You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
refactor(net): drop JS moq-transport FETCH serving - #8445
js/net no longer serves moq-transport FETCH. Every FETCH, standalone, filtered, or joining, gets NOT_SUPPORTED on its own stream; the session stays up. Deleted runFetch and its range, join registry, and FETCH_OK writer, plus the groupEnd hook only they used. The draft-20+ fill and #writeFetchGroup stay.
A JS draft-18 Next Object subscription carries its current group whole, from object 0. Without the joining FETCH answer, a draft-18 late joiner would lose the head of the current group, such as a catalog written once. Our subscribers already take a group stream that starts at object 0 as the whole group, and keep going after a refused joining FETCH, so this needs no subscriber change. A third-party subscriber that sends no joining FETCH also gets the head now. This brings back what the publisher did before moq-dev/moq#6945, but SUBSCRIBE_OK still reports the true Largest Object.
moq-net keeps listing a moq-transport peer's tracks with SUBSCRIBE_TRACKS.rs/ is unchanged from main. Cloudflare's relay (cloudflare/moq-rs, moq-transport/src/message/subscribe_tracks.rs and moq-relay-ietf/src/producer.rs) answers SUBSCRIBE_TRACKS with PUBLISH. Without our side we could no longer list tracks on that relay.
JS still refuses a received SUBSCRIBE_TRACKS with NOT_SUPPORTED. It now reads only the Request ID and skips the rest of the body, so JS's subscribe_tracks.ts is gone and the type is a bare SUBSCRIBE_TRACKS_ID constant. Rust still decodes the whole message, because it sends one when listing.
Docs (doc/concept/standard.md, doc/lib/js/net.md) and quest notes (m3/audit-interop.md, which lists FETCH cases with a moq-dev-js publisher as unsupported, and the m2/ietf-scope epic README, which now records that listing stays) describe the new scope.
Decisions
Track listing: ✅ keep moq-net's SUBSCRIBE_TRACKS listing as on main (rejected: delete it). Decided 2026-11-10 by the maintainer after the PR left the merge queue. The earlier premise that no third party consumes listings was wrong: Cloudflare's relay implements SUBSCRIBE_TRACKS.
What listing needs: ✅ restore everything this PR had deleted along with listing (rejected: restore listing alone). Against a real SUBSCRIBE_TRACKS publisher, each listed track arrives as a FORWARD=0 PUBLISH. So listing needs PUBLISH adoption in SubscribeOnce, the paused/resuming alias states that a reader resumes instead of subscribing again, and DUPLICATE_SUBSCRIPTION for a redundant draft-18 PUBLISH. Restoring rs/ whole from main is also the simplest result. It drops two of the earlier choices: the Rust "refuse SUBSCRIBE_TRACKS without decoding" and the IETF fuzz-kind renumbering, so the fuzz regression inputs are back as on main.
Epic remainder: ✅ title quest/m2/ietf-scope/README.md[S] and rewrite it as a standalone quest for the remaining docs work (doc/concept/, drafts/draft-lcurley-moq-ietf.md if it lists the extensions, the interop matrix) (rejected: finishing the docs in this PR, XS). Decided 2026-11-11 by the maintainer after the merge queue removed the PR: deleting the last child made the README a quest without a size, failing quest check.
✅ Keep the JS changes as they were (FETCH refused, draft-18 group served whole, SUBSCRIBE_TRACKS refused without decoding), per the same decision.
Choices made without asking
Serve the draft-18 group whole instead of keeping the joining FETCH answer. The alternatives were to keep JS joining-FETCH serving (against the quest's plan, which names feat(js): serve cached IETF FETCH, including draft-18 joining FETCH #6945's joining FETCH), or to delete it and accept the draft-18 head loss, which broke about ten JS integration tests and would break a catalog written once. Serving extra objects departs from Next Object's start, but costs less than the joining FETCH did, and per AGENTS.md the IETF requirements give way when they add cost without benefit.
Skip, not keep, the moq-transport FETCH cells with a JS publisher. A relay fills a downstream miss with an upstream FETCH, so with a native JS publisher on moq-transport the interop CLI FETCH and CLI FETCH history lanes can no longer pass. They are skipped there with a named reason, and test/interop/README.md says so. Over moq-lite, which our JS publishers use, both lanes still run and pass. Also deleted quest/m3/ietf-js-joining-fetch.md, whose goal this PR makes moot.
