[dccp] draft-ietf-dccp-tfrc-voip-03

[Sally Floyd <floyd at icir.org>]

We have just submitted a new version of draft-ietf-dccp-tfrc-voip.
The title is now: 

"TCP Friendly Rate Control (TFRC): the Small-Packet (SP) Variant", 

and it is available at:

http://www.icir.org/floyd/papers/draft-ietf-dccp-tfrc-voip-03.txt
http://www.icir.org/floyd/papers/draft-ietf-dccp-tfrc-voip-03.ps

I think this is ready for Working Group Last Call (unless
feedback indicates otherwise).  The draft has new simulations,
but nothing that changes the content of the presentation at the 
last IETF:

http://www.icir.org/floyd/talks/voip-Nov05.pdf
http://www.icir.org/floyd/talks/voip-Nov05.ppt

The problem with the TFRC-SP (small packets) variant of TFRC
specified in this draft is that it works great in an environment
where large and small packets are equally likely to be dropped
(e.g., a Drop-Tail queue in packets), but it is too aggressive
in an environment where small packets are less likely to be dropped
(e.g., a Drop-Tail queue in bytes).  Our current measurements
indicate that current paths from web servers don't generally treat
large and small packets much differently, but there has been
an unavoidable delay in getting this finished, and when it is all
in, it will be written up in a separate document.

Regardless of measurement results, my proposal would be that this
document be considered for Experimental.  However, I would propose
the version of Experimental that means "we don't recommend this for
widespread deployment, but we would like to see further experimentation
in the real world", rather than the version of Experimental that
means "this can be widely deployed, and one day might go to Proposed
Standard" (e.g., as done by NewReno, ECN, and HighSpeed TCP).

Feedback would be great.
  
Many thanks,

- Sally
http://www.icir.org/floyd/