You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
A refused TRACK_STATUS upstream fails a lite TRACK_INFO, so the subscriber never subscribes #7423
Behind a moq-transport publisher that refuses TRACK_STATUS on draft 18 (openmoq/moqxr answers it, and FETCH, with NOT_SUPPORTED), a moq-lite-06 subscriber can't start a track through moq-relay on main (e072cf6).
@moq/watch sends TRACK_INFO for catalog before subscribing.
The relay has no copy, so ietf::Subscriber::track_status asks the publisher with TRACK_STATUS.
The publisher refuses with NOT_SUPPORTED, and the query-only request fails ("a refused status fails the query-only or fetch-only request").
The relay logs moq_net::lite::publisher: track info error … track=catalog err=unsupported. The player never sends SUBSCRIBE, so the publisher never sees one.
The same publisher plays through moq-relay 0.17.8, before TRACK_INFO. It also plays through moqx, where the player speaks moq-transport and subscribes without a TRACK_STATUS. Only the moq-lite → moq-transport translation asks for one.
doc/concept/standard.md already says a reader whose publisher refuses TRACK_STATUS as NOT_SUPPORTED still subscribes. Could a TRACK_INFO query do the same: on a NOT_SUPPORTED TRACK_STATUS, learn the track from a SUBSCRIBE instead of failing?
Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.
Create Plan
🧪 Issue enrichment is currently in open beta.
You can configure auto-planning by selecting labels in the issue_enrichment configuration.
To disable automatic issue enrichment, add the following to your .coderabbit.yaml:
issue_enrichment:
auto_enrich:
enabled: false
💬 Have feedback or questions? Drop into our discord!
Same dead end with moqx (the relay, not moqxr) on draft 18 as moq-relay's upstream, with a different refusal: REQUEST_ERROR 0xFFFFFFFF, which is moxygen's internal CANCELLED ("peer closed stream before reply"). moqx sent it after a moq-dev publisher behind it reset the stream: moxygen FINs right after TRACK_STATUS, and the draft-18 publisher takes the FIN as a withdrawal; that part is #7438.
Two corrections to the issue above:
A NOT_SUPPORTED-only fallback would not have caught this. 0xFFFFFFFF maps to Error::Remote, and a_refused_track_status_never_subscribes already pins that both codes fail the query without subscribing.
Closing as won't fix (maintainer decision, 2026-10-31). A moq-lite TRACK_INFO behind a moq-transport publisher still needs that publisher to answer TRACK_STATUS. Falling back to a SUBSCRIBE to learn the track is the kind of guessing rejected on moq-dev/moq#7187, so the fix belongs in the publisher: moqxr should implement TRACK_STATUS. The moqx failure in the follow-up comment was #7438, which is planned in moq-dev/moq#7441. That quest also fixes the stale doc/concept/standard.md sentence you found.
Activity
coderabbitai commentedon Oct 30, 2026
🔗 Related PRs
#7187 - fix(net): subscribe blind when moq-transport discovery cannot prove absence [open]
📝 Issue Planner
Check the box below or use the
@coderabbitai plancommand to generate an implementation plan and prompts that you can use with your favorite coding assistant.🧪 Issue enrichment is currently in open beta.
You can configure auto-planning by selecting labels in the issue_enrichment configuration.
To disable automatic issue enrichment, add the following to your
.coderabbit.yaml:💬 Have feedback or questions? Drop into our discord!
michaelstuurman commentedon Oct 31, 2026
Same dead end with moqx (the relay, not moqxr) on draft 18 as moq-relay's upstream, with a different refusal: REQUEST_ERROR 0xFFFFFFFF, which is moxygen's internal CANCELLED ("peer closed stream before reply"). moqx sent it after a moq-dev publisher behind it reset the stream: moxygen FINs right after TRACK_STATUS, and the draft-18 publisher takes the FIN as a withdrawal; that part is #7438.
Two corrections to the issue above:
Error::Remote, anda_refused_track_status_never_subscribesalready pins that both codes fail the query without subscribing.doc/concept/standard.mdsentence I quoted came in with the NOT_SUPPORTED fallback (fix(net): an IETF TRACK_STATUS refused as NOT_SUPPORTED falls back to SUBSCRIBE #5409) and stayed when fix(net): register subscriptions before exposing track demand #5991 removed it. For query-only demand it is stale, not intended behaviour.(Written with Claude Opus 5.5)
kixelated commentedon Oct 31, 2026
I mean, how much of this is a moqx bug? It's really not hard to implement TRACK_STATUS
kixelated commentedon Oct 31, 2026
Closing as won't fix (maintainer decision, 2026-10-31). A moq-lite
TRACK_INFObehind a moq-transport publisher still needs that publisher to answerTRACK_STATUS. Falling back to a SUBSCRIBE to learn the track is the kind of guessing rejected on moq-dev/moq#7187, so the fix belongs in the publisher: moqxr should implement TRACK_STATUS. The moqx failure in the follow-up comment was #7438, which is planned in moq-dev/moq#7441. That quest also fixes the staledoc/concept/standard.mdsentence you found.(Written by Claude Opus 5.5)