Impact
Public API: none. Nothing in js/net's public API exposed FETCH serving; moq-net is unchanged.
Wire: no format change. A FETCH to a JS publisher (a relay's upstream fill included) gets NOT_SUPPORTED, an optional behavior in moq-transport. A JS draft-18 Next Object subscription now carries objects from 0 in its first group. Rust's SUBSCRIBE_TRACKS sending and both sides' refusal are unchanged.
On the head that restores listing (just check rerun after merging main again on 2026-11-11, which also took main's new FETCH-serving tests out of fetch.test.ts):
just check: passed.
just test interop --all: everything passed except one moq-lite-06 rust -> js-native-bun cell, which received only audio before the connection closed under load. It passed on rerun. This PR does not touch moq-lite or rs/.
just test interop --all --versions ietf: everything passed except one js -> gst cell (a 20 ms audio timestamp gap), with the 12 JS-publisher CLI FETCH cells per draft skipped as above. Six reruns of js -> gst across the IETF drafts all passed.
Follow-ups
None. The earlier js-web rerun note is gone: main's rewrite of the interop audit already has it confirm every red cell after the runner client bump.
Models
Size: M
Implementer: Claude Code, Claude Opus 5.5, medium effort
Implementer: Claude Code, claude-opus-5-5, medium effort (iterate, M)
Implementer: Claude Code, claude-opus-5-5, low effort (iterate, S)
Score: 8/10, The reduced FETCH-serving scope is consistently reflected in dispatch, documentation, and interop selection, with no actionable defects found; builds and tests were not run, as instructed.
…ting
js/net answers every moq-transport FETCH with NOT_SUPPORTED, so a draft-18
Next Object subscription carries its current group whole instead of a joining
FETCH head. moq-net no longer lists a moq-transport peer's tracks with
SUBSCRIBE_TRACKS: such a listing fails as unsupported, and a received
SUBSCRIBE_TRACKS is still refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
changed the title
moq-transport FETCH serving in JS and track listing are gone
refactor(net): drop JS moq-transport FETCH serving and Rust track listing
Nov 10, 2026
The reason will be displayed to describe this comment to others. Learn more.
Automated review by review (OpenAI)
Reviewed the full patch at fa33c64 and the affected request, subscription, alias, and group-join paths. No actionable findings.
Direction: the removals match the agreed IETF scope. Unsupported requests retain request-ID admission and per-request refusals; ordinary Rust subscription alias ownership and cancellation remain intact. The draft-18 JavaScript fallback sends the current group's head on the subscription while retaining the real Largest Object, consistent with the existing whole-group receiver path. All 28 changed fuzz regression inputs preserve their payloads and select the same message types after renumbering.
Verification limits: GitHub-only static review. No builds, tests, fuzzing, or interop runs performed; author-reported checks were not independently verified. Third-party behavior with the draft-18 whole-group fallback remains for the documented interop follow-up.
File: No configuration file from this repository applied
Commit: Unavailable
Layers: Organization UI
Review profile: CHILL
Plan: Advanced
Run ID: 42a870ef-3285-4688-a314-d1ad7882b69d
Commits
Reviewing files that changed from the base of the PR and between fa33c64 and f9cbd2e.
Files selected for processing: 22
doc/concept/standard.md
doc/lib/js/net.md
js/justfile
js/net/bench/fetch.ts
js/net/src/group.test.ts
js/net/src/group.ts
js/net/src/ietf/connection.test.ts
js/net/src/ietf/connection.ts
js/net/src/ietf/fetch.test.ts
js/net/src/ietf/malformed.test.ts
js/net/src/ietf/publisher.test.ts
js/net/src/ietf/publisher.ts
js/net/src/ietf/subscribe_tracks.ts
js/net/src/integration.test.ts
js/net/src/internal.ts
quest/m2/ietf-scope/README.md
quest/m2/ietf-scope/serving.md
quest/m3/audit-interop.md
quest/m3/ietf-js-joining-fetch.md
sh/gh/nightly.sh
test/interop/README.md
test/interop/interop.sh
No actionable comments were generated in the recent review.
