MOQ Wiki

Discussions - November 2026

Updated

Summary of active discussions in the MOQ ecosystem during November 2026. The late-October run-up lives in discussions-2026-10.

Activity (Oct 29 → Nov 1) — The post-interim lull is over on both layers: afrind merges eight draft-23 PRs into moq-transport (Oct-29→31), Will Law opens a five-PR MSF catalog/timeline burst, and the AUTH design team’s work reaches a privacy-pass PR — while the implementations run a clear draft-22 adoption wave ahead of the ~two-weeks-out IETF 127 hackathon: moqtransport cuts v0.15.0 (draft-22 alongside -18), moqlivemock brings its whole stack onto draft-22 (#164), moq-rs cuts v0.7.31, moqtail cuts its first relay release since July (relay@0.16.0), and moq-dev cuts v0.17.10 atop 200+ merges. The interop nightly settles off the Oct-29 spike into a 47–48% band (272/562 on Nov-1) as “ahead”/draft-22 cells creep up. Two new individual drafts surface — MoQ Feedback -01 and a volumetric-video draft-yuyou-moq-v3c-00.

With the 10/26 interim’s queue cleared, the WG’s editor turned a long backlog of agreed issues into merged text, while the media-format editor opened a fresh round of catalog/timeline PRs. The headline across the implementations is uniform: everyone is lighting up draft-22 alongside draft-18 — the interop target pair from interim-24 and Seattle, and the target set for the IETF 127 hackathon now roughly two weeks out. Slack was not checked this run.

Spec / WG: afrind clears eight draft-23 PRs; MSF opens a catalog/timeline burst

  • moq-transport lands eight merges Oct-29 → Oct-31 (all afrind; all draft-23 material — transport stays draft-22, no release), clearing much of the Oct interim queue:
    • #2040 “Describe where timeouts apply and the tradeoffs in setting them” (+21/−0, fixes #1743).
    • #2011 “Identify the unsupported property in UNSUPPORTED_EXTENSION” (+14/−0, fixes #1951) — adds an UNSUPPORTED_PROPERTY Error Detail carrying the Property Type of each Mandatory Track Property the sender doesn’t understand.
    • #2017 “Use a varint length for control messages” (+42/−21, fixes #1998) — the control-message Length becomes a varint bounded by a new MAX_CONTROL_MESSAGE_SIZE Setup Option (default 65535).
    • #2013 “Mark the end of a complete Group in FETCH” (+8/−5, fixes #1934) — extends the End-of-Track rule to Groups via an End of Group serialization flag.
    • #2018 “Allow overlapping namespace subscriptions” (+18/−32, fixes #1792) — removes the prefix-overlap prohibition for SUBSCRIBE_NAMESPACE / SUBSCRIBE_TRACKS (and the PREFIX_OVERLAP error), adding a SUBSCRIBE_TRACKS_ID parameter so each resulting PUBLISH identifies its SUBSCRIBE_TRACKS.
    • #2019 “Define the effects of updating a namespace subscription prefix” (+14/−6, fixes #2001).
    • #2042 “Prohibit coalescing REQUEST_UPDATEs that contain FILL_PARAMETERS” (+4/−1, fixes #1992).
    • #2039 “Treat a Property used outside its scope as a malformed Track” (+6/−0, fixes #2028).
  • New open PRs keep the draft-23 queue moving: #2049 “Default FILL_PARAMETERS Group Order to Ascending” (afrind), #2047 “Rename AUTHORIZATION_TOKEN to AUTH_PAYLOAD” (sharmafb), and #2045 “Describe Setup Options in 0-RTT” (afrind). The duplicate-Property fork (#2044 Hosna vs #2038 afrind) is still open. New issue #2048 (Swett) — “Fill-Fetch can send a lot of extra Objects with no (fast) way of cancelling them” — adds to the running Fill-Fetch design thread (#2043/#2034/#2032).
  • MSF: a five-PR editorial burst, all open (Will Law, Oct-30) reopening the catalog/media-timeline surface after the Oct-28 merges: #220 “Rename ‘track role’ to ‘track type’ and remove mimeType”, #219 “Clarifications on media timeline generation and modification”, #218 “Define accessibleWindowTime”, #217 “Fix media timeline catalog requirement”, and #216 “Updates to wallclock time reference and depends attribute”. None merged in the window, so the moq-msf page’s durable facts are unchanged — this is proposal-stage churn to watch into IETF 127.
  • moq-privacy-pass: AUTH-design-team text reaches a PR — privacy-pass #26 “Update moq privacy pass following progress of AUTH design team” (thibmeu, Oct-30, open). First privacy-pass movement reflecting the AUTH design team’s work (which earlier produced the englishm/moq-transport#12 SETUP-stream auth text); a signal the auth strand is converging across the privacy-pass / CAT / design-team venues.
  • loc, secure-objects, cmsf, catalog-format otherwise flat (cmsf’s deisenbacher capability-preflight PR #29 of Oct-15 is now echoed on the list — see below).

