[kixelated]

Plans the open issues without the quest label (#7573, #7579, #7646). It also records a /quest-import grill that reshaped how time-addressing handles timestamps. Each issue is closed by the PR that finishes its quest, not by this one.

New m1 quests

• JSON writes take a broadcast-clock timestamp (moq-c: a JSON write without a timestamp is ordered on the broadcast clock, which C media can't read, so every media track advertises ~8 s delay #7573): every binding's JSON write requires a timestamp, moq-c's NULL fails, and C gains moq_publish_catalog_now(). C media then stops advertising the 8 s delay.
• Catalog delay is the least a track trails (Catalog delay counts B-frame reorder: Order compares presentation timestamps written in decode order #7579): the estimator keeps the minimum trail, so B-frame reorder never becomes delay. No SPS depth.
• A lapsed request token resets with EXPIRED_AUTH_TOKEN (moq-net auth: expiry vs deny codes, renewal on session-authorized requests, tokenless PUBLISH_NAMESPACE, kind/path on auth::Request #7646 point 1).
• @moq/net group IDs past 2^53: needed now that legacy group IDs pass through verbatim.

#7646 points 2 to 4 go to the open ietf-token-auth PR #7211. I'm passing them to that PR's sub-agent, so they aren't in this PR.

Time-addressing edits

model.md, ietf.md, lite07.md, bindings.md, js.md, plus a note in max-age-defaults.md:

• Every producer write takes a timestamp, and nothing is ever wall-clock stamped. This replaces the 2026-10-27 audit's "producer arguments keep an optional timestamp", which recorded only sequencing.
• Legacy group IDs pass through verbatim as keys, with no sequence beside them.
• Scale-less tracks keep lite-07's Timescale 0. Their frames carry their group's key, and max-age counts bare units.
• One unit per track for keys and frames.
• moq-mux's catalog producer keeps a broadcast all timed or all scale-less.
• moq-transport drops our Timestamp property. LOC properties are proxied only between moq-transport peers, and Prior Group ID Gap stays as a best-effort drop signal.
• Left open: how lite-05/06 map on the way out.

Decisions (2026-11-03, maintainer, /quest-import)

• moq-c: a JSON write without a timestamp is ordered on the broadcast clock, which C media can't read, so every media track advertises ~8 s delay #7573 fix: NULL writes unordered; ✅ require a timestamp, with C now(); won't fix.
• moq-c: a JSON write without a timestamp is ordered on the broadcast clock, which C media can't read, so every media track advertises ~8 s delay #7573 scope: ✅ every binding in one quest; moq-c only; moq-c now, others later.
• Catalog delay counts B-frame reorder: Order compares presentation timestamps written in decode order #7579 fix: ✅ delay = minimum trail; subtract declared reorder; decode edges with SPS depth (rejected).
• moq-net auth: expiry vs deny codes, renewal on session-authorized requests, tokenless PUBLISH_NAMESPACE, kind/path on auth::Request #7646 expiry: no wire change; ✅ reset with EXPIRED_AUTH_TOKEN; won't fix.
• moq-net auth: expiry vs deny codes, renewal on session-authorized requests, tokenless PUBLISH_NAMESPACE, kind/path on auth::Request #7646 session-authorized update: ✅ keep refusing; allow moving to T's auth.
• moq-net auth: expiry vs deny codes, renewal on session-authorized requests, tokenless PUBLISH_NAMESPACE, kind/path on auth::Request #7646 tokenless out-of-limit PUBLISH_NAMESPACE: ✅ refuse UNAUTHORIZED; keep the hold.
• moq-net auth: expiry vs deny codes, renewal on session-authorized requests, tokenless PUBLISH_NAMESPACE, kind/path on auth::Request #7646 request kind/path: own quest; ✅ fold into ietf-token-auth; no.
• Ingest key: ✅ ID verbatim; first PTS, renumbered; per wire.
• Sequence beside the key: ✅ no; yes.
• Timestamp property: ✅ stop, LOC proxied only between moq-transport peers; read but don't send; keep sending.
• Prior Group ID Gap: drop; ✅ keep as a best-effort SUBSCRIBE_DROP signal.
• Lite-07 Timescale 0: ✅ keep; drop.
• Timescale meaning: ✅ one unit for keys and frames; flag legacy keys.
• Timed-broadcast enforcement: ✅ catalog producer; moq-net broadcast; convention.
• Scale-less frame timestamp: ✅ its group key; any increasing number.
• Scale-less max-age: ✅ bare units; doesn't apply.
• Milestone: ✅ m1 for all four new quests.

Review

Codex gpt-6.1-sol (medium) via quest review --base origin/main, head 1ad01fa06cf163c2bb00c5c149e425f236f1078d: no findings.

(Written by Claude Opus 5.5)

🤖 Generated with Claude Code

[pullfrog]

This run was cancelled before it finished.

  ^{｜ View workflow run ｜ via Pullfrog}

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: 1ad01fa

• quest/m1/time-addressing/model.md:101–108: Legacy publishers may retain sequential group IDs even when frame PTS uses milliseconds, but quest/m1/time-addressing/consumers.md directs HLS/DASH export to place groups by group start rather than first-frame PTS. Those IDs are not presentation timestamps, so the two plans can place legacy segments at incorrect times or collide on the grid. Preserve frame-PTS placement when keys are not known timestamps, or specify an equivalent mapping, and cover legacy ingest in export tests.

The explicit key/clock distinction is a useful direction; aligning the consumer plan with it is simpler than assuming all retained keys are timestamps. Static review only; I did not run tests.