Subject: Re: TCP problems resolved
To: Matthew Fredette <fredette@MIT.EDU>
From: David Brownlee <>
List: port-sun3
Date: 12/29/1999 00:38:13
	Is it possible for someone to tcpdump a session with and without
	net.inet.tcp.rfc1323 and send-pr the result (assuming they have
	the problem with rfc1323 enabled :)

	That we we could get a tcp guru to look at it :)


On Fri, 24 Dec 1999, Matthew Fredette wrote:

> [re: setting net.inet.tcp.rfc1323 to zero]
> > 	Hmm - do you see any problems to machines on your local net,
> > 	or is it all to remote hosts?
> > 
> > 		David/absolute
> FWIW, I also have to set net.inet.tcp.rfc1323 to zero to get correct
> network performance.  After reading the RFC a few years ago, it made
> sense that I would have to, since my 3/60 gets its connectivity
> through a 33.6k modem, and the RFC describes "TCP extensions to
> improve performance over large bandwidth*delay product paths" - not
> what I have at all.  (I also had to do this for my old NetBSD/i386
> machine.)
> My roommate's machine also sits behind the modem, running Linux
> (Debian 2.1), and it had the same problem with an equivalent solution.
> Later, when reinstalling the machine, he realized that answering "no"
> to a "do you have a fast network connection?" prompt would install
> things to set the rfc1323 value to no at startup.
> Maybe NetBSD could use a similar prompt and a new rc.conf variable?
> Matt
> --
> Matt Fredette
> "If you understood everything I said, you'd be me."  - Miles Davis