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
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,17 @@ nixpkgs still carries 1.64.1.

## Plan

Moved from m2 to m1 2026-11-14 (the m2 triage before the cut, under
/quest-afk, for the maintainer's review): ✅ m1, since a toolchain bump breaks
nothing and builds on no m2 work (rejected: m2, where the cut waited on
nixpkgs; m4 or m5, which hold it behind the cut or the next wire cut once
nixpkgs lands). [The release cut](/quest/m3/release/cut.md) requires it in
place of lite07-finalize, since the cut PR's interop lane runs with
`CI_STRICT` and fails the route key lane on a Chromium below 158;
[lite07-finalize](/quest/m2/lite07-finalize.md) runs the lane once with
npm's Playwright 1.65.0 before the rename as an early signal. This amends
the 2026-11-06 decision below that the lite-07 cut waits for the bump.

Decided 2026-11-06 (maintainer, /quest-complete, on Pullfrog's review of
[moq-dev/moq#8037](https://github.com/moq-dev/moq/pull/8037)): ✅ bump
Playwright, and the lite-07 cut waits for it. Not chosen: running installed
Expand Down
35 changes: 35 additions & 0 deletions quest/m2/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -217,3 +217,38 @@ rule here). Its lag rule reshapes `main`-only API the cut would publish (the
`stretch` setting, moq-audio's `Sink::due` and `bus::Positioned`), and the
cut should not ship the 2026-11-09 stretch model, which lets audio trail
video by up to 2 s.

Decided 2026-11-14 (m2 triage before the cut, under /quest-afk; the
maintainer asked that what remains be unblocked or punted, and picked these on
[moq-dev/moq#8768](https://github.com/moq-dev/moq/pull/8768)). m2 keeps only
what would need a breaking release had it landed right after the cut:

- ✅ Left m2: [the request loop](/quest/m3/cluster-routing/request-loop.md)
to m3's cluster-routing epic (a fixture and a fix, no wire);
[media consumers](/quest/m3/time-addressing-consumers.md) to m3 (its API
breaks ride the model; dropping hang's markers is tolerated by readers);
[moq-transport time addressing](/quest/m4/ietf-time-addressing.md) to m4
(low-priority compatibility, no named consumer); and the
[Playwright bump](/quest/m1/playwright-pin.md) to m1, which the release
cut requires in place of lite07-finalize (its strict interop lane needs
Chromium 158), with lite07-finalize checking Chromium 158 once by hand. The E2EE nonce
quest was removed: the nonce already binds the wire ID, which becomes the
start, and lite-07 by start rewords the draft. Rejected for each: staying
in m2.
- ✅ Stayed in m2 though not breaking itself:
[the interest set](/quest/m2/cluster-routes/interest-set.md), because
[the route stream](/quest/m2/cluster-routes/route-stream.md) needs it.
The route stream stays because lite-07 would otherwise freeze with one
router key per announce stream, which the route model doesn't check, and
keep the reopen-and-hand-over machinery for its lifetime (rejected: moving
both to lite-08).
- ✅ [Route acceptance](/quest/m2/cluster-routes/acceptance.md) stays in
the routes epic and gates the cut, per the 2026-10-13 rule that routes'
tier and partition tests run before it (maintainer; rejected: moving it to
m3 and fixing late routing flaws in lite-08).
- ✅ [Group drops](/quest/m2/group-drops.md) split: its lite-07 wire question
is [its own quest](/quest/m2/group-drops-wire.md), which starts now and
gates the freeze; the public `Ordered` API waits for the model.
[Catalog at timestamp](/quest/m2/catalog-at-timestamp.md) no longer waits
for `@moq/net` by start, and [the bindings](/quest/m2/time-addressing/bindings.md)
shrink to a rename, since the model already fills each field with the start.
11 changes: 10 additions & 1 deletion quest/m2/catalog-at-timestamp.md
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,16 @@ until media reaches that time; one that reads it late applies it at once.
A scale-less broadcast's catalog has no media time to wait for, so it applies
on arrival, as today.

Decided 2026-11-14 (the m2 triage before the cut, under /quest-afk, for the
maintainer's review): ✅ this waits only for the Rust model, not for
[@moq/net by start](/quest/m2/time-addressing/js.md) (rejected: keeping
that wait). A catalog's stamp is its frame's timestamp, which `@moq/net`
frames already carry (moq-json's `Timed.at`), and nothing here reads a group
start. The JS half checks that `js/hang`'s catalog producer stamps from the
broadcast clock, and makes it do so if not. It stays in m2: `rs/hang` and
`js/hang` return the stamp with each catalog, a breaking change to both
consumers, and the hang draft's rule is what lite-07 catalogs mean.

