[kixelated]

Problem

JavaScript refused compressed lite-07 tracks and could not publish them through the networking model. Rust publishers already use the negotiated format, so JS readers need the same bounded storage and plain-frame contract.

Approach

Add Info.compress, keeping model writes synchronous. Store each group's DEFLATE slices once, forward those slices unchanged on lite-07, and build one exact-size plain copy shared by application readers and older sessions. Datagrams have independent windows; malformed compressed datagrams are refused before delivery.

Charge declared plain sizes against the existing group limit before inflation, stop excessive output inside the decoder, retain committed prefixes on cuts, and measure refresh against stored bytes. Complete the JS quest and remove its references.

Impact

• Additive public API: optional @moq/net Track.Info.compress, default false. Application frame/datagram payloads remain plain, and writes remain synchronous. Codec classes and wire-ingest hooks remain private.
• Wire: implements the existing lite-07 codec and inflated-size fields. Lite-06 and other sessions receive plain payloads without compression flags. No new format or published-version change.
• Current consumers: JS applications opting into compressed data tracks, and JS subscribers receiving existing Rust compressed-track publishers through moq-relay.
• JS retains bounded stored and declared plain data per group; every plain reader shares the same inflated buffers.

Checks

• Final just ci passes on current main, including the merged fixture repair. No temporary overlay remains.
• All 2,486 net tests pass. Regressions cover shared inflation, hostile declarations, empty frames, synchronous admission independent of readers, cut prefixes, stored-byte refresh, immutable route metadata, FETCH, datagrams and lite-06 fallback.
• just test interop --all passes. Added Rust ↔ JS Node/Bun compressed exchange through moq-relay on lite-07, plus lite-06 plain fallback, with multiple frames per window and an empty frame.
• just drafts check and quest check pass.
• Maintained just js bench compression sweeps 1/4/16 publishers × 1/4/16 readers. Each publisher stores 740 bytes and inflates 71,680 bytes once for the fixture, independent of reader count.
• Consumer-only minified Bun entry: 31,410 gzip bytes versus 19,934 on main. Isolated decoder: 10,969; decoder plus encoder: 13,984, a 3,015-byte static encoder increment.

Alternatives

The maintainer approved the static encoder cost to preserve synchronous writes. Async pending writes and a preparation API were rejected. Both codecs now import one pako module so their helpers are shared.

Follow-ups

No open decisions. Native browser codec work is separately planned. The existing compression migration quest owns consumer migration and removing @moq/flate; this PR keeps that scope separate.

(Written by GPT-6)

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 2 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 4 included reviews currently available.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: ac39b8b4-f1e0-40ee-b9bd-4c03f187ba38

📥 Commits

Reviewing files that changed from the base of the PR and between b15af4a and 4aecc39.

⛔ Files ignored due to path filters: 1

• bun.lock is excluded by !**/*.lock

📒 Files selected for processing: 36

• doc/concept/moq-lite.md
• js/justfile
• js/net/bench/compression.ts
• js/net/bench/fixtures/compression-consumer.ts
• js/net/bench/frames.ts
• js/net/bench/lite-varint.ts
• js/net/package.json
• js/net/src/flate/codec.test.ts
• js/net/src/flate/decoder.ts
• js/net/src/flate/encoder.ts
• js/net/src/flate/group.test.ts
• js/net/src/flate/group.ts
• js/net/src/group.ts
• js/net/src/internal.ts
• js/net/src/lite/compression.test.ts
• js/net/src/lite/datagram.test.ts
• js/net/src/lite/datagram.ts
• js/net/src/lite/group.test.ts
• js/net/src/lite/group.ts
• js/net/src/lite/publisher.ts
• js/net/src/lite/subscriber.ts
• js/net/src/resume.test.ts
• js/net/src/resume.ts
• js/net/src/track.ts
• quest/m1/audit-size.md
• quest/m1/compress/README.md
• quest/m1/compress/app-range-widen.md
• quest/m1/compress/js.md
• quest/m1/compress/migrate.md
• quest/m1/lite07-finalize.md
• quest/m2/native-compression-adopt.md
• test/interop/README.md
• test/interop/clients/js-native/tail.ts
• test/interop/clients/rust/tail.rs
• test/interop/interop.sh
• test/interop/lite-varint.ts

• 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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-22T11:02:57.129922Z	4aecc39	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[kixelated]

Quest outcome: complete, with the maintainer-approved synchronous write contract and static encoder tradeoff. Final just ci, all 2,486 net tests, full just test interop --all, drafts and quest checks pass on the published head. Codex completed its review of 231a78db6e24ece0f021d648c533454e85d91681 with no findings.

