Meeting 2025-04-15

• Team calendar

  • Next on Necko triage: Sunil

• New contributors / Friends of Necko (5)– Thank you!

  • Jens Stutte [:jstutte]

    •  Bug: 1959367 Crash in [@ mozilla::net::nsStandardURL::nsStandardURL]

  • Dan Ballard (Tor Browser Dev) 

    • Bug: 1959305 When Parental Controls Service is disabled, DNS Over HTTPS settings fail to work with uncaught exception

  • Andrea Marchesini [:baku]

    • Bug: 1958422 cookieStore.delete() is unable to delete cookies that were set using cookieStore.set(), when invoked on a page whose URL includes a subdomain

  • Ryan VanderMeulen [:RyanVM]

    •  Bug: 1958041 Perma MinGW ld.lld: error: undefined symbol: HUF_decompress4X2_usingDTable_internal_fast_asm_loop

  • Alexander Kuleshov

    • Bug: 1959970 Merge code for unix and win in the NetAddrToPRNetAddr 

• Round Table

  • [acreskey] FYI, I’m now a Nimbus L3 reviewer; can approve start/stop experiments

• Active Projects roadmap Fx138 rel Apr 29, Fx139 code freeze April 24

  • Firefox Essential Internal Services Toughening[FEISTY] (in progress, TBD)

    • – enter updates –

  • DoH Enablement - Fenix and Focus (in progress, H2-2025)

    • (sekim) Informed the QA team that we are putting DoH on Focus on hold

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

    • CHIPS rolled out to beta users (100%) via Nimbus

    • 3PCD still will be pending chrome implementation

  • Firefox/Chrome benchmark test suite (Done, Apr 11)

    • Project Wrap-Up: Network Benchmark Framework

    • Plan to schedule a follow up meeting April 28 - May 02.

    • Neqo v0.13 improves the goodput for 300M scenarios.

      • 300M_40ms: 140 -> 149

      • 300M_80ms: 65 -> 100

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

    • Acreskey - can we remove this item?

  • Fast UDP I/O in Firefox (in progress, target Fx137)

    • Rollout on Beta

      • Still ongoing

    • 1 - 2 % rollout on release is next

  • Http3 upload speed

    • PR_POLL_WRITE patch merged.

    • 0RTT pull request merged

    • Neqo v0.13 landed in mozilla-central, i.e. Firefox 139.

  • QUIC stream receive window auto-tuning

    • Implementation merged

    • Neqo v0.13 landed in mozilla-central, i.e. Firefox 139.

  • Implement WebTransport based on HTTP/2 (in progress, Fx139)

    • (sekim) SendDatagram (and receive datagram) patch has been accepted, should be landed once we land its parent patches

    • [kershaw]

      • Patches for implementing Http2WebTransportStream are all in review

  • Interop 2025 - URL Pattern (internal review, Fx141)

    • – enter updates –

  • Support Compression Dictionary Transport (in progress, Fx141)

    • – enter updates –

  • Improve Fenix Performance in Slow and Transitional Networks (in progress, TBD)

    • Comparing Fenix to Chrome navigation over slow/fast networks 

  • Local Network Access (planning, TBD)

    • [Sunil] Project plan internal review complete

    • Meta Bug -  local-network-access

  • Performance work: 

    • DNS Cache Expiry:

      • Grace period experiment live Start: Apr 02, Enrollment end: Apr 09, End Apr 

30. 

• – enter updates –

• [acreskey] FYI, DNS perf issue visible on dashboards

  • Query

    • Reached out to Perftest to have this probe is monitored in initial histogram diff watch list

    • glam

  • Bug 1960020 -  Fenix top-level document dns lookup times regressed around April 6, 2025

  • [kershaw]
    Should we back out bug 1958018 and see if that resolves the regression?

If it does, we can revisit bug 1916418, which originally introduced the two additional DNS requests. Bug 1916418 was intended to check whether the CNAME matches the target name of the HTTPS record. Based on this query, only 0.2% of HTTPS records have such a mismatch.
As an alternative, we could consider checking the CNAME only when DoH is enabled, since the performance regression seems to be caused by the native DNS API.