Meeting 2024-06-25

• Team calendar

  • Next on Necko triage: Kershaw

• New contributors / Friends of Necko THANK YOU!!

  • amarchesini@mozilla.com

    • Bug: 1903110 cookie with empty name and value should be rejected 

    • Bug: 1768656 Deprecate and remove thirdparty.*sessionOnly prefs 

    • Bug: 1903415 Long cookie attrbiutes should be rejected 

  • ckerschbaumer@mozilla.com

    • Bug: 1883954 Consider moving HTTPS upgrade telemetry code into one place 

  • jdescottes@mozilla.com

    • Bug: 1903212 Expose a nsIDataChannel interface to JS 

  • mail@max-inden.de

    • Bug: 1903621 allow getsock* syscall in Rust binary 

  • twisniewski@mozilla.com 

    • Bug: 1796785 Addition of response status code to Resource Timing API 

    • Bug: 1842970 Pass failing xhr WPTs related to aborting XMLHttpRequests on workers

    •  Bug: 697151 Synchronous XMLHttpRequest (XHR) does not block readyState events for async XHR 

• Round Table

  • – no public topics this week  –

• Active Projects roadmap Fx128 rel July 09, Fx 129 soft freeze July 04 

  • Native HTTPS RR Implementation (in progress, target Fx129)

    • [kershaw] another potential blocker:

      • Bug 1895741 Intermittent valgrind-test | 28 bytes in 1 blocks are definitely lost at malloc / __res_context_send / __res_context_query / __res_context_querydomain

        • Need review

  • 3rd Party Cookie Deprecation (in progress, target Fx132)

    • [Ed] CHIPS partition limit: Reviews are coming back, making necessary fixes

    • Working to address review comments, may move to 133 TBD

  • Fetch keepalive: (in progress, Target Fx130) 

    • (greg) Reminder to ping Jason Kratzer once ms 1 has landed and we have added the fuzzing pref

      • [Sunil] Will do

    • [Sunil] Bug 1871378 - Incorporating review comments. MS1 can be marked done after merging this. We pass all WPTs now 

    • [Sunil] Will start work on MS2 next week. 

  • HTTP Priorities Experiments (in progress, target Fx130 )

    • [acreskey]

      • Bringing to DS Wednesday, Beta 128 (start as soon as approved)

      • https://mozilla-hub.atlassian.net/wiki/spaces/FIREFOX/pages/11047744/Experiments+Necko

      • Will run broad local tests

      • priorityHeader currently riding the trains; probably want to stop it

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

    • (greg) Can we use this environment to verify Fast UDP I/O(GSO)?

    • – enter updates –

    • [kershaw]

      • Test machines ready

      • Start working

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

    • (sekim) Landed %2e entered in URL bar not normalized leading to denormalized request, this bug may also fix Bug 1902331.

    • [Ed] Added some tests for adding remote settings defaultURI bypass list, made some fixes as well. 

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

    • [acreskey] Continue to hold on Android until further notice

    • (sekim) Dashboard for tracking networking_http_channel_disposition and networking_http_channel_onstart_status

    • (sekim) Landed Migrate HTTP_RESPONSE_STATUS_CODE probe to glean

    • (sekim) Added HTTP_RESPONSE_STATUS_CODE to the above dashboard.

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

    • Congrats Max Inden for getting first project approved!

    • Allow getsock* syscalls in Rust code unblocking milestone 1.

    • Do `cargo vet` `safe-to-deploy` _audit_ for `quinn-udp`. Needed to import `quinn-udp` into `neqo_glue`'s dependency tree.

    • Propose patch to remove `quinn-udp`'s dependency on `tracing-attributes`. One less dependency to audit for milestone 1.

    • Debug false positive CI check, requiring `safe-to-deploy` audit for `tokio` crate. See Phabricator comment.

    • Tested https://github.com/cloudflare/quiche/pull/1764 as an alternative to `quinn-udp`. We will proceed with `quinn-udp` given that it supports all the Firefox tier 1 platforms.

  • Performance work: 

    • [acreskey] Bug 1889771 - Increase network.http.max-connections on Android nightly (currently 40 (900 on desktop))

      • Via Performance Pulse dashboards, saw improvements of 5-20%, P50 onwards in probes where you could expect changes:

        • http channel page open to first_sent

        • http channel sub item open to first sent

        • transaction wait time

      • Backed out to double-check impact.

      • Proposal:

        • Land os-limits telemetry

        • Increase in nightly, 128 sockets (beta & earlier)

        • Uplift to beta to observe impact of single change

        • Consider Fenix release population experiment

    • New experiments:

      • Performance Calibration (A/A/A) on Fenix, live (release channel might be affected, some probes)

        • [acreskey] Completed and summarized:

          • Fenix Performance Calibration results

          • Top-level metrics show less variance than desktop, despite much smaller population sizes

      • OMT Decompression + html5Parser, cssLoader 

        • Experiment Brief

        • Data ticket