Skip to content

ietf: a TRACK_STATUS refused with any code but NOT_SUPPORTED fails the reader's subscription #5954

Closed

Description

@t0ms

Problem

moq export ts (current main, ad87ad3) against a moqx relay on draft 18 sometimes ends at once with 0xFFFFFFFF; reruns pass. At trace level it happened in two of three runs: the catalog read races its subscription, so the origin asks with TRACK_STATUS, and the relay answers REQUEST_ERROR with code 0xFFFFFFFF, which no draft defines. The subscriber falls back to SUBSCRIBE only on NOT_SUPPORTED (the Err(Error::Unsupported) arm after read_track_status_response in rs/moq-net/src/ietf/subscriber.rs). Any other refusal rejects the request, here the export's catalog subscription, and the export exits.

The undefined code is the relay's bug, and it's going to its operator. The question for moq-dev is whether a refused TRACK_STATUS should decide a subscription that SUBSCRIBE could still serve. As the comment on that arm notes, the query is often made on a reader's behalf before its subscription lands.

Proposal

Fall back to SUBSCRIBE on any TRACK_STATUS refusal except one that answers for the track itself (it doesn't exist, or the subscriber may not have it), so SUBSCRIBE gives the authoritative answer. A narrower option: fall back on INTERNAL_ERROR and on codes outside the draft's registry. Either way, a test would have a publisher refuse TRACK_STATUS with an undefined code and check that the subscription still opens.

Would you take either? I can send the pull request.

(Written by Claude Opus 5.5)

Activity

  1. added
    questBeing tracked/planned in a quest. See `quest/`
    on Oct 19, 2026
  2. kixelated commented on Oct 19, 2026

    @kixelated
    Collaborator

    Decision: the undefined code is the relay's bug, but moq-net walked into it through its own race: a subscribing reader's request becomes visible before its subscription, so it is sent as TRACK_STATUS. The fix closes that race at the source and deletes the NOT_SUPPORTED fallback (no in-tree publisher refuses TRACK_STATUS any more). The fallback is not widened. Upstream issues get filed for relays that refuse TRACK_STATUS.

    Tracked by quest/m1/track-status-race.md.

    Planned in #5969. Anyone who wants to donate tokens can run /quest-start quest/m1/track-status-race to pick it up.

    (Written by Claude Opus 5.5)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    questBeing tracked/planned in a quest. See `quest/`

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions