[quicwg/base-drafts] Asymmetry in the handling of SETTINGS frame (#1846)

[Kazuho Oku <[e-mail]>]

[Section 2.3 of the editor's draft](https://quicwg.org/base-drafts/draft-ietf-quic-http.html#rfc.section.2.3) forbids a server from sending data on any other stream until it receives a SETTINGS frame on the control stream, while such requirement does not exist for the client even for the 1-RTT case.

I hereby propose to remove the requirement for the server as well. The requirement is a source of unnecessary complexity; it requires a latchet that blocks the server to send a HTTP response until it receives the first octet of the control stream.

IMO, not knowing the QPACK table size cannot justify having the blocking rule for the server-only; the same rule does not exist for the client and therefore a client is allowed to start sending requests without knowing the QPACK table size. Having symmetry has the potential of simplifying the specification and implementations. For example, a HQ stack can send HTTP messages without using the dynamic table until it receives the SETTINGS frame on *both client-side and server-side*.

Note that even if we relax the specification, the endpoints will have the freedom to block until it receives the first octet of the control stream.

-- 
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/1846