On Wed, Jul 07, 1999 at 11:58:29AM -0400, Mark Allman wrote:
> 
> Mohit-
> 
> We can certainly continue to argue about whether better estimators
> are attainable or whether we may be able to add a mechanism that
> figures out that a spurious rexmt has been sent and performs some
> sort of "undo" on the cwnd/ssthresh adjustment (so we can live with
> more aggressive estimators that are wrong sometimes).  The part of
> your original note that struck me was your recommendation that your
> new code be included in the production netbsd code.  The whole point
> of my note of this morning was that my opinion is that this may not
> be the best time to go messing with the current code until this RTO
> stuff is a bit better understood.

I guess I don't see why that would be the case.  He doesn't actually
change the RTO estimation algorithm, so I don't see how his code
would make things any _worse_.  When research reveals more information
on how to tune RTO estimation, presumably we can add that to Mohit's
code just as we could have to what was there before, no?

Thor