Discussions - October 2026

Summary of active discussions in the MOQ ecosystem during October 2026. The late-September run-up lives in discussions-2026-09.

Activity (Sep 30 → Oct 2) — draft-22 lands eleven days before Seattle interop, carrying one wire change that the “identical to -20” framing left out. Draft-14 starts leaving the interop matrix, and the auth design team gets its own channel.

The spec side moved in a single evening: on Thu Oct-1 afrind merged three PRs between 18:15 and 21:31 UTC, tagged draft-ietf-moq-transport-22 at 21:33, and the datatracker posted it at 21:36. He then spent the next hour labelling issues for Seattle. The interop runner shed its first draft-14 cells. On Slack, imquic became the first relay with public draft-20/21 support, and testing found bugs in it within hours.

draft-22 is published, and it is not quite wire-identical to -20

draft-ietf-moq-transport-22 (165 pages, expires 2027-04-04) is the Seattle interop target named at interim-24 together with draft-18. Release notes are PR #1965 (afrind, approved by Suhas).

• One control-plane change. LOCATION_FILTER (0x21) replaces its Length field with an explicit Location Filter Type (#1953, Swett, merged Oct-1 18:17 UTC). The types are 0x00 none, 0x01 relative start, 0x02 absolute start, 0x03 + group end, 0x04 absolute range, 0x05 Next Object; any other value is a PROTOCOL_VIOLATION. A draft-20/21 decoder will misread it. Duke’s Sep-21 “-22 is functionally identical to -20 except the ALPN” therefore holds everywhere except this one parameter.
• Everything else is editorial, most of it merged on main from Sep-6 to Sep-11 and held out of the move-only -21. The last piece was #1946 (afrind, +243/−190, merged 18:15 UTC after vasilvv’s approval four minutes earlier). It restructures Publisher and Namespace Discovery: a Namespace Prefix Matching section, SUBSCRIBE_TRACKS moved under Publishing and Receiving Tracks, and a ladder diagram. Per the PR, 2,343 of 2,379 sentences are unchanged.
• Not in -22: Zero-length Track Names (#1954) and DEFAULT PUBLISHER PRIORITY as a parameter (#1957), both approved, plus INCLUDE_PAYLOAD (#1955), the FETCH relative-filter ban (#1960) and Filter Sets (#1894).

Full itemisation in moq-transport.

WG tracker: three merges, then a pre-Seattle triage pass

• New issues.
  • #1962 (Miniero, Sep-30) asks how FILL_PARAMETERS works with SUBSCRIBE_TRACKS. afrind: the publisher should act as if a SUBSCRIBE had carried them, echo them in PUBLISH and open a fill fetch stream. He concedes the text is unclear and is “not sure if it actually will work.”
  • #1963 and #1964 (Duke, Oct-1) cover parameter restrictions. Delivery timeouts in REQUEST_UPDATE and “any SUBSCRIBE parameter” in SUBSCRIBE_TRACKS (which would include RENDEZVOUS_TIMEOUT) need limits, and EXPIRES should be allowed in every _OK.
• Seattle prep, Oct-1 18:45–19:18 UTC.
  • afrind marked #1948 (NEW_GROUP_REQUEST far-future ID), #1889 (pause-mechanism interaction: core, or the Top-N/SSTS extension docs?) and #1850 (payload-compression properties) BLOCKED with “discuss in Seattle?“.
  • On the Quint-found stale Largest Object bug (#1958): “let’s make a PR to have something to discuss with the wg in Seattle.”
  • On #1959, Swett’s point that a FETCH end may also need a live Largest Object puts the premise of #1960 in doubt.
  • Zanaty and afrind debated multi-range FETCH (#899). Zanaty pointed out that multi-range location filters would bring back the Length field that #1953 just removed.
• Metadata scope (#1857): michalhosna (Sep-30) argues that mid-group relays never see object 0, so SUBGROUP_/OBJECT_DELIVERY_TIMEOUT silently fall back to the Track value. He wants subgroup-level metadata in the data plane.
• Totals: moq-transport 3 merges (#1946, #1953, #1965), 1 new PR, 3 new issues, 2 closed (#841, #1937). Other moq-wg repos had no merges.

