2025年3月26日(水) 20:32 Lucas Pardue <[email protected]>: > Hi Alan, > > Speaking as an individual, I'm unclear what people want documented over > and above what is already captured in RFC 9000 Section 2.4 "Operation on > Streams" [1], which defines a set of functions on streams application > protocols can rely on. > > Furthermore, the applicability draft (RFC 9308) has even more text about > how application might use streams in Section 4 [2]. > > I'm of the opinion this is already an ample amount of detail about *QUIC*, > that doesn't tread on implementation-specific concerns. >
+1. IMO, API describes two things: what can be done and how they can be done. When developing QUIC v1, we made the choice to specify the former but leave the latter to each QUIC stack. Considering how different each QUIC stack is, I think we made the correct choice. IIUC, some QUIC stacks run on top of the Unix Socket API, hiding the sockets. Others run as a codec that provides and consumes QUIC packets, letting the application pass the QUIC packets to / from the UDP connection. Some QUIC stacks have read / write calls. Others invoke application-provided callbacks for writing data to be sent / storing data that have been received. These choices have their own benefits and drawbacks, but different QUIC stacks are designed for different deployments in mind, and have different design goals. I do not think we can agree on a single API that defines how applications and QUIC stacks should communicate, nor would it be a good idea to do so. > Other specs, like WebTransport can (and should) document their world where > appropriate. > > Cheers > Lucas > > > [1] - > https://datatracker.ietf.org/doc/html/rfc9000#name-operations-on-streams > [2] - https://www.rfc-editor.org/rfc/rfc9308.html#name-use-of-streams > > On Wed, Mar 26, 2025, at 07:38, Michael Welzl wrote: > > …. following up, I would also like to raise awareness of RFC 9622. This > doesn't specifically expose QUIC (part of the point! explained in RFC > 9621), but QUIC was *very much* a part of the design logic and discussions > underlying this API. > > Cheers, > Michael > > > On 25 Mar 2025, at 22:28, Alan Frindell <[email protected]> > wrote: > > During the QUIC session at IETF 122 while discussing QMux (formerly QUIC > on Streams), a couple folks at the mic advocated for documenting the “QUIC > API”. I wanted to raise awareness that this is something like this is > covered today in the WebTransport Overview draft, section 4 “Transport > Features”. > > > https://www.ietf.org/archive/id/draft-ietf-webtrans-overview-09.html#section-4 > > Thanks > > -Alan > > > -- Kazuho Oku