## Required

- [Rust groups are addressed by their start](/quest/m2/time-addressing/model.md) - the catalog producer stamps from the broadcast clock
- [@moq/net addresses groups by start](/quest/m2/time-addressing/js.md) - the same for js/hang
7 changes: 7 additions & 0 deletions quest/m2/cluster-routes/acceptance.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,13 @@ replace are retired.

## Plan

Decided 2026-11-14 (maintainer, /quest-complete on
[moq-dev/moq#8768](https://github.com/moq-dev/moq/pull/8768)): ✅ stays in
m2 and gates the cut, per lite07-finalize's 2026-10-13 rule that routes'
tier and partition tests run before the cut (rejected: moving it to m3 and
fixing late routing flaws in lite-08). A flaw found here is fixed on
lite-07-wip before it freezes.

Decided 2026-11-12 (maintainer, /quest-plan on the quests still written for
Babel): acceptance is of the path rule
([epic](/quest/m2/cluster-routes/README.md) Plan item 27), on the
Expand Down
2 changes: 1 addition & 1 deletion quest/m2/cluster-routes/flapping-publisher.md
Original file line number Diff line number Diff line change
Expand Up @@ -64,4 +64,4 @@ Work:

## Related

- [A request loop heals once routes settle](/quest/m2/cluster-routes/request-loop.md) - another failure mode from the same BGP comparison
- [A request loop heals once routes settle](/quest/m3/cluster-routing/request-loop.md) - another failure mode from the same BGP comparison
7 changes: 7 additions & 0 deletions quest/m2/cluster-routes/interest-set.md
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,13 @@ carries it.

## Plan

Decided 2026-11-14 (maintainer, /quest-complete on
[moq-dev/moq#8768](https://github.com/moq-dev/moq/pull/8768), the m2 triage
before the cut): ✅ stays in m2 (rejected: moving the route stream and the
interest set to lite-08). It changes no wire or API on its own, but
[the route stream](/quest/m2/cluster-routes/route-stream.md) needs it, and
no quest requires one in a later milestone.

Decided 2026-11-10 (maintainer, /quest-plan on lite-07 routing). The model
goes first because the set's edit rule is the one new rule in this redesign,
and every earlier routing bug was an ordering nobody had enumerated.
Expand Down
7 changes: 7 additions & 0 deletions quest/m2/cluster-routes/route-stream.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,13 @@ and lite-06 and older and moq-transport are untouched.

## Plan

Decided 2026-11-14 (maintainer, /quest-complete on
[moq-dev/moq#8768](https://github.com/moq-dev/moq/pull/8768), the m2 triage
before the cut): ✅ stays in m2, gating lite-07's freeze (rejected: moving the route stream and the
interest set to lite-08). Without it lite-07 would freeze with one router key per announce
stream, which the route model doesn't check, and keep the reopen-and-hand-over
machinery for its lifetime.

Decided 2026-11-10 (maintainer, /quest-plan on lite-07 routing).

Why: today a side's requests ride the streams it opened and its answers ride
Expand Down
64 changes: 64 additions & 0 deletions quest/m2/group-drops-wire.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
# [M] A moq-lite subscriber always learns which group it abandoned

## Goal

No group a moq-lite subscriber gives up on vanishes because its stream was
stopped or evicted before its header was read: the subscriber always learns
the ID of a group it abandoned. moq-transport reports what its wire allows,
and the PR records any gap there. Whether this needs a lite-07 wire change is
settled before lite-07 freezes.

## Plan

Split 2026-11-14 from [the public drop signal](/quest/m2/group-drops.md) (the
m2 triage before the cut, under /quest-afk, for the maintainer's review):
✅ the wire half reads no group key, so it starts now, and
[lite-07's freeze](/quest/m2/lite07-finalize.md) waits on it instead of on
the public API (rejected: keeping one quest, which holds this wire question
until the Rust model lands).

Planned 2026-11-04 with the maintainer, from
[moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838). Under that
benchmark's overload policy, overcommitted cells hang. Sometimes the subscriber
drops a group without `recv_group` ever returning it: either the group's stream
was evicted before its header was read, or the group was evicted and so was the
FETCH that refetched it. The 4 MiB-frame cell hits this on every run; the
256-stream cell sometimes does.

