Repository navigation
docs(hang): the hang-04 changelog describes the catalog that ships - #9133
Conversation
Quest-Claim: 4ace436e3bb1d1d6
Rewrite the draft's moq-hang-04 changelog as the change from the published hang-03, dropping steps main took on the way and adding the Discovery, TTML, CMAF run, and LOC framing changes it missed. Complete the list of catalog changes an older consumer can ignore, and give the data track bitrate its range. The shared parity corpus now covers stats, echo, path spelling, includes, the sync clock, every data track field, text, and the remaining video and audio fields. It found Rust accepting a data track bitrate above 2^53 - 1 that JS refuses; Rust now refuses it too. Finishes the hang04 epic. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found.
Approach: sound. Diffing the changelog against the published datatracker hang-03, not release, is the right baseline for a reader of the published draft. The one code change brings Rust's data track bitrate into line with JS and with Rust's own media bitrate.
Reviewed changes
I reviewed all 7 files. I checked the changelog against the published hang-03 table of contents and checked the Rust bound against the existing bitrate fields and the JS schemas.
- Changelog baseline: I checked the rewritten entries against the published
draft-lcurley-moq-hang-03. It had Discovery, a root Clock, Timeline (holdingarchive), Recording, and MPEG-TS Service Information. It had nostats,echo,include, HLS section, Version section, data track estimates, or audiowarmup. So the new "Removed the Discovery section", "Replaced the rootarchiveandclocksections… Removed the Timeline section", and the dropped viewer feedbacklatencyentry all match what was published. - Version section: The longer safe-to-ignore list matches the rule in the paragraph before it: only a
trackother than the ID raises an entry to 4. - Data track
bitraterange: The draft now gives 0 to 2^53 - 1. Rust'sJsonConfig/BinaryConfiguseNotNull<crate::snapshot::Number>, the same asaudio/mod.rsandvideo/mod.rs, and this matchesu53Schemainjs/hang/src/catalog/{json,binary}.ts. - Parity corpus: Each new range field gets a boundary pair (2^53 - 1 accepted, 2^53 refused). Rust and JS both keep an unknown
modeorformatand leave it for selection to skip, and the corpus records that. - Quest bookkeeping: The epic README is deleted, and the links to it in
quest/m2/README.mdandquest/m3/ffi-catalog-skipped.mdare removed.
claude-opus-5-5 | 𝕏
|
No actionable comments were generated in the recent review. WalkthroughThe draft adds catalog field details and reorganizes the hang-04 changelog. The quest documentation marks hang-04 as done, removes its planning document, and removes a related link. JSON and binary bitrate serialization now uses a bounded numeric type. New parity fixtures cover catalog validation and bitrate, jitter, and delay boundaries. Merge Risk:
|
|
Merge pick (maintainer away, granted the coordinator merges it confidently recommends): ✅ /quest-merge, accepting the decisions listed in the PR body. (Written by Claude Opus 5.5 with high) |

Finishes quest/m2/hang04/README.md, the hang-04 epic's own work: the last read before the freeze. Every child had already merged, so this deletes the epic.
Summary
drafts/draft-lcurley-moq-hang.md'smoq-hang-04changelog so it describes the change from the published hang-03 (the datatracker's draft-lcurley-moq-hang-03, which matchescddededdfb). The baseline is notrelease: that branch also carries backports hang-03 never had. I diffed the 03 source against main section by section.latencyremoval (hang-03 never hadstats), and "wall-clock flush lateness" in thejitterentry (a release-only state; hang-03 defined a flush delay with frame-duration examples).includeentries into one, merged the data trackbitrate/jitteranddelayentries, and rewrote thesyncentry to say what replaced what (rootarchiveandclock, and the timeline track) without listing fields that hang-03 never had (replayEpoch).data_offsetcounted from themoof, LOC's codec payload using the legacy framing, andinterlacedas an identity field.preference,instance,color,projection,stereo,interlaced, and audiolayout, which is what the code already does (onlytrackraises an entry to version 4). The data trackbitratenow has a range (0 to 2^53 - 1), so the root rule's "out of range" has something to apply to.rs/hang/fixtures/catalog-parity.json, which bothrs/hangandjs/hang'scatalog/parity.test.tsload, gains 52 cases. Before,statsandechohad only mis-shaped cases. Now there are cases for acceptedstatsandecho, echo andbroadcastpath spelling, include entries, thesync.clocktimescale and wall bounds, everyjsonandbinaryfield and its range, unknown data track modes and text formats, text roles, and the remaining video and audio fields (stereo,stalled,color, VR180,interlaced, a lone display aspect, ambisonic, PCM, versions 3 and 4).jsonorbinarytrackbitrateabove 2^53 - 1, which JS refuses. Rust now bounds it the way it bounds a videobitrate(crate::snapshot::Number).Decisions
Choices made while the maintainer is away; each can be undone in review:
release. The README says hang-03 is what the datatracker andreleasecarry, butreleasehas backported fields (renditiondelay, data track estimates) that the datatracker's 03 lacks. The changelog tells a reader of published 03 what changed, so it has to diff against what was published.../../xat parse time; resolution checks it. The corpus only covers path spelling.modeand textformatare kept at parse. Both languages already keep them, the same way an unknown containerkindis kept, and leave them for selection to skip. The corpus records that rather than changing either side.Validation
bun test src/catalog/parity.test.tsinjs/hang: 422 pass. A first probe failed on the root-escape cases (dropped, see Decisions) and on the unknown mode and format cases (both languages keep those entries; the expectations were corrected).cargo test -p hang --lib: 143 pass. Before the Rust fix,shared_parity_fixturesfailed onjson bitrate beyond 2^53: accepted. That run is the independent check: the same corpus run through both implementations.cddededdfbto main, and the rendered datatracker text of hang-03.just drafts check,just check, andquest check --base origin/mainpass.Public API and wire
hang::catalog::JsonConfig/BinaryConfignow refuse abitrateabove 2^53 - 1, matching JS (a behavior change, not breaking).bitrategains a range, and the changelog and Version text are editorial.Caused by
bitratebounded to 2^53 - 1 in JS but unbounded in Rust (found by check)Follow-ups
None.
Models
Size: S
Implementer: Claude, claude-opus-5-5, high effort
(Written by Claude Opus 5.5 with high)
🤖 Generated with Claude Code
Review
Rewritten by
quest reviewafter each round.Head: bcded05
Base: origin/main at 77dd8fe
Risk: medium,
the base has no RISK.md to name what is always high or customer-facing, so every change is mediumClassified: medium by default
Reviewer: Codex, gpt-6.1-sol, medium effort
Session:
01a1f12e-82d4-70c2-ae55-632db2144546, round 1Score: 9/10,
The Rust bitrate bounds match the existing JavaScript schemas and draft, with shared boundary fixtures and no actionable defects found by source review, though builds and tests were not run as instructed.Rounds: 1; first round 0; findings 0: fixed 0, refuted 0, open 0
Findings: none