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
2 changes: 2 additions & 0 deletions quest/m2/cluster-routing/routes/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -534,6 +534,8 @@ benchmarks.
- [Route session integration](/quest/m2/cluster-routing/routes/session.md) - one router per origin node drives the origin and every lite-07 announce stream
- [Announcement availability](/quest/m2/cluster-routing/routes/availability.md) - announcements report availability, and a let-go is offline rather than END, in Rust and JS
- [js/net route layer](/quest/m2/cluster-routing/routes/js.md) - `js/net` encodes, decodes, and resolves the route layer
- [Browser tabs prove their route key](/quest/m2/cluster-routing/routes/browser-proof.md) - current Firefox and Chromium tabs open lite-07 sessions with a route-key proof the relay verifies, checked nightly
- [Playwright bundles the spec exporters](/quest/m2/cluster-routing/routes/playwright-pin.md) - the toolchain's Playwright bundles Chromium 158+ and Firefox 160+, so the browser proof lane asserts both tabs proven
- [A declined lite-06 route is reconsidered when the lite-07 path fails](/quest/m2/cluster-routing/routes/lite06-failover.md) - a mixed-version relay fails over to the lite-06 route it declined
- [A covering predecessor's takeover shows under a mount](/quest/m2/cluster-routing/routes/mount-takeover.md) - lite-07 presents a takeover beneath a mount as lite-06 does
- [Route layer acceptance](/quest/m2/cluster-routing/routes/acceptance.md) - forwarding fixtures ported and the model deleted, incident fixtures, end-to-end checks, and benchmarks
Expand Down
78 changes: 78 additions & 0 deletions quest/m2/cluster-routing/routes/browser-proof.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,78 @@
# [M] Browser tabs prove their route key to the relay

## Goal

Before the lite-07 cut, a lite-07 session from a current Firefox tab and a
current Chromium tab connects to `moq-relay` with its route-key proof
verified, and a nightly browser test keeps it that way.

A tab proves a key whenever its WebTransport exposes `exportKeyingMaterial`
and the page can generate an Ed25519 key; it sends no proof when either is
missing (`js/net/src/lite/route_key.ts`). The relay derives each WebTransport
session's secret per draft-ietf-webtrans-http3 section 4.8
(`rs/moq-tokio/src/quic/web/session.rs`, `rs/moq-uring/src/quic/web.rs`), and
ends the session with a protocol violation when a proof does not verify
(`rs/moq-net/src/lite/setup.rs`). So a browser whose exporter disagrees with
the relay cannot open a lite-07 session at all. Tests compare the Rust and JS
transcript vectors and both Rust exporter ends, but no real browser exporter
has ever run against the relay
([moq-dev/moq#7937](https://github.com/moq-dev/moq/pull/7937) left this gap).

Browser support as of 2026-11-06: Firefox 155 shipped `exportKeyingMaterial`
on an older draft (context optional, output fixed at 32 bytes), and Firefox 160
(stable 2026-11-10) follows the spec's three arguments
([MDN BCD #30999](https://github.com/mdn/browser-compat-data/pull/30999)).
Chromium enabled it by default in Canary, targeting Chrome 158
([crbug 556304550](https://issues.chromium.org/issues/556304550)), and mixes
the session's CONNECT stream into the derivation.

## Plan

Decided 2026-11-06 (maintainer, /quest-complete while queueing
[moq-dev/moq#8022](https://github.com/moq-dev/moq/pull/8022)):

- ✅ **Cover Firefox and Chromium**, and treat a mismatch as a bug that
blocks the cut. It joins the Routes epic, which gates the cut. Not chosen:
Firefox only until Chrome 158 ships; not planning it.
- ✅ **A nightly Playwright lane.** Firefox and Chromium tabs in the interop
suite (`test/interop`, nightly and the release gate) dial a relay on
Comment thread
kixelated marked this conversation as resolved.
Resolved
lite-07 and assert the relay accepted the tab's proven key. A browser
build without the exporter fails the lane loud rather than passing
unproven. Not chosen: a one-time manual check before the cut.
Comment thread
kixelated marked this conversation as resolved.
Resolved
- ✅ **Bump Playwright; the cut waits for it.** The pinned 1.64.1 bundles
Chromium 156 and Firefox 157, so neither tab proves today.
[Playwright bundles the spec exporters](/quest/m2/cluster-routing/routes/playwright-pin.md)
carries the bumps. Not chosen: installed browser channels; an
informational lane that does not gate the cut.
- ✅ **A browser whose exporter cannot match sends no proof.** If Firefox
155-159's older-draft export differs from the relay's derivation, the JS
client tells it apart at setup and sends no proof there, so the tab
connects unproven. Never send a proof the relay will reject. Not chosen:
requiring Firefox 160+; ignoring 155-159.

Start by running both browsers against a local relay and comparing the
exported bytes with the relay's for the same session. Fix any mismatch at the
root cause, on whichever side departs from the draft. The pinned Firefox 157
is an old-draft build, so the old-draft comparison can run now.

Name the old-draft predicate from that run, not from the exporter's arity:
WebIDL's `length` counts only required arguments, and the old draft's
`context` is optional, so arity does not tell the drafts apart. If the
Comment on lines +58 to +60

@pullfrog pullfrog Bot Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

WebIDL length does separate the drafts: the old draft's optional context leaves one required argument, while the spec's three required parameters leave three. The current wording contradicts the Goal above it and rules out a length < 3 predicate that works.