✅ **The wire mechanism is this quest's call, with numbers** (maintainer,
2026-11-04). Compare:

- **The publisher answers STOP_SENDING with SUBSCRIBE_DROP.** Lite already
sends SUBSCRIBE_DROP on the control stream when a group fails at the
publisher (`GroupServe::settle_failed`, `rs/moq-net/src/lite/publisher.rs`,
versions with a track stream), but a subscriber's STOP_SENDING doesn't
produce one.
- **The subscriber reads a group's header before it stops or evicts that
stream.**

Doing both is fine if each is needed. Start with the maintainer's question:
why does the subscriber stop or evict a stream before reading its header at
all? If receive-window eviction can wait for the header, take the no-wire
path and remove this quest from
[lite07-finalize](/quest/m2/lite07-finalize.md)'s Required.

Take the numbers from the receive-windows benchmark's overcommitted cells (256
streams, 4 MiB frames, follow; `rs/moq-tokio/tests/integration/receive_credit.rs`):
each must account for every group sent as delivered or abandoned, counted
inside moq-net until [the public drop signal](/quest/m2/group-drops.md) lets
the benchmark count them itself. Compare their throughput and the
control-stream bytes each mechanism adds. Cover moq-transport too, or say why
it differs. A wire rule applies to `js/net`'s publisher and subscriber in the
same PR.

The drop set owns the per-cursor drop runs and their reasons
([moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476)); this changes
only whether an abandoned group is known, not how a cursor judges a missing
one.

Public API: none. Wire: possibly lite-07-wip (SUBSCRIBE_DROP for a stopped
group), decided here.

## Related

- [The public drop signal](/quest/m2/group-drops.md) - surfaces what this makes known to the application
89 changes: 28 additions & 61 deletions quest/m2/group-drops.md
Original file line number Diff line number Diff line change
@@ -1,81 +1,46 @@
# [M] A group a subscriber gives up on is always reported
# [S] A group a subscriber gives up on is always reported

## Goal

Every group a moq-lite subscriber gives up on surfaces to the application: no
group vanishes because its stream was stopped or evicted before its header was
read. On the wire, the subscriber always learns the sequence of a group it
abandoned; moq-transport reports what its wire allows, and the PR records any
gap there.
Every group a moq-lite subscriber gives up on surfaces to the application.
In the API, `track::Subscriber`'s drop signal (today the crate-private
`poll_drop`, which the ordered reader and resume use to skip gaps) is public,
so the receive-windows benchmark and applications can count losses instead of
waiting forever.
waiting forever, and that benchmark's overcommitted cells finish with
`delivered_frames` plus the reported drops accounting for every group sent.

## Plan

Planned 2026-11-04 with the maintainer, from
[moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838). Under that
benchmark's overload policy, overcommitted cells hang. Sometimes the subscriber
drops a group without `recv_group` ever returning it: either the group's stream
was evicted before its header was read, or the group was evicted and so was the
FETCH that refetched it. No public API reports that group, so a reader that
paces on delivered-or-evicted groups stalls. The 4 MiB-frame cell hits this on
every run; the 256-stream cell sometimes does.
[moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838): no public API
reports a group the subscriber dropped without `recv_group` returning it, so a
reader that paces on delivered-or-evicted groups stalls. Split 2026-11-14 (the
m2 triage before the cut, under /quest-afk, for the maintainer's review):
[the wire half](/quest/m2/group-drops-wire.md) makes every abandoned group
known and decides lite-07's part; this publishes it. It stays in m2 because
the `Ordered` change below is breaking.

- ✅ **Milestone m1** (maintainer, 2026-11-04): it changes public API and
possibly lite-07 wire, so it lands before the cut.
- ✅ **The wire mechanism is this quest's call, with numbers** (maintainer,
2026-11-04). Compare:
- **The publisher answers STOP_SENDING with SUBSCRIBE_DROP.** Lite already
sends SUBSCRIBE_DROP on the control stream when a group fails at the
publisher (`GroupServe::settle_failed`, `rs/moq-net/src/lite/publisher.rs`,
versions with a track stream), but a subscriber's STOP_SENDING doesn't
produce one.
- **The subscriber reads a group's header before it stops or evicts that
stream.**