Walkthrough
The changes remove Rust upstream track listings and JavaScript FETCH serving. Both implementations refuse incoming SUBSCRIBE_TRACKS requests with NOT_SUPPORTED. JavaScript also refuses incoming FETCH requests. Rust subscriber setup no longer adopts held listing subscriptions, and incoming PUBLISH offers are declined. The changes remove related protocol types, helpers, tests, and benchmarks, and update documentation to describe the revised behavior.
Merge Risk: Minimal · previous review fa33c
This change removes JavaScript FETCH serving and Rust track listing, and refuses these requests with NOT_SUPPORTED while the session stays up. No concrete merge-blocking defect remains. The author still needs to rerun the js-web publisher interop cells behind a joining FETCH, as the PR notes.
Pre-merge checks: 4
Check name
Status
Explanation
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Title check
The title clearly identifies the main JavaScript change: removing moq-transport FETCH serving. It is concise and directly related to the changeset, although it does not mention the Rust track-listing …
Description check
The pull request description directly addresses the changeset, including removal of JavaScript FETCH serving, SUBSCRIBE_TRACKS behavior, related code removal, documentation updates, and test results.
Full details: Title check
Explanation
The title clearly identifies the main JavaScript change: removing moq-transport FETCH serving. It is concise and directly related to the changeset, although it does not mention the Rust track-listing changes.
Finishing Touches
Simplify code
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 1 remain after this review.
The reason will be displayed to describe this comment to others. Learn more.
Important
Removing JS FETCH serving also removes the only path that fills a relay's upstream miss, which the IETF interop CLI FETCH history lane still exercises against a native JS publisher. One now-impossible m3 quest and a few removed-FETCH references are also left behind.
Reviewed changes
JS publisher stops serving FETCH.runFetch, the draft-18 join registry, the range and joining answer paths, and the groupEnd hook are gone; Connection refuses every FETCH NOT_SUPPORTED on its own stream while the session stays up. The draft-20+ fill path stays.
Draft-18 Next Object subscriptions carry the current group whole.filterRange now takes the version and starts a pre-draft-20 nextObject at {largest.group, 0}, so the head arrives without a joining FETCH while SUBSCRIBE_OK still reports the true Largest Object.
Rust stops listing a moq-transport peer's tracks.ietf/subscribe_tracks.rs and ietf/subscriber/tracks.rs are deleted with the paused/resuming alias states, PUBLISH adoption, and DUPLICATE_SUBSCRIPTION; a received PUBLISH is always declined and a local listing is aborted with Error::Unsupported without sending SUBSCRIBE_TRACKS.
SUBSCRIBE_TRACKS is a bare refusal. Rust and JS read only its Request ID and skip the body; the type is a SUBSCRIBE_TRACKS_ID constant in both.
Fuzz kinds renumbered from 27 to 25 with the committed regression inputs rewritten to the same effective decoder, plus the doc and quest notes.
⚠️ The IETF interop CLI FETCH history lane can no longer pass with a JS publisher
test/interop/README.md documents that a native JS publisher answers groups 0 through 2 of the CLI FETCH history lane from track.requests(), and that this "runs on moq-lite and every supported IETF draft". The relay fills a downstream miss by sending an upstream ietf::Fetch, so with every FETCH refused there is no path for those groups, and run_fetch_round <js-runtime> <sub> history keeps running unchanged.
The follow-up in the description only covers the joining-FETCH js-web cells, so this lane was not reconciled. Note that run_tails gates its native-JS publisher loops on needs (which also consults the subscriber list), so those cells also run on the PR lane where js-native-* are subscribers only.
Technical details
# The IETF `CLI FETCH history` lane still assumes a JS publisher answers FETCH## Affected sites-`test/interop/interop.sh:2684` — `run_fetch_round "$runtime" "$sub" history` still runs for `js-native-node` and `js-native-bun`.
-`test/interop/README.md:76-79` — states the JS publisher "answers groups 0 through 2, which it never cached" and that the lane runs on every supported IETF draft.
-`rs/moq-net/src/ietf/subscriber.rs:6998` — `ServeFill` fills a local miss by writing `ietf::Fetch::ID` upstream, a Standalone FETCH on draft-18 and a Filtered absolute FETCH from draft-20.
-`js/net/src/ietf/connection.ts:401-406` — every FETCH is now refused `NOT_SUPPORTED`.
