Re: Next steps for Multipath QUIC

[Mikkel Fahnøe Jørgensen <[e-mail]>]

I believe it is good to have experimental results as extensions, or
otherwise, fleshing out the various concepts before committing to anything.

One aspect that isn’t clear to me is the symmetri of multipath, and also
for migration post V1.
I for one would like to see more symmetri for data center deployments, at
least for multipath.

Mikkel


On 20 November 2020 at 10.30.39, Lars Eggert ([e-mail]) wrote:

Hi,

Lucas and me got asked what we'd see as the next steps for multipath
support for QUIC. Here's our current thinking:

The QUIC WG remains the default venue for continued open discussion of
multipath experiments and results. The goal for this discussion is to help
steer towards identifying whether a multipath design can emerge that

(1) addresses a number of the use cases discussed during the interim, or
new use cases that are similarly motivated

(2) sees implementation and experimentation/deployment interest from a
number of production stacks

(3) is (as much as possible) a minimally-scoped extension to QUIC v1 that
cleanly integrates with its concepts and other adopted extensions

Until this has happened, we think it'd be premature to adopt one of the
individual multipath designs that have been proposed as a WG work item.

We hope this provides some clarification.

Thanks,
Lars and Lucas