Re: [Tsvwg] Input for draft-paxson-tcp-rto-00.txt

[Mark Allman <[e-mail]>]

I have been watching this thread with some obvious interest.  My
high-oder bit goes something like this...  

First, we have studies that suggest a 1 second minimum RTO is
conservative.  There are suggestions floating around (including
those presented by Reiner) that very well may be better than using a
1 second minimum.  I don't care to dispute Reiner's assertion that
his ideas are superior.  However, I'd rather see a bunch of evidence
that the scheme is better before we decide to change away from
something that is known to be nicely conservative in favor of a
scheme that is more aggressive.

I would hope those working on "better" RTO estimators will
thoroughly investigate the new estimators.  And, that these
investigations will feed into IETF standards track documents at some
future point (if the algorithms turn out to be appropriate).
However, I see no reason to block this document on the possibility
of getting evidence of a new and better RTO estimator at some point
in the future.

allman