-run-unpaced.md) - a moq-transport FETCH run is no longer held at group boundaries for a slow reader, matching lite-07 diff --git a/quest/m2/unbounded-groups/fetch-run-unpaced.md b/quest/m2/unbounded-groups/fetch-run-unpaced.md new file mode 100644 index 00000000000..76effd60446 --- /dev/null +++ b/quest/m2/unbounded-groups/fetch-run-unpaced.md @@ -0,0 +1,36 @@ +# [S] Remove FETCH-run pacing + +## Goal + +A moq-transport FETCH run reads its stream at the speed it arrives, as a +lite-07 history run does: no wire request is held at group boundaries for a +slow reader. A reader that falls off the cache tail gets `Evicted`, then +refill-once or a refresh, on both wires. + +## Plan + +Decided 2026-11-07 (maintainer, /quest-unblock): ✅ no flow control between +groups on either wire (rejected: (a) holding lite-07 group streams unread and +admitting them in the fill's order; (b) narrowing the subscription's ranges +to the next owed group; (c) fetching paced runs a group at a time). Pacing +between groups is head-of-line blocking: under backpressure the head is +evicted and cancels everything queued behind it. SUBSCRIBE is best-effort; per-subscription flow control was considered and not judged worth it. +This deletes the planned lite-07 pacing quest and undoes what +[moq-dev/moq#8117](https://github.com/moq-dev/moq/pull/8117) added for +moq-transport FETCH. + +- Remove `track::Fill::poll_run_paced` (`rs/moq-net/src/model/track/fill.rs`) + and the FETCH stream's `poll_pace` hold that calls it + (`rs/moq-net/src/ietf/subscriber.rs`), with #8117's tests and any reader or + test-transport hooks only they use. +- Keep `TrackState::evictions`: `track/miss.rs` reads it too. +- Out of scope: [moq-dev/moq#8017](https://github.com/moq-dev/moq/pull/8017)'s + pacing of a batch fill's local handler by its fastest reader, which holds + no wire stream. +- Update the epic README's prose that describes FETCH-run pacing, if any. + +Public API: none. Wire: none. + +## Related + +- [Unbounded groups](/quest/m2/unbounded-groups/README.md) - the epic whose eviction and refill rules a slow reader now falls back on diff --git a/quest/m2/unbounded-groups/run-fill-pacing.md b/quest/m2/unbounded-groups/run-fill-pacing.md deleted file mode 100644 index a508a069f4e..00000000000 --- a/quest/m2/unbounded-groups/run-fill-pacing.md +++ /dev/null @@ -1,67 +0,0 @@ -# [M] Lite-07 history runs are paced - -## Goal - -A lite-07 copy bringing a long run of history to a slow reader (a range fill, a -recovery, or the reader's own history range on the copy's subscription) -backpressures the upstream once it sees a downstream reader lose a group to -eviction, so a slow reader of a long run stays within the tail instead of -falling off it. - -## Plan - -Added 2026-11-06 (maintainer, follow-up planning in `/quest-complete`), from -[moq-dev/moq#8017](https://github.com/moq-dev/moq/pull/8017), which paces a -batch fill's handler by its fastest reader. Pace on evidence, not up front: an -unpaced run is fine while every reader keeps up. - -Landed in [moq-dev/moq#8117](https://github.com/moq-dev/moq/pull/8117): the -evidence is the track's eviction count (`TrackState::evictions`, bumped when a -reader loses a group it judged cached), and a moq-transport FETCH run holds its -fetch stream unread at each group boundary until a reader picked up the group -before (`Fill::poll_run_paced`), through the session's pause budget. - -What remains is lite-07, which that PR's prototypes found harder than the plan -assumed: - -- History reaches a lite-07 copy three ways, all unpaced: the copy's upstream - subscription carrying a reader's own history range (the common case, and not a - fill), `RangeFills` asking a held fill's run through that subscription, and a - `Recovery` subscription asking again for what the subscription already - accounted for. Each arrives as one group stream per group, concurrently. -- Holding each arriving group stream unread until the fill's readers picked up - the one before admits streams in arrival order, not the fill's, so groups a - reader takes last evict each other first. A slow reader of 40 groups on a - 4-group cache still got 10. -- Fetching a paced fill a group at a time instead (lite's `FillServeRun`, paced - since #8017) got that reader 26 of 40, but costs a round trip per group, which - #8017 and [moq-dev/moq#7339](https://github.com/moq-dev/moq/pull/7339) - rejected, and paces neither the subscription's own history range nor several - fills at once. -- #8017's per-fill pacing does not hold when one reader has several fills out, - split by groups already cached or by refills: each fill moves on as the reader - picks up its answer, so together they still outrun it. This affects batch - handlers and FETCH runs too. - -The mock transport moves a whole run in an instant, before any reader polls, so -a test needs the evidence first: one slow reader loses part of a run, then a -second slow reader of another long run gets all of it. - -Public API: none expected. Wire: none. - -## Waiting - -- Maintainer: pick lite-07's pacing mechanism. (a) Hold arriving group streams - unread and admit them in the fill's order, within the session's pause budget, - extended to the history range a reader's subscription carries: no round trips, - but out-of-order streams spend the pause budget and read on unpaced once it - runs out. (b) Narrow the subscription's ranges to the next owed group once - paced: in-order, but a SUBSCRIBE_UPDATE and a round trip per group. (c) Fetch - paced runs a group at a time: reuses `FillServeRun`, but a round trip per group - and the subscription's own range stays unpaced. (d) Drop lite-07 pacing and - rely on refill-once. Recommended: (a), since it is the plan's "hold the - upstream stream" without round trips; pair it with pacing per reader across - its fills (a read picks up one answer per group it takes, all of them while it - waits), so several fills of one reader cannot outrun it. An agent cannot take - this: (b) and (c) reopen #7339's rejection, and the per-reader change rewrites - #8017's settled pick-up rule.