Subject: Re: kern/9085: enabling RFC1323 support causes some TCP connectionsto stall
To: Kevin Lahey <firstname.lastname@example.org>
From: Mark Allman <email@example.com>
Date: 02/28/2000 21:24:45
> I continue to believe that as we get fatter and fatter pipes,
> it'll be more and more important to have stuff like this turned
> on. OTOH, when folks like Jamshid Madavi and Mark Allman are
> against something, I start to wonder what I'm overlooking...
I see the point... But, my claim is that it doesn't matter how fat
the pipe is. If the content is small, such that you're never going
to be able to utilize the bandwidth (because of TCP's congestion
control) it doesn't matter, so why burn the bits?
The ideal TCP in my head has autotuned socket buffers and some new
"large window extension" that is not like RFC1323 at all. What we
really want to do is to be able to turn on the large window
extensions when we need them (i.e., when cwnd exceeds 64KB), not at
the beginning of the transfer. We should not have to know apriori
that we are or are not going to need the extensions. Because, as
you point out, that is a hard question to answer in general... Is
the content big enough to use big windows? Is the pipe fat enough?
Is the congestion modest enough? Etc.
Of course, replacing RFC 1323 with something new and better is a
huge deployment headache.
(Also note that I am not saying 1323 is junk. I just believe we
have learned a few things since its publication.)