Replacing the "RED Manifesto" with an "AQM manifesto" in the IETF

[Jim Gettys]

The ICCRG meeting at last week's IETF went very well, as did a variety of
live demos of fq_codel.

You can find the ICCRG slide sets here:
https://datatracker.ietf.org/meeting/86/materials.html

 and, though the sessions were not video'ed by the IETF, Dave and I used
our phones and may put some low quality video up later.

See: http://www.ietf.org/blog/

 and

https://plus.google.com/u/0/107942175615993706558/posts/A1seNKANmDg

My great thanks to Dave Taht and Comcast to pull off a live demo of
bufferbloat helping drive bufferbloat's reality home to people.  Such live
demos are always a huge amount of work.

The tsvarea meeting resulted in consensus to work toward establishing a
working group and updated set of best practices and RFCs regarding AQM
recommendations. First up was establishing a new mailing list for "aqm",
which can be joined here:

https://www.ietf.org/mailman/listinfo/aqm



Now is the time to see how standards sausage is made!

Some advise was published in the ICCRG meeting about how to proceed further:

http://www.ietf.org/proceedings/86/slides/slides-86-iccrg-4.pdf


Those of you familiar with RFC 2309 may know that it was informally called
the "RED manifesto", and aware that RED's shortcomings doomed it. But such
a document (as a public statement of the IETF that this problem must be
urgently solved).

Fred Baker (who was IETF chair recently) just issued an initial internet
draft:

http://datatracker.ietf.org/doc/draft-baker-aqm-recommendation/?include_text=1


This one is intended to obsolete RFC2309, and is mostly RFC2309 with the
old stuff ripped out, and not a lot of new added. YET! Please join the AQM
list to discuss what the new advice (AQM manifesto) should look like.

Since multiple AQM algorithms can co-exist (e.g. CoDel & PIE), and multiple
flow queuing algorithms, we expect that "one size fits all" is unlikely,
but documenting them (so that purchasing RFP's can reference them) and
explaining their best areas of use; best guess is we'll see a number of
informational RFC's and BCP's result, though standards track for the
algorithms are not out of the question.

The research and development in this area is still very young.  This is the
beginning of a long road. As Matt Mathis put it in the ICCRG meeting, the
results (which you can see repeated in all the slide sets in the ICCRG
meeting) are compelling, and it is important to start the deployment
process without years of optimization: seldom do you see orders of
magnitude improvement shown as everyone did at the meeting.

I would like to thank all of you (and Dave Taht, Kathy Nichols and Van
Jacobson in particular) for helping us to get to this point.  I started as
a lone voice in the wilderness and felt very alone.

With all your help and support there is now a growing chorus and we have
billions of devices to deploy to, and many problems yet to solve.

Again, thanks to all.
                                 Jim
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.bufferbloat.net/pipermail/bloat-devel/attachments/20130317/a8369a17/attachment-0002.html>