Re: [aqm] publishing algorithms

[Dave Taht <[e-mail]>]

On Tue, Apr 1, 2014 at 5:11 AM, Wesley Eddy <[e-mail]> wrote:
> Hello AQMers.  As chairs, Richard and I had been planning to let
> the evaluation guidelines converge and then use those to guide
> adoption of algorithm documents.
>
> However, we now think there may be value in not waiting so long,
> and getting some algorithm documents moving along more quickly.

I saw this came out on april 1, and have a tendency to ignore stuff
that comes out that day. :)

> We hope you can provide some feedback on the plan below:
>
> 1) Starting soon, we may look to adopt a small number of algorithm
>    drafts for Experimental, with the goal that by doing so, it will
>    increase the number of eyeballs and independent reviews of them, and
>    enhance the quality, since people may be implementing to the drafts
>    in order to test using the evaluation guidelines.  Each algorithm
>    *must* clearly identify which types of use cases / scenarios it is
>    targeted for.

We submitted the codel and fq_codel drafts for consideration prior
to ietf 89, and have got no feedback from anyone as yet. The first
went into tsvwg

https://datatracker.ietf.org/doc/draft-nichols-tsvwg-codel/?include_text=1

and the second is seemingly homeless as yet:

https://tools.ietf.org/html/draft-hoeiland-joergensen-aqm-fq-codel-00

As they were being developed independently, they have a few nits
in relating to each other that need to be resolved, presently.

If this is a new requirement, and it seems like it is, then those will
need to be revised more fully, although each has caveats and
use cases described already late in the document.
(sections 6 and 9 in the first, section 10 in the second)

Are you proposing an explicit use cases and scenarios section?

> 2) Adoption of an algorithm spec as a working group draft will require
>    working group consensus that the algorithm looks attractive to
>    experiment with for the stated scenarios, and multiple parties will
>    plan to be looking at it, testing, analyzing, providing feedback,
>    etc.

+1

I note that I have another draft in progress that I hope to complete before
ietf-90 describing multiple deployed aqm + packet scheduling + classification
systems in the hope to get the overall conversation on track.

http://snapon.lab.bufferbloat.net/~d/draft-taht-home-gateway-best-practices-00.html

and I'd like to talk to that at ietf 90 if I'm not too broke to go. Otherwise
hawaii.

>
> 3) The evaluation guidelines / scenarios drafts being worked on
>    separately will guide the later selection of one or more Experimental
>    algorithms to become Proposed Standards with applicability
>    statements for the scenarios they have been evaluated in.

At this point I must confess to being deeply dissatisfied with the
progress in the AQM guidelines document. I don't know what to do about it,
I've tried to talk about the issues being totally missed in things like this:

http://www.bufferbloat.net/projects/cerowrt/wiki/Smart_Queue_Management

and here on the list...

and at the moment I'd prefer more folk have better models and understanding
of actual traffic types on various media types before even trying to
write anything down as to what valid tests are.

I think I have finally come up for what I'd like this group to be doing,
which is at least partially out of scope for the charter, but the acronym is
nice: "CQM" - "Comprehensive Queue Management", and a goal is to
try and design something that leverages the work other groups such
as rmcat are doing.

> We're interested to know if the working group thinks this sounds like
> a good idea, bad idea, or any other thoughts.

By all means, let's get folk fiddling with the working algorithms and code
that exists today and is distributed now with every Linux box. Perhaps
some more BSD support will show up soon, or hardware support in
some other gear other than in what I describe above.

I'm all in favor of new algorithms too, if anybody's got some...

>
> --
> Wes Eddy
> MTI Systems
>
> _______________________________________________
> aqm mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/aqm



-- 
Dave Täht

NSFW: https://w2.eff.org/Censorship/Internet_censorship_bills/russell_0296_indecent.article