-`quest/m3/audit-interop.md:101-104` — the audit note covers only the third-party `js-web` publisher cells behind a joining FETCH.
## Required outcome- The IETF pass of this lane must not fail for an intended removal: either skip the `history` cells whose publisher is `js-native-*` with a named reason (the `run_finishes` quarantine comment is the local pattern), or record them as expected failures somewhere the harness reads.
-`test/interop/README.md` must stop claiming the lane covers moq-transport with a JS publisher.
## Open questions for the human- Is the intent to drop the whole lane for moq-transport, or should the JS publisher keep serving cached/fill FETCH for upstream fills while only joining FETCH goes away?
ℹ️ Removed-FETCH references left behind
Three places still describe the deleted behavior. The first is an open quest whose goal this PR makes impossible; the other two are docs and comments that now contradict doc/concept/standard.md.
Technical details
# Stale references to JS FETCH serving## Affected sites-`quest/m3/ietf-js-joining-fetch.md` — goal is "the JS publisher answers a joining FETCH after a quick finish", which cannot happen any more. Still linked as open from `quest/m3/README.md:121`. The PR deletes `quest/m2/ietf-scope/serving.md`, so retiring this one in the same pass keeps the tree coherent.
-`doc/lib/js/net.md:53` — "This serves lite FETCH, single-group IETF FETCH, and lite-07 ranged subscriptions from the same track". js/net no longer serves IETF FETCH; only lite FETCH and lite-07 ranges still reach `track.requests()`.
-`js/net/src/ietf/publisher.ts:833-835` — the `SUBSCRIBE_OK.largest` comment says a Next Object subscription starts right after it and recovers the head with a fill or a joining FETCH. On draft-18 it now starts at object 0 of that group.
-`js/net/src/ietf/publisher.test.ts:2851-2854` — the same claim in the `LARGEST_OBJECT is the live edge` doc comment (the assertion itself is still correct).
## Required outcome- The quest is deleted (or rewritten to a closed note), and the three references describe the draft-18 whole-group behavior or drop the IETF FETCH claim.
The reason will be displayed to describe this comment to others. Learn more.
Important
The merge integration itself is clean, but the IETF interop CLI FETCH history concern from the previous review is unchanged, so this still is not mergeable as-is. run_fetch_round <js-runtime> <sub> history needs groups 0-2 that the relay can only obtain with an upstream ietf::Fetch (FetchRun at rs/moq-net/src/ietf/subscriber.rs:7148), while the JS publisher now serves track.requests() only through the SUBSCRIBE-fill path; PR CI's --publishers rust hides that, nightly's --all --versions ietf does not.
Reviewed changes
Listing removal re-applied to main's refactored state machine.OnceStep::Adopt, Write's held, Answering::held, OnceRequest::token_grant, Established::alias, the Paused/Resuming alias states, and pause/want/unwant/adopted/entry_info/entry_ended/is_entry/poll_listing/listen/poll_handoff/buffer_forward are all gone from ietf/subscriber.rs; nothing else from the merge was dropped.
Shared helpers inlined locally.tracks::TrackKey became type TrackKey = (Path, Arc<str>) and tracks::listing_end became a byte-identical done_end.
RunBroadcast refuses listings instead of serving them. It holds a broadcast::Lister and aborts each wanted listing with Error::Unsupported; the loop terminates because Lister::abort clears the readers, returning poll_wanted to Pending.
Secondary Rust files carry only the matching removals.RequestKind::Publish, DUPLICATE_SUBSCRIPTION, Publish::into_owned, request_stream::Cancel, and the SUBSCRIBE_TRACKS_ID wiring in session.rs/publisher.rs; datagrams.rs follows the new TrackKey.
Verified at 30d7aed1b:cargo check -p moq-net --all-features warning-free, cargo test -p moq-net --all-features --lib ietf:: → 838 passed, --test integration tracks_ietf → 1 passed. js/net/src/ietf/ is untouched since the prior review.
… stale FETCH claims
A relay fills a downstream miss with an upstream FETCH, which a JS publisher
now refuses on moq-transport, so the interop CLI FETCH and history cells with a
native JS publisher are skipped there with a named reason. The JS net docs, two
comments, and the m3 JS joining FETCH quest no longer describe IETF FETCH
serving.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The reason will be displayed to describe this comment to others. Learn more.
