Re: [quicwg/base-drafts] hq: Does ignored data/frames cause flow control updates? (#1580)

[MikkelFJ <[e-mail]>]

My comments are purely in relation to transport. I can't say if there is unclear language in the HQ spec.

Even if HQ is a primary use case for QUIC, the transport cannot be concerned with how an application uses data. So it HQ (the app in this case) layer needs to notify the transport when to advance MAX or it must have another semi-automatic way to signal this as I suggested earlier. So the answer to the question must be a yes - the app must help transport advance flow control as long is it expects to receive more data - on that stream in particular, and on streams in general due to the global flow control credits. Also if a given range is ignored.

When a stream is terminated using RST there is an exact stream offset included. Data up to that offset counts against the global flow control while the max for the specific stream no longer matter because it is retired and pending data will not wait for more credits on that stream.

Due to race conditions data might be sent after the RST offset and there could be multiple RST's. However, if there are multiple RST's they must agree, or it is a protocol violation.

I assume that data sent after the RST offset will be rolled back from global credits counting, but I'm not entirely sure about this. Perhaps this should be clarified?

-- 
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/1580#issuecomment-406017758