Re: [quicwg/base-drafts] Are Separate Uni and Bidi Stream Limits Meaningful? (#2358)

[Martin Thomson <[e-mail]>]

Though functionally a bidirectional stream with accompanying STOP_SENDING looks like a unidirectional stream, it is unlikely that an application protocol will treat it as being unidirectional, because the type of stream is input to processing - bidirectional streams have different semantics.  So I think that we *can* rely on application protocols dealing with this case properly.

But the split has utility.  At least in h3, this avoids the problem @LPardue describes.  Firefox treats responses to requests and pushes very differently from a flow control perspective and the split there makes things considerably more efficient for us.  MAX_PUSH_ID is insufficient - we want that to be much higher than the number of concurrent unidirectional streams so that we can learn about potential pushes.  The actual limit needs to be lower, particular since the utility of push is... questionable.

I can only speculate about other protocols, but I think that we have enough justification here to keep the design as-is.

-- 
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/2358#issuecomment-456576462