Automated review by review (OpenAI)
Reviewed the updates through 453b165 against the prior review and current base, including the Rust subscriber merge integration. No new actionable findings.
Direction: the interop follow-up addresses the previously reported unsupported FETCH cells. The shared run_fetch_round guard records explicit skips for native-JS publishers on moq-transport in both cached and history modes, while retaining moq-lite and Rust-publisher coverage. Removing the obsolete joining-FETCH quest and correcting the related documentation is consistent with the scope reduction.
Verification limits: GitHub-only static review; no builds, tests, or interop runs performed. Author-reported passing runs were not independently verified. The documented third-party joining-FETCH interop follow-up remains.
The reason will be displayed to describe this comment to others. Learn more.
✅ No new issues found.
Reviewed changes — the delta since the prior pullfrog review (30d7aed1b..453b165cf, 8 files, +17/−39):
Interop FETCH cells now skip a native JS publisher on moq-transport.run_fetch_round records a named skip when $VERSION is moq-transport-* and $pub is js-native-*, before opening a relay. Both the cached and history calls share the function, and run_tails only passes rust/js-native-node/js-native-bun, so no other publisher is affected.
The interop README states the skip. Both lanes run on moq-lite and every supported IETF draft except a native JS publisher's moq-transport cells, with the upstream-FETCH refusal named as the reason.
Stale FETCH-serving claims corrected.doc/lib/js/net.md, the SUBSCRIBE_OK.largest comment in publisher.ts, and the LARGEST_OBJECT doc comment in publisher.test.ts now describe the draft-18 whole-group start from object 0 instead of a joining FETCH.
The now-impossible joining-FETCH quest is gone.quest/m3/ietf-js-joining-fetch.md is deleted, its quest/m3/README.md link removed, and quest/m3/audit-ci.md notes that the JS half no longer exists.
The prior review's blocking finding is resolved by the first two bullets: the CLI FETCH and CLI FETCH history lanes no longer fail on an intended removal, and the README no longer claims coverage the publisher cannot provide. I verified the skip guard's version/publisher conditions against VERSION_LIST (set per pass), the two run_fetch_round call sites, and the absence of any remaining ietf-js-joining-fetch reference; bash -n test/interop/interop.sh passes. No Rust or TS behavior changed in this delta.
Cloudflare's relay answers SUBSCRIBE_TRACKS, so restore moq-net's listing
and the PUBLISH adoption, paused/resuming states, and DUPLICATE_SUBSCRIPTION
it needs, as on main. Docs and the ietf-scope epic record why.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
changed the title
refactor(net): drop JS moq-transport FETCH serving and Rust track listing
refactor(net): drop JS moq-transport FETCH serving
Nov 11, 2026
The reason will be displayed to describe this comment to others. Learn more.
Automated review by review (OpenAI)
Reviewed the update from 453b165 through b87eb24 against the base. No new actionable findings.
Direction: retaining Rust SUBSCRIBE_TRACKS listing is consistent with the revised scope. The comparison confirms no remaining Rust changes relative to the base: PUBLISH adoption, paused/resuming aliases, request decoding, and fuzz dispatch/seeds are restored together. The previously reviewed JS FETCH refusal, draft-18 whole-group fallback, and explicit unsupported interop-cell skips remain unchanged; the documentation now distinguishes receiving a listing request from requesting a third-party listing.
Verification limits: GitHub-only static review; no builds, tests, fuzzing, or third-party interop runs performed. Author-reported checks were not independently verified. GitHub currently reports merge conflicts, so the eventual conflict resolution still needs review.
The reason will be displayed to describe this comment to others. Learn more.
Important
The branch no longer merges into main (mergeable_state: dirty), and three of the quest edits target quest/m3 files that main rewrote or deleted since this branch's last merge. The Rust revert itself is clean: rs/ no longer appears in the PR diff, so the PR is now the 24 JS, doc, and interop files only.
Reviewed changes — the delta since the prior pullfrog review (453b165cf..b87eb2473, 48 files, +3382/−209; only the docs and one quest README survive in the PR diff):
The Rust listing removal is reverted whole.rs/moq-net (including ietf/subscribe_tracks.rs, ietf/subscriber/tracks.rs, request_stream::Cancel, RequestKind::Publish, DUPLICATE_SUBSCRIPTION) and the committed fuzz/regressions/ietf/* inputs are back to the base, along with IETF_KINDS = 27, so no Rust file is part of this PR any more.
