Re: [quicwg/base-drafts] Required state for retaining unacked RETIRE_CONNECTION_ID frames is unbound (#3509)

[Kazuho Oku <[e-mail]>]

@marten-seemann 
> My fear regarding #3547 is that no matter how high we choose the limit, there will always be a loss pattern that will trigger a protocol violation.

Thank you for raising the concern, I think the recommended behavior when exceeding that limit is to stop sending RCID frames for some sequence numbers rather than triggering a protocol violation. That's actually what I had in mind hence called it "leaky."

The benefit of stopping retiring some sequence numbers is that the worst case becomes endpoints running out of CIDs - something that is expected to happen on ordinary connections too. Note also that in HTTP/3, reestablishing a connection in 0-RTT is always a viable option.

To summarize, I think that the concern can be resolved for HTTP/3 by changing the recommended behavior from resetting the connection to stop sending RCIF frames for some sequence numbers.

-- 
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/3509#issuecomment-607220688