Re: [quicwg/base-drafts] encoder stream can deadlock (#1420)

[Mike Bishop <[e-mail]>]

It's also worth noting that we really have two situations here:

On the encoder stream, nothing says the encoder has to send all updates in a single block.  It can query for available flow control, write that much, and wait for more credit.  The only time it deadlocks with this approach is when a single header is bigger than the flow control credit being issued, in which case don't put those huge Kerberos tickets in your dynamic table, for the love of God.

On the request stream, you do have to send the whole header set all at once, which is harder.  However, you're also bounding the size of headers you'll accept on any given stream; we can recommend that the flow control window be large enough to transfer the expected header set or that you be willing to buffer as needed.

-- 
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/1420#issuecomment-395217431