The Rust listing claims return to the docs.doc/concept/standard.md, doc/concept/stats.md, and quest/m5/net-wrap/e2ee.md are base text again, and their statements about Rust answering FETCH and listing a peer's tracks with SUBSCRIBE_TRACKS match the restored code.
The listing decision is recorded.quest/m2/ietf-scope/README.md adds the "Kept (decided 2026-11-10, maintainer)" bullet, naming Cloudflare's relay as the consumer and noting that a received SUBSCRIBE_TRACKS stays refused.
The two doc lines that remain changed (the one-group-at-a-time paragraph and the "any FETCH to JavaScript" refusal list) still describe the shipped JS behavior, and nothing under quest/, doc/, or test/ refers to ietf-scope/serving any more.
⚠️ The quest notes now target files main has deleted or rewritten
main is at 114b14fb6 while this branch last merged it at 30d7aed1b. Since then the audit consolidation reshaped quest/m3, so the PR's note edits no longer apply: quest/m3/audit-ci.md does not exist on main, and the js-web paragraph in quest/m3/audit-interop.md that the PR rewrites is gone too. Resolving this by keeping the branch's side would resurrect deleted files and reword text main removed.
Technical details
# Quest notes target pre-#8437 audit files## Affected sites-`quest/m3/audit-ci.md:92-94` — the PR replaces the "JS joining FETCH owns the JS half" wording, but `main` deleted this file (#8437 merged the 13 recurring audits into six). Modify/delete conflict; the note has nowhere to land.
-`quest/m3/audit-interop.md:101-104` — main's version no longer contains the "Every `js-web` publisher cell behind a joining FETCH" paragraph this PR rewrites.
-`quest/m3/README.md:124` — `main` still links `/quest/m3/ietf-js-joining-fetch.md`, which this PR deletes; the link removal must be re-applied at the resolved head or the link breaks.
-`quest/m2/ietf-scope/README.md` — main changed only the `Related` link (`audit-simplify.md` → `audit-code.md`); the new "Kept" bullet is unaffected but the file is in conflict.
-`js/net/src/internal.ts`, `js/net/src/ietf/publisher.ts`, `js/net/src/integration.test.ts`, `js/justfile` — main also changed these, in unrelated regions (the `arrivalRuns` hook, `saturatingAdd`/`checkedAdd` in `#refusal`, the `until()` deadline, a `tla` recipe). Keep main's hunks.
## Required outcome- Merge `main` and re-apply the PR's quest notes onto main's files: drop the `audit-ci.md` edit (or move the JS-half note to whichever consolidated audit now owns it), and re-target the `audit-interop.md` note to its new structure.
-`quest/m3/README.md` must not link the deleted `ietf-js-joining-fetch.md` at the resolved head.
- Confirm `rs/` still matches `main` after resolution: main changed `rs/moq-net/src/model/track.rs`, `ledger.rs`, `lite/publisher.rs`, `tests/integration/socket.rs` and added `rs/moq-net/verify/*` after this branch's merge, so a careless "keep ours" would resurrect pre-main code.
## Open questions for the human- Does the "The JS half is gone" note still earn a home once the audits are consolidated, or is it covered by `doc/concept/standard.md` alone?
The reason will be displayed to describe this comment to others. Learn more.
✅ No new issues found.
Reviewed changes — the delta since the prior pullfrog review (b87eb2473..6b65f945a, the merge of main at 114b14fb6; 4 quest/doc files):
The branch merges main cleanly again. GitHub now reports mergeable: MERGEABLE (was dirty), and the merge touches no code: git diff --name-status against main lists only the 23 intended PR files.
The quest/m3/audit-ci.md edit is dropped.main deleted that file when #8437 consolidated the audits, so the modify/delete conflict is resolved by keeping main's deletion rather than resurrecting the file; no "JS half" note is left orphaned.
The ietf-js-joining-fetch.md link removal is re-applied.quest/m3/README.md no longer links the quest this PR deletes, and main's new Cost audit entry is kept.
The audit-interop.md note is re-targeted onto main's structure. The FETCH-refusal bullet now extends main's "Unsupported or down" list instead of rewording the paragraph main removed; the resolved text has no duplicated lines.
quest/m2/ietf-scope/README.md keeps the "Kept" decision bullet on top of main's Related link change.