Doing both is fine if each is needed. The numbers come from the
receive-windows benchmark's overcommitted cells (256 streams, 4 MiB frames,
follow; `rs/moq-tokio/tests/integration/receive_credit.rs`), which
[moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838) skips until
this lands: each must finish, with `delivered_frames` plus the reported
drops accounting for every group sent. Compare their throughput and the
control-stream bytes each mechanism adds. Start with the maintainer's question:
why does the subscriber stop or evict a stream before reading its header at
all? If receive-window eviction can wait for the header, the wire half may
shrink to that. Cover moq-transport too, or say why it differs.
- **Public drop signal.** Make the drop signal public in Rust. Choose its
shape here: a separate read named like the existing ones (for example
`recv_drop` returning the given-up run and its error), or a drop variant on
`recv_group`, the way container consumers already read `Frame | Gap | End`.
Prefer folding it into an existing read if that reads as naturally. The
ordered reader and resume keep using it internally. Its first consumer is
the receive-windows benchmark. Mirror it in `@moq/net` and the bindings only
for a named consumer (for example a player or container reader that skips
gaps), per the API audit rule that new
surface needs one; name it in the PR or leave the mirror out.
- **After the drop set.** This waited (2026-11-05) for Drop set, which
rewrote the lite-07 SUBSCRIBE_DROP wording and the drop signal, and
finished in [moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476) on
2026-11-12. `poll_drop` now takes no `Delivered` argument and yields the
given-up run with its `Hole` (`rs/moq-net/src/model/track.rs`,
`track/hole.rs`). Publish the API against that record.
- The drop set owns the per-cursor drop runs and their reasons. This quest
reports what a cursor gives up; it doesn't change how a cursor judges a
missing group.
- ✅ **Milestone** (maintainer, 2026-11-04): it changes public API, so it
lands before the cut.
- **After the drop set.** Drop set finished in
[moq-dev/moq#8476](https://github.com/moq-dev/moq/pull/8476) on 2026-11-12.
`poll_drop` now takes no `Delivered` argument and yields the given-up run
with its `Hole` (`rs/moq-net/src/model/track.rs`, `track/hole.rs`). Publish
the API against that record. The drop set owns the per-cursor drop runs and
their reasons; this reports what a cursor gives up.
- Mirror it in `@moq/net` and the bindings only for a named consumer (for
example a player or container reader that skips gaps), per the API audit
rule that new surface needs one; name it in the PR or leave the mirror out.

Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and remaining quests): ✅ key the drop signal by group start and land after
Decided 2026-11-07 (maintainer, /quest-plan from the API sweep of open PRs and
remaining quests): ✅ key the drop signal by group start and land after
[time addressing's model](/quest/m2/time-addressing/model.md), so its public
API never ships sequence-keyed (rejected: landing first and rekeying with
the model). ✅ Drops surface as a `Gap { range, error }` item on
`track::Ordered`, which already skips gaps through `poll_drop`, mirroring
the container consumer's `Read::Gap` (`rs/moq-mux/src/container/consumer.rs`);
`recv_group` is unchanged (rejected: a separate `recv_drop` read; a drop
variant on `recv_group`). Answer the maintainer's question above first: if
receive-window eviction can wait for the header, take the no-wire path, and
remove this quest from [lite07-finalize](/quest/m2/lite07-finalize.md)'s
Required.
variant on `recv_group`).

Open, for whoever starts this, with the maintainer: a `Gap`'s start-keyed
`range` doesn't say how many groups it covers once starts are sparse, yet
Expand All @@ -91,11 +56,13 @@ Callers match the new item, and a caller that ignores drops takes the group
arm; the PR's MIGRATE notes say so. `recv_group` and the unordered cursor are
unchanged, so a reader that needs drop accounting reads through `Ordered`.
`@moq/net` and the bindings mirror it only with a named consumer.
Wire: possibly lite-07 (SUBSCRIBE_DROP for a stopped group), decided here.
Wire: none; [the wire half](/quest/m2/group-drops-wire.md) owns it.

## Required

- [Address groups by timestamp](/quest/m2/time-addressing/model.md) - the drop range is keyed by start
- [A moq-lite subscriber always learns which group it abandoned](/quest/m2/group-drops-wire.md) - the benchmark's cells finish only once no abandoned group is lost

## Related

- [The receive-windows benchmark](/quest/m3/receive-windows-evicted.md) - its overcommitted cells wait on this (skipped by [moq-dev/moq#7838](https://github.com/moq-dev/moq/pull/7838) until it lands)
Loading