Hi all, Just a reminder that this thread, and Ian's comment that we're picking at, predates the "Next steps for Multipath" email that Lars sent on November 20 [1].
There's levels of nuance here: a self-test experiment, interoperable experiments in developer environments, experiments that emulate a production-like deployment, experiments on in-situ production equipment, and so on. Interest, ability and time in all of these levels is not universal. So my _personal_ take is that folks would like to establish confidence in a multipath design, as if it were deployed in the production environment that it is intended, before adopting it as a solution. We've seen before with stream priorities that a rush to design a signalling mechanism, while punting on scheduling, can result in defining something that is not generally useful - the antithesis of standardisation. I think it's fine to have the logical separation between signalling and scheduling but the practical engineering reality is that they get deployed together. Cheers Lucas [1] - https://mailarchive.ietf.org/arch/msg/quic/rcPf7u9AHIGwNr6j0ZqrqFujVvk/ On Wed, Dec 2, 2020 at 5:08 PM Spencer Dawkins at IETF < [email protected]> wrote: > Hi, Ian, > > I'm not Mirja, of course, > > On Wed, Dec 2, 2020 at 10:29 AM Ian Swett <[email protected]> wrote: > >> Thanks Mirja, that deployment setup sounds like it could make it >> difficult to get deployment experience prior to publishing an RFC? Given >> the complexity of multipath, shipping an RFC without deployment experience >> seems unappealing to me, but maybe there is an opportunity to get some >> deployment experience earlier? >> > > At least some of the presentations at the October virtual interim on QUIC > multipath were not 3GPP ATSSS-related. > > I agree with Mirja's point on the difficulty of obtaining experience at > scale in a 3GPP network for an experiment, but I'm hoping there is enough > commonality between other people interested in multipath and QUIC that > SOMEONE will be able to report experiments for a proposal that would also > work for 3GPP. > > That's one of the reasons I started working on > https://datatracker.ietf.org/doc/draft-dawkins-quic-what-to-do-with-multipath/ > - to see what was common across people with different use cases. > > Best, > > Spencer >
