Hi, Lars,

Thank you, and Jana, for tugging on this.


On Thu, Oct 8, 2020, 01:13 Lars Eggert <[email protected]> wrote:

> Hi,
>
> On 2020-10-6, at 10:21, Lars Eggert <[email protected]> wrote:
> > we decided to take up Ian's suggestion to schedule a virtual interim
> meeting to discuss the multipath topic. To find a suitable time before the
> next IETF meeting, please fill out the Doodle at
> >
> > https://doodle.com/poll/iqmap34cszi4qub4
> >
> > by the end of Thursday, Oct 8. We'll pick a slot on Friday, Oct 9 and
> announce it.
>
> reminder to please state your preferences by the end of today.
>
> > On each day, there are two four-hour slots that I think are workable,
> given our participant distribution.
>
> We received feedback that four hours are maybe more than needed, and a
> long slot like this makes scheduling somewhat more difficult.
>
> Speaking as one chair - but without having discussed this to great lengths
> with my co-chairs - the goal for the interim is to try and come to
> consensus on what to do about the multipath work item our current charter
> includes.
>
> A 2h slot is likely OK for this. We'll identify a 4h slots based on the
> Doodle poll and will try to select a 2h window within that slot that
> further maximizes attendance.
>
> Multipath was included in our initial charter, because it was being
> investigated for Google QUIC, which was the main input to our work. As can
> be observed by discussions over the past year or so - and especially during
> the last few weeks - whether the QUIC WG or the IETF should work on
> multipath support for QUIC, and on what time line - is a contentious topic.
> Many options have been expressed in this discussion so far, ranging from
> "remove the charter item and BOF the topic" to "adopt draft-deconinck and
> start the work now" and other options in between.
>
> The purpose of the interim is to provide a higher-bandwidth venue for this
> discussion, in the hopes that some ways forward emerge that might reach
> consensus.
>

We really do need to trim back the tree of potential paths forward, even if
the interim meeting can't converge on a single path forward.

This goal seems sufficient to help the chairs flesh out an agenda.

Other things could be on the agenda, of course.

Best,

Spencer

Personally, I believe that is essential that any complex engineering on
> QUIC, such as the addition of multipath, is motivated by the needs of
> large-scale deployments and driven by engineers that are in a position to
> feed back experiences and data from such deployments into the design
> process in a timely manner. This is how the WG managed to develop the
> existing QUIC specifications in a relatively short amount of time (given
> the changes we made to the initial QUIC design). It would be unfortunate if
> we gave up this proven model of design and engineering.
>
> Thanks,
> Lars
>

Reply via email to