- From: Watson Ladd <watsonbladd@gmail.com>
- Date: Tue, 30 Jul 2024 17:31:15 -0700
- To: Patrick Meenan <patmeenan@gmail.com>
- Cc: Josh Cohen <joshco@gmail.com>, Julian Reschke <julian.reschke@gmx.de>, ietf-http-wg@w3.org
So I think there's a number of things to think about here, inspired by sitting in MoQ (sorry I missed httpbis). There's the Web, HTTP, and the underlying transport technology. The relation of HTTP requests to the underlying connection technology is very different in H1, H2, and H3. At the same time a lot of intermediaries only handle H1, and aren't really going to change. There's a number of assumptions that sink into HTTP processing that break a lot of the things people want to do that can end up pretty deep. MASQUE doesn't run over WebSockets but directly over QUIC. MOQT can map onto WebTransport, but that's a bit oddly layered. Maybe what we should do for efficient state diffs is trifurcate the issue. There is retrieving an object something HTTP can do easily, and subscribing to a stream of updates from a point is something that's easier over QUIC+occasional submission. Put the three together and you get Braid like semantics. This does come with a intermediary challenge: I don't know how we do it end to end easily, but one could map to polling transient resources marked as uncacheable to emulate streaming or fit on top of WebSockets. Sincerely, Watson -- Astra mortemque praestare gradatim
Received on Wednesday, 31 July 2024 00:31:31 UTC