Drafts / mailing list: MoQ Feedback -01, CMSF capability preflight, IETF 127 agenda, a Feb interim

  • MoQ Feedback revs to -01 (draft-liu-moq-feedback-01, 2026-10-30): Yanmei Liu (Alibaba) posted a “-01: summary of changes since -00” note to the list and separately requested an IETF 127 agenda slot to present the draft — the first WG-meeting-facing push for the feedback/quality-signalling category. See moq-feedback.
  • A new volumetric-video individual draft surfaces — draft-yuyou-moq-v3c-00 (datatracker 2026-10-29), “Transport of V3C Bitstreams over MOQT” — V3C (Visual Volumetric Video-based Coding) over MoQ, from the same yuyou author as moq-conditional-filtering. Not yet discussed on the list; logged as newly spotted, a media-packaging direction to watch (no wiki page yet).
  • “Capability preflight for CMSF” (David C. Eisenbacher, list, 2026-10-30) brings the cmsf #29 idea to the WG — a way to verify a client’s capabilities before authorizing a track in CMSF, with content protection as the first probe type.
  • IETF 127 agenda forming: Will Law replied on the “IETF 127 Agenda Requests” thread and Yanmei Liu filed an “IETF 127 Agenda Request — MoQ Feedback” (both Oct-30). (For context, the WG’s IETF 127 sessions were scheduled Oct-23 and announced by Westerlund Oct-19.)
  • Interim scheduling: Duke replied in “Proposals to host an interim 8-11 February” (Oct-29) — a distinct thread from his Oct-26 “Proposed Virtual Interim dates” (Nov 30 / Dec 14 / Jan 11 / Jan 25), whose Nov 9 concerns deadline still stands with no reply in the window.
  • The weekly GitHub digest resumed Nov-1 (“Weekly github digest (Media Over QUIC Activity Summary)”, Repository Activity Summary Bot, 2026-11-01 — the first since Oct-25) and there was no consensus call in the window. MoQ Monthly still #2 (2026-05-31).

Implementations: a draft-22 adoption wave ahead of the IETF 127 hackathon

  • moqtransport v0.15.0 (Oct-30, “draft-22 alongside draft-18”, #29, +7,811/−1,435, tobbee) — the Go library now speaks both draft-22 and draft-18 in one build, ALPN-selected per session.
  • moqlivemock #164 (Oct-31, +699/−337) brings mlmpub/mlmsub/mlmrel/mlmtest onto draft-22 alongside draft-18 on moqtransport v0.15.0 — so a relay can bridge a draft-18 upstream to a draft-22 downstream; on main, no new tag (v0.17.0 stands). Companion #163 fixes multi-publisher namespace retention in mlmrel (keep a namespace until its last publisher withdraws).
  • moq-rs v0.7.31 (Oct-31, moq-transport crate v0.16.5) — restore draft-18 FETCH in moq-pub/moq-sub (#303), bound request-error reasons (#299), reset stopped subgroups until FIN ack (#298).
  • moqtail cut its first tagged release train since July — relay@0.16.0 + moqtail-rs@0.16.0 + moqtail@0.14.0 (Oct-30), finally tagging the July→Oct relay work. Merge #451 (lminiero: return the track alias in a subscribe response); new open #452 (kkroo: an MMTP track receiver, tracking the MMT interest).
  • moq v0.17.10 (Oct-30; moq-cli v0.14.10, libmoq v0.6.20, moq-gst/moq-ffi v0.4.20, obs-moq v0.6.20) atop a 200+ “codex” merge run tilting to security/transport hardening: client-cert verification on tls:// (#7358), keep-alive from the negotiated idle timeout (#7347), an errored track end keeps its history (#7349), max_delay defaults to 0 (#7403), ES256/EdDSA-only peer grants (#7303), surround Opus through CMAF dOps (#7336); CI now requires Moxygen interop in the smoke gate (#7415) and a lite-07 padding datagram is being reserved before the next moq-lite cut (#7419). The churn only accelerated on Nov-1 — ~106 merges in one day (188 since the v0.17.10 tag), still no new release: lite-07 padding datagrams defined (#7435), a breaking chore! remove WASM support (#7475), zeroing the automatic keep-alive interval (#7449), rejecting truncated frame headers (#7444), and three congestion-control candidates opened — SCReAMv2 (#7547), GCC (#7546) and BBR3 (#7544) — plus a feat(net)! fetch as a range part 1 (#7545, open). No tag → these stay in log, not on the moq-dev page.
  • google/quiche moqt (not flat): three commits — PUBLISH_STATE_NOTIFY parser/framer + delete request-id from GOAWAY (c592dffc0, Oct-29), consolidate SubscriptionFilter anchoring + LARGEST_OBJECT validation (211f88f2b, Oct-30), update kMaxGroupId/kMaxObjectId (cc6c73f82, Oct-30).
  • warp-player joins the draft-22 wave — open #207 (tobbee, Nov-1) “feat(transport): speak MOQT draft-22 alongside draft-18”, extending the Eyevinn draft-22 adoption to the browser player (v0.16.0 tag still stands; not yet merged).
  • Flat: moq-js (#73 open), birneee/quiche_moq.

Interop: settles off the Oct-29 spike into a 47–48% band

  • The nightly matrix held at 562 cells four straight nights (Oct-29 → Nov-1) and the pass count settled off the Oct-29 49.5% high: 278 → 266 → 268 → 272. The Nov-1 00:19 cut — 562 / 272 pass / 255 fail / 29 skip (48.4%), 352 at-target · 164 ahead · 46 behind. The one steady drift is “ahead” creeping 161 → 164 (at-target 355 → 352) as endpoints declare draft-22 against the draft-18 target — the implementation-side draft-22 wave starting to show in the roster. Two new open runner PRs in the window: #180 (riedlse — a CMSF test area with a cmsf-live-join case and a stitcher-moq-media client, Oct-30) and #181 (OJII3 — registering moq-mbt, a MoonBit draft-18 client, Oct-31); neither merged. See interop-runner.

Graph View

DraftsConceptsPeopleImplDiscInteropOtherTags
DraftsConceptsPeopleImplDiscInteropOtherTags

Backlinks