Re: [quicwg/base-drafts] HTTP/3 Zero-weighting (#2723)

[Robin Marx <[e-mail]>]

I just wanted to come back to this PR for a moment.

Seeing as we are now seemingly moving closer to HTTP/2 with the re-introduction of exclusive priorities and sending PRIORITY frames on the control stream, I feel that maybe there might be room for this to be re-considered.

I originally proposed it as a counterpart to #2700 that was much more closely aligned to the typical H2 setup. Now that the Orphan Placeholder has been merged, I think it makes sense to consider the 0-weight semantics in a broader setting. While it is still a departure from HTTP/2 in general, I feel it's close enough to be considered in the QUIC wg. 

(sidenote: it seems that the nodes under the orphan placeholder are still sent in Round Robin fashion rather than FIFO, which I thought we agreed is sub-optimal as default behaviour. Any idea if this was intentional @MikeBishop? If the orphan gets FIFO semantics, this aligns with the 0-weighting in this proposal).

If there is interest, I can look into maybe also incorporating @DaanDeMeyer's suggestion https://github.com/quicwg/base-drafts/pull/2723#issuecomment-493668705 of making it symmetrical for weights of 255 as well. 

I am currently finishing up a body of experiments comparing various prioritization schemes, and this is one of them. I should have some concrete numbers to share in the coming weeks. If it is decided that we do not immediately want to make progress on this (e.g., it should be decided by the HTTP wg), I would appreciate the PR to remain open until we can discuss the measurement results. 

-- 
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/pull/2723#issuecomment-502478532