Summary: We are introducing Happy Eyeballs v3, a new
connection-establishment algorithm that replaces the previous
DnsAndConnectSocket-based approach. The new implementation races IPv4 and
IPv6 connection attempts in parallel using a Rust-based state machine
<https://github.com/mozilla/happy-eyeballs>, and can also race QUIC/HTTP/3
against TCP-based HTTP/1.1 or HTTP/2 when an HTTPS record or Alt-Svc
response header indicates that HTTP/3 is available.

Bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1953459

Specification: The algorithm is specified in
draft-ietf-happy-happyeyeballs-v3
<https://datatracker.ietf.org/doc/draft-ietf-happy-happyeyeballs-v3/>.

Standards Body: IETF.

Platform coverage: Available on desktop and Android.

Preference: network.http.happy_eyeballs_enabled

DevTools bug: Not applicable.

Extensions Bug: Not applicable.

Use Counter: N/A

Standards-Positions Discussion:
https://github.com/mozilla/standards-positions/issues/1192

Other browsers:
- Chromium: According to
https://chromeenterprise.google/policies/happy-eyeballs-v3-enabled/ , this
has been supported since Chrome 136.
- WebKit: No signal yet

web-platform-tests: There is no WPT coverage for connection-establishment
behavior. We have added xpcshell tests covering H1, H2, and H3 0-RTT
accepted/rejected paths, dual-family race scenarios, fallback to IPv4, and
LNA denial.

Performance: Connections to dual-stack hosts are faster because IPv6 and
IPv4 are raced simultaneously rather than tried sequentially. We also
expect to see increased HTTP/3 usage, because Happy Eyeballs v3 can race
QUIC/HTTP/3 against TCP-based HTTP/1.1 or HTTP/2 and select the faster
working path.
If you observe unexpectedly long connection times, page not loading, or any
regression in connection behavior, please file a bug blocking Bug 1953459
<https://bugzilla.mozilla.org/show_bug.cgi?id=1953459>.

Thanks,
Kershaw

-- 
You received this message because you are subscribed to the Google Groups 
"[email protected]" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/a/mozilla.org/d/msgid/dev-platform/CAMy_zmP1f0i%3DOXwPVyCVckE1dGZUGgf3XqHMM8TPbic6q9u5fw%40mail.gmail.com.

Reply via email to