Open
Bug 1582525
Opened 6 years ago
Updated 3 years ago
Avoid racing 6 connections for sites known to have been HTTP2 in the past
Categories
(Core :: Networking, enhancement, P5)
Core
Networking
Tracking
()
NEW
Performance Impact | medium |
People
(Reporter: jesup, Unassigned)
Details
(Keywords: perf:pageload, Whiteboard: [necko-triaged])
Right now, fetching from a site such as y.simg.com which is an HTTP2 server will spin up 6 connections and they race to connect and negotiate TLS, and then when one connects and realizes it's http2, it collapses back to one connection (and any queued requests get processed). This causes some reordering of load results (perhaps), and certainly causes extra load on the CPU and network, and on the server.
If we could cache (along with TLS Token Persistence?) that the server had previously been known to be http2, we could make one connection, and then only open more if we find it's no longer an http2 connection.
Updated•6 years ago
|
Whiteboard: [qf] → [qf:p2:pageload]
Updated•6 years ago
|
Priority: -- → P2
Whiteboard: [qf:p2:pageload] → [qf:p2:pageload][necko-triaged]
Updated•4 years ago
|
Priority: P2 → P5
Updated•4 years ago
|
Severity: normal → N/A
Updated•3 years ago
|
Performance Impact: --- → P2
Keywords: perf:pageload
Whiteboard: [qf:p2:pageload][necko-triaged] → [necko-triaged]
You need to log in
before you can comment on or make changes to this bug.
Description
•