You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@DerafshAtur — v1.9.15 (CI building now) ships HTTP/2 multiplexing on the relay leg. This is the architectural fix for the slowness you and others reported. The TLS pool tuning that PR #751 added in v1.9.11 becomes much less load-bearing under h2 — one connection multiplexes ~100 concurrent streams, no head-of-line blocking, no pool refill loop work to do.
Once you upgrade to v1.9.15:
ALPN negotiates h2 against the Google edge automatically — no config change needed
All relay traffic (Apps Script, exit-node, full-mode tunnel) goes over one connection
Falls back to the existing HTTP/1.1 keep-alive path automatically if h2 isn't viable in your network
If for any reason h2 misbehaves on your specific network / deployment, set "force_http1": true in config.json to revert to exactly the pre-PR behavior.
Closing as fixed. Reopen if v1.9.15 doesn't deliver — paste the config (auth_key masked), Recent log, and which mode you're on (apps_script / direct / full).
Thanks for the persistent push on this; "feels slower" is hard to action without users confirming, and you and the other commenter on the thread were the signal that pushed h2 up the priority queue.
[reply via Anthropic Claude | reviewed by @therealaleph]
Activity
mrezaee1994 commentedon May 6, 2026
yes it seems 1.9.14 is slower than 1.9.10 for some strange reasonsit seems culprit is block doh was off
DerafshAtur commentedon May 6, 2026
i did this setting too before someone try to off/on it the main issue is its still slower
chore: cut v1.9.15 — h2 multiplexing + block QUIC + UI a11y + GitHub …
therealaleph commentedon May 6, 2026
@DerafshAtur — v1.9.15 (CI building now) ships HTTP/2 multiplexing on the relay leg. This is the architectural fix for the slowness you and others reported. The TLS pool tuning that PR #751 added in v1.9.11 becomes much less load-bearing under h2 — one connection multiplexes ~100 concurrent streams, no head-of-line blocking, no pool refill loop work to do.
Once you upgrade to v1.9.15:
h2against the Google edge automatically — no config change neededIf for any reason h2 misbehaves on your specific network / deployment, set
"force_http1": trueinconfig.jsonto revert to exactly the pre-PR behavior.Closing as fixed. Reopen if v1.9.15 doesn't deliver — paste the config (auth_key masked), Recent log, and which mode you're on (apps_script / direct / full).
Thanks for the persistent push on this; "feels slower" is hard to action without users confirming, and you and the other commenter on the thread were the signal that pushed h2 up the priority queue.
[reply via Anthropic Claude | reviewed by @therealaleph]