Suggested change
Name the old-draft predicate from that run, not from the exporter's arity:
WebIDL's `length` counts only required arguments, and the old draft's
`context` is optional, so arity does not tell the drafts apart. If the
Name the old-draft predicate from that run: WebIDL's `length` counts only
required arguments, so the old draft's optional `context` leaves one required
argument and the spec's three required parameters leave three, making
`length < 3` the discriminator. If the

old-draft bytes differ, gate on the Firefox major version, since 155-159 is
a closed set. Record the predicate and the build it was checked against
here.

The relay never acknowledges a proof, and a session that sent none also
stays open, so the lane observes both halves: the page reports whether
`proveRouteKey` returned a proof (`js/net/src/lite/connection.ts`), and the
session staying open past SETUP shows the relay accepted it, since a bad
proof ends it with a protocol violation (`rs/moq-net/src/lite/setup.rs`).

The lane keys its expectation on each tab's reported browser version:
Chromium 158+ and Firefox 160+ must prove, Firefox 155-159 must send none,

@pullfrog pullfrog Bot Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The lane expectation hard-codes "send none" for Firefox 155-159, but the client gate above is conditional on the measured bytes differing. If the bytes match the relay's derivation (the client already passes an empty context and asks for 32 bytes), a 155-159 tab proves and this expectation fails a correct client.

Suggested change
Chromium 158+ and Firefox 160+ must prove, Firefox 155-159 must send none,
Chromium 158+ and Firefox 160+ must prove, Firefox 155-159 must send
none when their old-draft export differs and prove when it matches,

and a build older than its browser's first exporter is "not yet covered".
Any other outcome fails the lane, and the release gate also fails while a
browser is not yet covered, so the cut waits on the
Comment on lines +72 to +75

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[P2] Require both browsers to prove before the release gate passes

The release check here rejects only not yet covered, while Firefox 155-159 sending no proof is an expected passing outcome. At the first planned bump, Playwright 1.65.0 bundles Chromium 158 and Firefox 159 (also confirmed by its browsers.json): Chromium can satisfy proven, Firefox can satisfy sends no proof, and neither is not yet covered. The release lane described here therefore goes green without ever verifying a Firefox proof, contrary to the stated decision that the cut waits for both engines and the Firefox 160 bump. The separate Required quest records the prerequisite but does not make this acceptance condition enforce it.

Keep the old-Firefox fallback case as a valid compatibility test, but make release readiness fail for either not yet covered or sends no proof; require proven for both engines. Add the Chromium-158/Firefox-159 intermediate combination as a gate regression case.

[Playwright bump](/quest/m2/cluster-routing/routes/playwright-pin.md). The lane
prints both browser builds, and they stay current with the toolchain
refresh that bumps the pin.
41 changes: 41 additions & 0 deletions quest/m2/cluster-routing/routes/playwright-pin.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
# [S] Playwright bundles the spec exporters

## Goal

The toolchain's Playwright moves to a release whose bundled Chromium is 158
or later and Firefox 160 or later, the first builds whose
`exportKeyingMaterial` follows draft-ietf-webtrans-http3, so the
[browser proof lane](/quest/m2/cluster-routing/routes/browser-proof.md) can
assert both tabs proven.

The root flake's `pkgs.playwright-driver` supplies `PLAYWRIGHT_BROWSERS_PATH`
and `PLAYWRIGHT_VERSION` (`flake.nix`), and the npm pins
(`test/interop/clients/js/package.json`, `bun.lock`) must match it
(`test/smoke/freshness.sh`). As of 2026-11-06 both are 1.64.1, bundling
Comment on lines +12 to +14

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
and `PLAYWRIGHT_VERSION` (`flake.nix`), and the npm pins
(`test/interop/clients/js/package.json`, `bun.lock`) must match it
(`test/smoke/freshness.sh`). As of 2026-11-06 both are 1.64.1, bundling
The root flake's `pkgs.playwright-driver` supplies `PLAYWRIGHT_BROWSERS_PATH`
and `PLAYWRIGHT_VERSION` (`flake.nix`), and the npm pins
(`test/interop/clients/js/package.json`, `bun.lock`) must match it
(`sh/js/playwright.py`, which `sh/js/workspace.py` runs over every in-tree
manifest; `test/smoke/freshness.sh` guards only the smoke pin). As of 2026-11-06 both are 1.64.1, bundling

Chromium 156 and Firefox 157. npm's latest, 1.65.0, bundles Chromium
158.0.8127.18 and Firefox 159.0; no release bundles Firefox 160 yet, and
nixpkgs still carries 1.64.1.

## Plan

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
Chrome and Firefox channels instead of the bundled builds; leaving the lane
informational so the cut does not wait.

Take each step as nixpkgs carries it: 1.65.0 or later proves the Chromium
tab, and the first release bundling Firefox 160 or later proves the Firefox
tab. With each bump, move the lane's expectation for that browser from "not
yet covered" or "sends no proof" to "proven". Refresh the flake lock and the
npm pins together, as `smoke-toolchain.yml` describes. Overriding
`playwright-driver` in the flake was considered and not chosen: it would
rebuild and pin browser binaries outside nixpkgs.

## Waiting

- nixpkgs: `playwright-driver` at 1.65.0 or later on `nixpkgs-unstable`;
check `pkgs/development/web/playwright/driver.nix` there.
- Playwright: a release whose bundled Firefox is 160 or later; check
`browsers.json` in `npm view playwright-core dist-tags.latest`'s tarball,
then wait for nixpkgs to carry it.
Loading