The maintained publisher × reader benchmark confirms one shared inflated copy, independent of reader count. The standard sh/size/bundles.ts first-load gzip measurements are net 113,506 bytes, watch 161,891, watch-ui 174,261 and publish 150,744. The separate consumer-only probe and 3,015-byte isolated encoder increment are recorded in the PR body.

No open decisions or additional follow-up quests. Existing consumer migration and native codec plans retain their separate scope. The PR is ready; merge and queue actions remain with the batch coordinator.

(Written by GPT-6)

[kixelated]

Merged origin/main into the branch (2102dba); no force-push.

• js/net/src/track.ts: kept both static hook assignments, this PR's hooks.writeDatagram and main's sourceStart from moq-dev/moq#6206. The rest of main's js/net changes (feat(js)!: finish a broadcast cleanly, serving its tail until stale #6151 linger, fix(net): preserve source start for descending JS subscriptions #6206 held first group, fix(net): abandon parked group opens on expiry and eviction #6214 parked opens) auto-merged and are independent of the compression paths.
• Quests: migrate.md keeps main's two new Required entries and drops the finished JavaScript one; app-range-widen.md no longer links the deleted compress/js.md and now just says to mirror the regression in js/net.
• test/interop/interop.sh: tail results now name their lane (compress moq-lite-07-wip rust -> js-native-node), so a failing compression lane isn't printed identically to the plain one.

On the merged head: just check passes (net 2,511 tests, 0 fail), just test interop --tail passes every plain and compression lane on lite-07-wip and lite-06 for js-native-node and js-native-bun, and the rust/js/js-native matrix passes. The Codex review of 231a78d had no findings, and everything since is mechanical, so enabling auto-merge.

(Written by Claude Opus 5.5)

[kixelated]

Merged main after the queue dropped this for a conflict with moq-dev/moq#6150 (lite-05/06 timestamp dropping).

• frameDecoder and readFrames now take one internal FrameFormat { scale, timed, compress } so both the dropped-timestamp and declared-size reads apply; callers in the subscriber, bench, and test/interop/lite-varint.ts updated.
• The publisher threads main's Stamp alongside compress everywhere timescale used to go; TRACK_INFO keeps liteTimescale(info) plus the compression flag.
• infoDefaults keeps both the compress type check and main's unchecked validation.
• quest/m2/native-compression-adopt.md (from moq-dev/moq#6227) required the js.md quest this PR deletes; dropped that finished blocker.

Checks: just check, all 2,519 js/net tests, biome, quest check --base origin/main, and just test interop --all pass.

(Written by Claude Opus 5.5)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f87cd23c94

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[kixelated]

Merged origin/main (4aecc39) after the queue dropped this for a conflict.

• quest/m1/compress/README.md: both sides removed their own Required entry (js.md here, ietf.md via feat(net): negotiate IETF compressed tracks from draft17 #6228), so both lines are gone.
• test/interop: test(interop): run the matrix on each even moq-transport draft #6246 moved the tails into a per-version run_tails pass. The compress lanes now pin their own lite versions like the drop lanes (COMPRESS_VERSIONS=(moq-lite-07-wip moq-lite-06)), run only in the auto pass, record as case=compress version=<v>, and are planned in plan_pass. run_tail_pair <pub> <sub> [<mode> <version>] keeps main's pinned-version/join path when no mode is given, and the version check now covers the compress lanes.
• tail.rs/tail.ts: compress joins main's validated mode list; Rust's new write() takes compress, with a shared sizes() for writer and reader.

Checks: quest check (and --base origin/main), js/net tsc + 2524 tests, biome, interop client tsc, clippy + rustfmt on interop-tail, and interop.sh --tail with rust/node/bun: all 21 cases pass, including the 8 compress lanes.

(Written by Claude Opus 5.5)