Auth: the design team gets a channel, and DPoP folds into C4M

• #moq-auth-design-team (quicdev Slack) was created by Mike English on Sep-30, right after a design-team call. He posted “Chris’s detailed worked example from today’s call” as a Mermaid sequence diagram.
  • The plan is to “add a section to the MOQT draft that describes this new AUTH message and how it gets used”, drafted first in English’s englishm/moq-transport fork before a WG PR.
  • Chris, presumably C4M co-author Chris Lemmons, offered a matching C4M PR to illustrate usage. English wondered whether DPoP “would be a separate draft … or even the main thing we need?”
  • Aman Sharma suggested a skeleton draft with sections people claim. Suhas and Cullen Jennings have joined the channel.
• DPoP into C4M: Suhas’s CAT-4-MOQT PR #53 (Oct-1, +224/−251) “folds generic-dpop-proof into this draft per Security AD guidance.” That would retire draft-nandakumar-moq-generic-dpop-proof as a separate document. No reviews yet.
• The design team has the Oct-15 10:00 slot on the Seattle agenda.

Seattle: agenda rev-04, slides due Oct-8, and a request for more auth time

• Duke, “Seattle Agenda” (Oct-1 17:30 UTC, permalink): “Magnus and I have hammered out a Seattle Agenda.” This is the first list announcement; the Sep-28 upload was never announced. Slides are due by Oct-8, and the MOQT editors are asked to post a detailed issue/PR breakdown.
  • Rev-04 adds “Deployment Results: SWITCH_FROM and Group metadata” (Riedl, Wed 13:20) and doubles the Auth Design Team to 60 minutes (Thu 10:00).
  • Still no slot for MSF, CMSF, LOC or LOCMAF.
• afrind (19:23 UTC, permalink): an hour can’t cover the 11 open auth-labelled MOQT issues. He and Swett offer some of their issue time if the design team brings slides and leads. No chair reply yet.
• Oct-26 confirmed as interim-30 (Duke 20:23 UTC: “There are no comments. 26 October is the date”). The IESG announced it at 21:11 UTC: 16:30–18:00 UTC, agenda “MoQT Issues”, session page.
• Other list traffic: the “I-D Action” for transport-22 (21:36 UTC) and Swett’s explanation of the wire change (Oct-2 01:48 UTC, permalink). “Draftification” (top-N / SSTS) has no replies yet, and its list-discussion deadline is Oct-13. The weekly GitHub digest bot has not posted since Sep-13.
• Datatracker: transport-22 was the only MoQ submission (140 total since Sep-29). All other WG documents are unchanged (loc-04, msf-01, cmsf-01, secure-objects-01, privacy-pass-auth-03, c4m-01). Not renewed: draft-duke-moq-subscribe-rewind-02 (see joining-fetch) expires Oct-4 and herz-moq-nmsf-01 Oct-9. No new liaisons; no new minutes. IETF 127’s sessions are still unscheduled. See interim-meetings.

MSF: media timeline and the DVR window

Gwendal Simon filed msf #213 (Sep-30). The Media Timeline requirements still refer to a type identifier, and he proposes packaging: "mediatimeline" instead.

On msf #205 (DVR) he prefers a catalog field covering both timeline tracks, named accessibleWindowTime (ms) or accessibleWindowGroup rather than “DVR”. He also points out that the first object of each timeline group grows without bound: about 43k records (~1.5 MB) for a 24 h stream with 2 s groups. His proposed fix: “a publisher SHOULD omit records of Groups that are no longer accessible.” No MSF merges.

Slack: imquic goes to v20/v21, and FETCH ordering bugs surface

• Miniero (#moq, Oct-1 10:13 UTC) updated imquic and his public relay (lminiero.it:9000) to v20/v21: LOCATION_FILTER and FILL_PARAMETERS. He expects it “will break in a million pieces once it’s tested with different endpoints”. Fill FETCH works, but more than one FETCH per subscription (one in SUBSCRIBE, another in REQUEST_UPDATE) is probably broken. His runner client also now implements the new rendezvous-timeout case.
  • Aman Sharma found a FETCH ordering bug within hours. With subgroup 0 holding objects 0 and 2 and subgroup 1 holding 1 and 3, the relay sent 0, 2, 1, 3. Miniero fixed ascending order at once; descending order was still broken after a quick second fix. Sharma’s FETCHes were plain standalone requests, not triggered by FILL_PARAMETERS.
• #moq-interop-runner (Sep-30): Yu You ran moxygen’s conformance Section 8 (PUBLISH) against the Nokia relay and everything passed. afrind said results differ over longer RTTs and that the tool has known bugs to be fixed soon.
• #moq-rs, #moq-js, #libquicr: silent.

