Meeting 2024-09-03

• Team calendar

  • Next on Necko triage: Kershaw

• New contributors / Friends of Necko THANK YOU!!

  • Andrea Marchesini [:baku]

    • Bug: 1903403 Reject cookies containing %x9 

    • Bug: 1913161 Implement a max-age cap for cookies 

  • Andreas Farre [:farre]

    • Bug: 1912551 Crash in [@ mozilla::net::HttpBaseChannel::ShouldFilterOpaqueResponse] - MOZ_DIAGNOSTIC_ASSERT(ShouldBlockOpaqueResponse()) 

  • Mark Banner (:standard8)

    • Bug: 1914620 Enable remaining ESLint rules on dom/websocket  

  • Eric Chen

    • Bug: 1914995 Remove an unnecessary #include in MicrosoftEntraSSOUtils.mm 

  • Tannal2409

    • Bug: 1915613 Remove test_http3_large_post_telemetry.js since the probes are expired. 

  • Ted Campbell [:tcampbell]

    • Bug: 1915668 Support PUT method in ObliviousHttpChannel 

  • Alexandra Borovova [:Sasha]

    • Bug: 1916105 Perma /fetch/api/basic/request-upload.h2.any.html | Fetch with POST with empty ReadableStream - assert_equals: expected "" but got "[object ReadableStream]" when Gecko 131 merges to beta on 2024-09-02 

• Round Table

  • [valentin] 1894109 - Not caching byte range request responses 

    • Do we want to change our range caching decision logic?

    • Next: Valentin Identify how many HTTPChannels are range requests assess impact and effort

    • May need to do this as part to Compression Dictionary Transport also 

  • [valentin] 1915982 - isLocalIPv4 should consider 0.0.0.0/8 a local address per RFC1122 

    • Yes, lets consider as local address

  • [valentin] 1915848 - Pref off HTTP2 push 

    • It’s become a webcompat risk. I’ll send the intent to unship email soon.

    • Chrome disabled 2yrs ago, Safari TBD

• Active Projects roadmap Fx130 rel Sep 03, Fx131 freeze Aug 29

  • 3rd Party Cookie Deprecation (in progress, target Fx133)

    • No changes. 3PCD will proceed in a phased approach

      • 131: Enable in nightly all windows

      • 132: Enable in PBMode and ETP Strict to release

      • 134: All windows in release (targeting January)

      • 136: “Migration”/deletion of TCP cookies (non-chips partitioned)

  • Fetch keepalive: (in progress, Target Fx132) 

    • – enter updates – 

  • HTTP Priorities Experiments (in progress, target Fx 132)

    • Need to ship and confirm if priorities get reset

  • Firefox/Chrome benchmark test suite (in progress, Fx132)

    • Lars/Max to sync with Kershaw next week 

    • – enter updates – 

    • [kershaw]

      • Bug 1895362 Writing browsertime scripts for testing download and upload

        • WIP

        • Initial result shows no obvious difference between NSPR an quinn-udp

      • Bug 1895363 Reuse our current alerting/perfherder system to track results

        • WIP

  • Interop 2024 URL (in progress, target H2 2024)

    • [Ed] Patch fixup for defaultURI re-enable   

  • necko-glean: (sekim) Landing glean patches using #ifndef ANDROID blocks tracking such blocks using Bug 1901798  

    • [Acreskey] discussion on probes to remove

      • Necko probe migration discussion

    • – enter updates –

  • Fast UDP I/O in Firefox (in progress, target H2 2024)

    • Milestone 1 - end-to-end integration

      • Android x86 seccomp failure on recvmmsg fixed upstream in quinn-udp v0.5.5.

        • Mozilla-central update

      • Fix minor regression introduced in phabricator#D212959, not tracking bytes sent and received on socket for `about:networking#sockets`. See phabricator#D220379.

    •  Milestone 2 - ECN

      • Early blackhole detection on path migration by Lars merged.

        • Thus mitigating the impact of Android API Level 25 failing to send ECN marked datagrams.

        • See https://github.com/quinn-rs/quinn/pull/1975 for details.

      • Pending Neqo v0.9.0 release to bring https://github.com/mozilla/neqo/pull/2072 and https://github.com/mozilla/neqo/pull/2081 into mozilla-central 

      • Thus unblocking ECN pref and metrics https://phabricator.services.mozilla.com/D220062

    • Milestone 5 - multi-packet state machine

      • Proof-of-concept, removing allocation in UDP receive path.

        • https://github.com/mozilla/neqo/pull/2076

        • Has a mere 2% performance improvement on Download benchmark.

        • Proof-of-concept for send path in progress.

    • Glean metrics http3_udp_datagram_size_send and http3_udp_datagram_size_received

      • Ported buffered metrics proposal (mozilla/glean#2945) into phabricator#D216034. Local testing no longer shows significant CPU usage by Glean on 10 GB download.

      • Discussed steps forward with Jan-Erik (Glean team). They are able to reproduce the performance issue seen in phabricator#D216034 and have a local patch fixing it in Glean directly.

  • Performance work: 

    • – enter updates –

    • New experiments:

      • OMT Decompression + html5Parser, cssLoader for Beta

        • Experiment Brief

        • Data ticket

        • [Sunil] Experiment will re run, check with Sunil

          • We had meetings at MozWeek

          • Improving one of the probes

          • Looking for sites that show significant wins/losses with a local patch