I verified that rs/ still matches main byte for byte (the prior run's "keep ours" risk), that main's own edits inside the touched JS files survived (arrivalRuns in internal.ts, until() in integration.test.ts, saturatingAdd/checkedAdd in publisher.ts's #refusal, the tla recipe in js/justfile), and that no conflict markers or references to ietf-js-joining-fetch, ietf-scope/serving, or audit-ci.md remain anywhere in the tree. No specialist dispatch was warranted: the delta is conflict resolution in quest notes, with no behavioral surface to probe.
quest land watch: the merge queue removed this pull request at 442b670 because its merge_group build failed (failed_checks).
The land fixer works on it in the background, and the watcher re-enqueues it at the back of the queue once it is fixed or judged flaky, or hands it back here with what it needs.
codex, gpt-6.1-sol, medium effort land fixer note:
The scratch merge with base reproduces the queue failure. Base alone passes nix develop --command bash sh/quest.sh check (518 documents). Base has quest/m2/ietf-scope/serving.md; this PR deletes it, leaving the README without children and requiring a sized title. Choosing the size of the remaining work needs a maintainer decision, and the fixer instructions say never to edit quest/. No PR branch was changed.
Failure log lines 300–304:
ci: check quest failed in 2s
Dev-shell quest: /nix/store/pwg8xw04qkakb329k00lfrw99vrb0rb9-quest-0.1.0/bin/quest
quest: quest/m2/ietf-scope/README.md: quest title must be '# [XS|S|M|L|XL] Title'
Ran the full selected gate with nix develop --command just ci base on the scratch merge. Every other selected module passed: Biome, JavaScript checks/builds/tests, audio quality, Markdown, repository tooling checks/tests, shell, large files, text, TOML, Nix, and justfile. Only check quest failed. The checkout is restored to base; the result file remains uncommitted.
codex, gpt-6.1-sol, medium effort land fixer: nix develop --command just ci base fails: quest: quest/m2/ietf-scope/README.md: quest title must be '# [XS|S|M|L|XL] Title'; deleting the last child makes the epic a standalone quest.. Recommendation: Choose a size for the remaining work and add its tag to the README title; the fixer may not edit quest/.
ci: check quest failed in 2s
ci: check shell cancelled in 2s
##[group]check quest
Dev-shell quest: /nix/store/pwg8xw04qkakb329k00lfrw99vrb0rb9-quest-0.1.0/bin/quest
quest: quest/m2/ietf-scope/README.md: quest title must be '# [XS|S|M|L|XL] Title'
The reason will be displayed to describe this comment to others. Learn more.
✅ No new issues found.
Reviewed changes — the delta since the prior pullfrog review (442b6706b..f9cbd2e2b) is the quest README rewrite plus a merge of main; no JS behavior changed.
The epic README becomes a standalone [S] docs quest.quest/m2/ietf-scope/README.md retitles to [S] Docs state the frozen moq-transport scope, restates the goal as the remaining docs work (doc/concept/, the IETF draft, and the interop matrix), records the 2026-11-11 sizing decision, and gains the [S] bracket that a leaf quest needs once the last child is deleted. No child links remain, and its Related link to audit-code.md resolves.
The merge keeps main's newer changes inside the PR-touched files.#8444's in-band AUTH removal and #8474's MAX_CACHE_DURATION change sit alongside the PR's FETCH and SUBSCRIBE_TRACKS refusals, and git diff --name-status main...HEAD still lists only the 22 intended files (3 deletions); rs/ remains byte-identical to main.
Verified at the head:bunx tsc --noEmit -p net/tsconfig.json is clean, cd js && bun test net/src reports 3106 pass / 1 skip / 0 fail, bash -n passes on test/interop/interop.sh and sh/gh/nightly.sh, and nothing in the tree references ietf-scope/serving or ietf-js-joining-fetch.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
kixelated
changed the title
kixelated
left a comment
![pullfrog[bot]](https://avatars.githubusercontent.com/in/1768019?s=60&v=4){kind=small}
Finishes quest/m2/ietf-scope/serving.md, part of the IETF scope freeze.
