On Wed, Jul 13, 2022, at 19:02, Gorry Fairhurst wrote: > I'm unsure why you think the QUIC protocol is not the best place to > consider such an extension. QUIC has various features that make it a > much easier place to do this type of change - the design of the > transport, the ability to update implementations, and a design which > implies pacing, security, etc.
Simple test: if you took "QUIC" out and replaced it with DCCP, the draft would still work. This is a congestion control document, not a QUIC one. Sure, QUIC is a great testbed. I expect that this will work there. It will also work very nicely with TCP. Google was doing something very similar ~10 years ago in TCP, though maybe less measured and careful.
