search

ietf-wg-webtrans / draft-ietf-webtrans-http2

WebTransport using HTTP/2
Other
7 stars 5 forks source link

Should H2 be permissive about flow control violations #94

Open ekinnear opened 10 months ago

ekinnear commented 10 months ago

H2 matches H3 in being permissive about flow control violations around the stream limit.

the asynchronous nature of the protocol

A server will only be unsure if it reduces the limit and the SETTINGS frame hasn't been acknowledged yet. But then it is sure, because everything opened before the SETTINGS+ACK uses the larger limit and everything after uses the reduced limit... So maybe we can completely rewrite this little bit.

_Originally posted by @martinthomson in https://github.com/ietf-wg-webtrans/draft-ietf-webtrans-http2/pull/91#discussion_r1368470323_

ekinnear commented 10 months ago

Discussed in editor's meeting:

If you're above the limit, you lower that limit, and the peer ACKs that lowering, you only enforce the "don't go above the limit" when a new session is opened, so it doesn't matter how long the peer takes to close everything down, they have to get below the limit if they want a new session.

DavidSchinazi commented 10 months ago

Chair: discussed in the WG session at IETF 118. No objections raised.

@achernya also pointed out that there might be some interesting interactions between SETTINGS_WEBTRANSPORT_MAX_SESSIONS and DRAIN_WEBTRANSPORT_SESSION that could lead to suboptimal incentives around keeping draining sessions around for longer than needed. @ekinnear took an action item to look into this.

ekinnear commented 1 month ago

(Pinged to discuss in more detail)