Summary
NOT_SUPPORTEDon its own stream; the session stays up. DeletedrunFetchand its range, join registry, and FETCH_OK writer, plus thegroupEndhook only they used. The draft-20+ fill and#writeFetchGroupstay.rs/is unchanged frommain. Cloudflare's relay (cloudflare/moq-rs,moq-transport/src/message/subscribe_tracks.rsandmoq-relay-ietf/src/producer.rs) answers SUBSCRIBE_TRACKS with PUBLISH. Without our side we could no longer list tracks on that relay.subscribe_tracks.tsis gone and the type is a bareSUBSCRIBE_TRACKS_IDconstant. Rust still decodes the whole message, because it sends one when listing.doc/concept/standard.md,doc/lib/js/net.md) and quest notes (m3/audit-interop.md, which lists FETCH cases with a moq-dev-js publisher as unsupported, and them2/ietf-scopeepic README, which now records that listing stays) describe the new scope.Decisions
main(rejected: delete it). Decided 2026-11-10 by the maintainer after the PR left the merge queue. The earlier premise that no third party consumes listings was wrong: Cloudflare's relay implements SUBSCRIBE_TRACKS.FORWARD=0PUBLISH. So listing needs PUBLISH adoption inSubscribeOnce, the paused/resuming alias states that a reader resumes instead of subscribing again, andDUPLICATE_SUBSCRIPTIONfor a redundant draft-18 PUBLISH. Restoringrs/whole frommainis also the simplest result. It drops two of the earlier choices: the Rust "refuse SUBSCRIBE_TRACKS without decoding" and the IETF fuzz-kind renumbering, so the fuzz regression inputs are back as onmain.quest/m2/ietf-scope/README.md[S]and rewrite it as a standalone quest for the remaining docs work (doc/concept/,drafts/draft-lcurley-moq-ietf.mdif it lists the extensions, the interop matrix) (rejected: finishing the docs in this PR, XS). Decided 2026-11-11 by the maintainer after the merge queue removed the PR: deleting the last child made the README a quest without a size, failingquest check.Choices made without asking
CLI FETCHandCLI FETCH historylanes can no longer pass. They are skipped there with a named reason, andtest/interop/README.mdsays so. Over moq-lite, which our JS publishers use, both lanes still run and pass. Also deletedquest/m3/ietf-js-joining-fetch.md, whose goal this PR makes moot.Impact
Public API: none. Nothing in js/net's public API exposed FETCH serving; moq-net is unchanged.
Wire: no format change. A FETCH to a JS publisher (a relay's upstream fill included) gets NOT_SUPPORTED, an optional behavior in moq-transport. A JS draft-18 Next Object subscription now carries objects from 0 in its first group. Rust's SUBSCRIBE_TRACKS sending and both sides' refusal are unchanged.
Caused by
Not a bug fix: the freeze removes features added on purpose (moq-dev/moq#6945, moq-dev/moq#7793, moq-dev/moq#7794, moq-dev/moq#8117, moq-dev/moq#8231).
Checks
On the head that restores listing (
just checkrerun after mergingmainagain on 2026-11-11, which also took main's new FETCH-serving tests out offetch.test.ts):just check: passed.just test interop --all: everything passed except one moq-lite-06rust -> js-native-buncell, which received only audio before the connection closed under load. It passed on rerun. This PR does not touch moq-lite orrs/.just test interop --all --versions ietf: everything passed except onejs -> gstcell (a 20 ms audio timestamp gap), with the 12 JS-publisherCLI FETCHcells per draft skipped as above. Six reruns ofjs -> gstacross the IETF drafts all passed.Follow-ups
None. The earlier js-web rerun note is gone:
main's rewrite of the interop audit already has it confirm every red cell after the runner client bump.Models
Size: M
Implementer: Claude Code, Claude Opus 5.5, medium effort
Implementer: Claude Code, claude-opus-5-5, medium effort (iterate, M)
Implementer: Claude Code, claude-opus-5-5, low effort (iterate, S)
🤖 Generated with Claude Code
(Written by Claude Opus 5.5 with medium)
Review
Rewritten by
quest reviewafter each round.01a1cc1d-600d-7ba1-9fb7-42f7f98d89ef, round 1The reduced FETCH-serving scope is consistently reflected in dispatch, documentation, and interop selection, with no actionable defects found; builds and tests were not run, as instructed.