Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions quest/m1/interop-coverage/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@ coverage on `main` at `3ab6cbf81d`.

## Required

- [Draft-18 TRACK_STATUS FIN](/quest/m1/interop-coverage/draft18-status-fin.md) - a draft-18 TRACK_STATUS followed at once by a FIN is answered, so moqx-forwarded status requests reach moq-dev publishers
- [JS namespace lifecycle](/quest/m1/interop-coverage/js-namespace-done.md) - the runner's moq-dev-js client moves to an `@moq/net` that withdraws namespaces by reset, so subscribe-namespace-lifecycle passes on Cloudflare draft-18 and aiomoqt
- [JS subgroup data through imquic](/quest/m1/interop-coverage/js-imquic-objects.md) - the runner's moq-dev-js client moves to an `@moq/net` patch that writes data stream headers in one write, so data-subgroup-basic passes against imquic
- [JS subscribers FETCH](/quest/m1/interop-coverage/js-ietf-fetch.md) - `@moq/net` fetches past groups over moq-transport, and joins the CLI FETCH round
Expand Down
33 changes: 33 additions & 0 deletions quest/m1/interop-coverage/draft18-status-fin.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,33 @@
# [S] A draft-18 TRACK_STATUS followed at once by a FIN is answered

## Goal

On draft 18, a moq-net publisher (Rust and `@moq/net`) answers a TRACK_STATUS whose requester FINs right after sending it, with TRACK_STATUS_OK or REQUEST_ERROR, instead of resetting the stream with CANCELLED. A reset or STOP_SENDING still withdraws it. Every other draft-18 request keeps treating a requester FIN as a cancel. A unit test sends TRACK_STATUS plus an immediate FIN on drafts 18, 20 and 22 and gets an answer on each. `doc/concept/standard.md` no longer says a reader whose publisher refuses `TRACK_STATUS` still subscribes, where that reader only queries or fetches.

## Plan

Decided 2026-10-31 (maintainer, `/quest-import` of [moq-dev/moq#7438](https://github.com/moq-dev/moq/issues/7438) and [moq-dev/moq#7423](https://github.com/moq-dev/moq/issues/7423)).

moxygen FINs right after TRACK_STATUS on draft 18+, since nothing can follow one. The WG agrees this is normal (moq-wg/moq-transport#1634), and draft 19 §3.3.2 says a FIN is not a cancellation. On draft 18, `rs/moq-net/src/ietf/request_stream.rs:11` `fin_cancels` treats any requester FIN as a withdrawal. So a TRACK_STATUS forwarded by a moxygen-based relay such as moqx is reset, and a moq-lite viewer behind moq-relay never subscribes.

- Scope of the FIN rule:
- ✅ TRACK_STATUS only: a FIN only ends a TRACK_STATUS requester's input, on every draft. Our own draft-18 requesters keep cancelling other requests by FIN.
- FIN never cancels on draft 18 (except SUBSCRIBE_NAMESPACE/SUBSCRIBE_TRACKS, §6.1)
- Won't fix
- The fix is in [moq-dev/moq#7439](https://github.com/moq-dev/moq/pull/7439), an external contributor's draft that allows maintainer edits, with exactly this scope. Rust passes `fin_cancels: bool` through `poll_cancel`/`poll_withdrawn`/`until_withdrawn`/`resolve`, and `serve_track_status_stream` passes `false`. JS `RequestStream` takes `{ finCancels }`. Review it, mark it ready, and land it. Push the doc fix below onto it, or land it as a follow-up commit.
- #7423 asked that a TRACK_INFO behind a publisher refusing TRACK_STATUS fall back to subscribing:
- ✅ Won't fix + doc fix: publishers must answer TRACK_STATUS. moqxr should implement it, which the maintainer called "really not hard". This matches the 2026-10-29 rejection of blind subscribing on [moq-dev/moq#7187](https://github.com/moq-dev/moq/pull/7187).
- Fallback subscribe
- Keep open, defer
- Doc: `doc/concept/standard.md:31` ends with "a reader whose publisher refuses `TRACK_STATUS` as `NOT_SUPPORTED` still subscribes". That is left over from the fallback that [moq-dev/moq#5991](https://github.com/moq-dev/moq/pull/5991) removed. Reword it so only a subscribing reader still subscribes, while a query-only or fetch-only reader fails, as `a_refused_track_status_never_subscribes` (`rs/moq-net/src/ietf/subscriber.rs`) pins.
- Not here: moqx's GROUP_ORDER parameter in TRACK_STATUS_OK, which draft 18 refuses. That is upstream moxygen's fix, facebookexperimental/moxygen#264, tracked by [Moxygen pin bump](/quest/m1/moxygen-pin-bump.md).

Public API: none. Wire: a draft-18 TRACK_STATUS is no longer withdrawn by a FIN alone. `release` refuses TRACK_STATUS, so nothing needs backporting.

## Closes

- [#7438](https://github.com/moq-dev/moq/issues/7438) - draft-18 TRACK_STATUS followed by a FIN is reset instead of answered

## Related

- [Moxygen pin bump](/quest/m1/moxygen-pin-bump.md) - the other half of playing through moqx on draft 18