Re: [tsvwg] path forward on L4S issue #16

[Ingemar Johansson S <[e-mail]>]

Ok. So I take it that you dismissed L4S as a crackpot idea and gave it a blind eye for a few years. That was your decision and now you stand with equipment that only speaks RFC3168, you could have corrected that issue a few years ago but chose not to do it just because you ignored ongoing work.
Please forgive me but I fail to see why your ignorance is a problem for IETF?
/Ingemar

Hämta Outlook för iOS<https://aka.ms/o0ukef>
________________________________
Från: Jonathan Morton <[e-mail]>
Skickat: Wednesday, June 17, 2020 6:54:26 PM
Till: Ingemar Johansson S <[e-mail]>
Kopia: Rodney W. Grimes <[e-mail]>; Sebastian Moeller <[e-mail]>; [e-mail] <[e-mail]>
Ämne: Re: [tsvwg] path forward on L4S issue #16

> On 17 Jun, 2020, at 7:27 pm, Ingemar Johansson S <[e-mail]> wrote:
>
> With that in mind, shouldn't it have been on the table to consider the
> possible future existence of L4S in the different realizations of fq-codel
> (and OpenWrt) development?

Codel and its progeny were developed on the assumption that RFC-3168 was not going to be obliterated by an incompatible protocol overlaid upon the same codepoints.  A reasonable assumption, given that RFC-3168 was well established as an RFC at that time and widely implemented (if not very widely used in practice), and that DCTCP was explicitly known and accepted to be unsafe to use outside of controlled environments.

And I have to say, Codel represented a genuine step forward in the state of the art for AQM.  It is far easier to configure correctly than most of its predecessors, and does not over-react to brief bursts of traffic (ie. shorter than the estimated RTT), unlike most RED derivatives.  I think it's entirely fair to credit the present increase in RFC-3168 ECN deployment to Codel's existence.  I note in passing that Codel was explicitly designed to match TCP AIMD behaviour, not high-fidelity 1/p responses.

IIRC, when L4S first appeared, the idea of deploying DCTCP over the Internet was rightly dismissed as being completely crackpot.  So everyone in the Codel community simply ignored the RITE project as irrelevant, when it became clear that was the approach they were taking.  Unfortunately, the IETF took them more seriously, and now here we are.

If and when the L4S team manage to put forward a workable solution to the basic problem, I'm sure we will be able to have a more fruitful discussion.  For the time being, we can only eliminate certain classes of solution as unworkable.

 - Jonathan Morton