Re: [tsvwg] FIRST-CUT Agenda for TSVWG in Prague

["Holland, Jake" <[e-mail]>]

Hi Gorry,

I'm writing to express an interest in the SCE work.

I note that it's being considered but not committed on the
agenda, and I'd like to request its inclusion, if possible:

On 2019-03-12, 02:42, "Gorry Fairhurst" <[e-mail]> wrote:
-------------- Requested made or talk being considerd ------------------
Dave Taht: Some congestion experienced (requested, linked to ECN)
             draft-morton-taht-tsvwg-sce

(My sympathies on how full the agenda is, I know you have some tough
choices...)

The SCE draft is early work, but I find it especially interesting
for a few key reasons:

- it has backward-compatible semantics for ECT(1).  This comes
  with some possibly significant ramifications for incremental
  deployability.
- it seems to me there might be an interesting potential for
  responses that vary differently with the congestion control,
  rather than a response tuned to Reno's cwnd characteristic,
  as described in section 2.1 of the dualq draft [1].
  (this is especially relevant in light of some of the interest
  we've seen in some alternate congestion control proposals like
  BBR.)
- SCE has some advantages over L4S with regard to IPR constraints,
  which may be particularly relevant for AQMs in CPE devices based
  on OpenWRT or other open platforms. [2]

These seem to me like some major points in favor of supporting its
examination and consideration by the wg, (or perhaps tsvarea, in
case this raises questions about options like re-opening aqmwg or
something).

I also acknowledge there's a significant point in favor of not
bothering, which is the advanced state of the (incompatible) L4S
experiment proposal.

If there's wg consensus that L4S has become a de-facto standard and
it's therefore unproductive to look at incompatible proposals for
ECT(1) semantics, I'll withdraw my request.

However, I'd also like to register my vote against that position,
and argue that it may be useful to at least briefly discuss options
(or criteria) for running separate experiments based on each proposal.

If you do manage to find a way to fit this talk in, I'll also suggest
that it might be useful to arrange the schedule so that questions about
consensus on the L4S documents can easily be deferred until we get a
chance to hear about SCE, since they might have some bearing on one
another.

Thanks for your consideration.

Regards,
Jake

[1]
https://tools.ietf.org/html/draft-ietf-tsvwg-aqm-dualq-coupled-08#section-2.1
[2]
https://datatracker.ietf.org/ipr/search/?submit=draft&id=draft-ietf-tsvwg-aqm-dualq-coupled