[quicwg/base-drafts] HTTP/3 priorities are too complex for a majority of implementers to implement (#2739)

[ianswett <[e-mail]>]

A number of implementers have indicated a lack of interest(ie: I'll do something else for cases when I control both client and server) or complete unwillingness to implement existing HTTP/3 priorities.

I believe a core goal of HTTP/3 should be that the majority of implementations will support the specified priority scheme.  If that's not true, we should move priorities to an extension, but that may be a regression relative to HTTP/2, which is unappealing.

Measurement indicates that <<50% of servers implement H2 priorities(https://github.com/andydavies/http2-prioritization-issues) and clients use a fixed scheme they designed once and never changed. 

A key comment I hear repeatedly is that the tree structure is too complex, in particular requests depending upon requests.  The idea of proxy to origin using groups/placeholders for backend connections keeps coming up, but it's not actively used at the moment based on all input I've heard, so I heard a large amount of support for either making that an extension or optional in the same way push is optional in H3.

This is somewhat of a meta-issue, but it keeps coming up and I believe it's extremely important to achieve consensus on something that will be implemented.  As everyone keeps reminding me, that's where we failed with H2.  Let's not let a coin flip decide this time.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/2739