Interop: draft-14 starts leaving the matrix, and the pass rate hits a high

• Oct-1 00:32 cut: 463 cells / 223 / 238 / 2. Runner #131 (moqx drops d14) merged Sep-30 15:27 UTC and removed 12 draft-14 cells, all already failing. The behind band went 117 → 105, its first change since Sep-2.
• Oct-2 00:28 cut: 463 / 227 / 235 / 1 (49.0%), the best pass rate of any cut with 50+ cells. Most of the +4 is the moqtail relay recovering from 4 → 7 passes with no moqtail commits in between.
• Hidden failure: all 18 docker cells against the moq-dev-rs relay have failed since Sep-24, because moq-relay 0.15 renamed --server-bind → --listen. Runner #134 (riedlse) fixes it and is the first runner PR to declare draft-21/22, for stitcher-moq via moq-tokio 0.19.20.
• Other runner PRs: libquicr → draft-18 (#133) and a namespace-lifecycle test (#135) are open. The nightly still targets draft-18. See interop-runner.

Implementations: imquic reaches v21, moq-dev ships a breaking relay, and the -22 retargeting begins

• imquic merged draft-20/21 to main (#38, Oct-1, +2,110/−704) and runs it on the public relay. Same-day fixes corrected FETCH ordering (see Slack above).
• moq released moq-relay 0.16.0 on Sep-30 (breaking auth parity, #4319; tag and crates.io only). 0.17.0 is already queued (#4596, which removes cluster gossip).
  • About 100 merged PRs, including a conformance push “ahead of the Seattle interop”. #4610 (+1,363/−330) decodes every legal request on drafts 14–22 and answers unsupported ones with NOT_SUPPORTED rather than closing. #4253 finishes the moxygen-compatibility line.
  • The wave ran through the night of Oct-1→2: five more net/mux merges by 03:41 UTC, among them #4685 (+268/−328) refusing a per-request SUBSCRIBE_TRACKS with NOT_SUPPORTED (the data-plane companion to #4610) and abandoned-FETCH / spliced-group fixes (#4689, #4691). #4702 bumps moq-noq to 1.3.3 for the RUSTSEC-2026-0185 advisory — a dependency fix, not a protocol change.
  • moq-lite-07 wire work continues behind a -wip flag (#4455).
• OpenMOQ:
  • moq-playa closed its draft-21 PR #20 in favour of a long-lived feature/draft-21 branch, and is renaming its npm scope to @openmoq/* (#21).
  • moq2ts conforms its catalog and media timeline to MSFTS-02 / MSF-01 (#4, Simon).
  • moqx: CI and release-tarball PRs (incl. #782, gmarzot, making the stats stack private), plus afrind’s pipelined joining-FETCH fix (#783, open). It stays on drafts 16 + 18.
• shaka-player merged SCTE-35 via MSF event timeline tracks (#10668, +1,428/−34). #10670 (catalog join via FILL_PARAMETERS on draft-20/21) is open. v5.3.0 is still unreleased.
• libquicr merged 11 PRs on draft-18. One of them, #953, handles a peer SETUP that arrives before or after the local one, which fixes interop with moqx.
• quiche (Duke): session-level draft-18 GOAWAY, and STOP_SENDING on every Reset().
• moxygen: proxied-PUBLISH failover (Aman Sharma) and per-object moqtest deadlines. kSupportedVersions is still {14, 15, 16, 18}, which keeps runner #132 waiting.
• moq-rs: no merges. New PRs bring back draft-16 Joining FETCH as a standalone FETCH (#237) and keep subscriptions alive when a superseded subgroup is cancelled (#239). HenrySchlesinger also filed seven issues, including the draft-18 relay closing the session on a moq-dev SUBSCRIBE_NAMESPACE (#243).
• moqtail: #394 (malformed upstream FETCH → MALFORMED_TRACK) and a thorough review of the SSTS PR #374. aiomoqt: 0.12.0 work continues on a branch (#40: priority, draft-20), and aiopquic 0.5.0 is in a release PR.
• Quiet: moq-js, birneee/quiche_moq, and all Eyevinn repos (moqlivemock v0.16.1, warp-player v0.16.0, moqtransport v0.14.0 and locmaf remain the newest).
• Not yet on -22: no tracked implementation has announced the draft-22 Location Filter Type. moq-dev’s shim and Paramount’s relay already negotiate moqt-22; whether they use the new filter encoding was not checked.
• Community: MoQ Monthly is still at #2. No wiki issues.