Subject: Re: tcp_do_rfc1323 vs Linux - any solid understanding?
To: Ken Hornstein <firstname.lastname@example.org>
From: Peter Galbavy <email@example.com>
Date: 06/10/1996 16:54:26
> I'm sorry, but this has not been my experience at all. I first noticed
> these problems when connecting to Solaris and Irix boxes.
I am talking about Ethernet for my problems.
> I'd like to point out that in my particular case, turning off VJ header
> compression made everything work, and the performance between NetBSD and
> Solaris/Irix boxes is now comparable. If you could give more exact info
> on why you think the NetBSD rfc1323 implementation is "sub-optimal", I'd
> like to hear it. But in all honestly, every problem I've ever heard that
> was caused by rfc1323 IP options was always the fault of some intervening
> system, not the NetBSD system itself.
The problem arises when you have no control over the path between two hosts.
For example a SUP to sup.netbsd.org may go through a shit load of various
routers and ancillary equipment. If some fool has configured VJ header
compression (the default for many PPP connections) on a line between me
and the server, then bang. This is of course assuming it is the VJ + 1323
interaction that is causing some of my problems.
On Ethernet, turn 1323 off seems to make NFS over TCP more responsive.
Just a feel at the moment, but I can check later.
Never the less, on WAN connections I suspect it is a good defensive practice
to turn it off by default, and to a degree on LANs too.
Peter Galbavy firstname.lastname@example.org
@ Home phone://44/973/499465
in Wonderland http://www.wonderland.org/~peter/