> John Kohl writes:
> > I've submitted a PR enhancement request to add a sysctl option to turn
> > on/off RFC1323.  I've been using those changes it at every reboot (to
> > turn off RFC1323) for about 2 months now; it works great.
> 
> The IETF ought to be told that 1323 is causing performance
> problems. Are they truly caused by 1323 and not by the implementation
> thereof?

I have just rebooted with tcp_do_rfc1323 set to 0. And:

rfc0822.txt: 107337 bytes received in 27.88 seconds, 3.76 K/s.

This is *much* better than the few hundred bytes a second with it turned
on. Still not the 7K/s I have seen before but nevertheless good. As
for line traffic and a data point, I just got 5.5K/s from a SunOS box.

I second John's request for a sysctl variable. It was something I was
thinking of trying.

The IETF should be made aware I think, since 4.4 lite based systems
are really the only ones that have 1323 support. Any other OS'es
out there support it yet ? Alternatively, has Van Jacobson got any
ideas, since that man seems to know everything else about IP ?

Regards,
-- 
Peter Galbavy                                           peter@wonderland.org
@ Home                                                 phone://44/973/499465
in Wonderland                              http://www.wonderland.org/~peter/
                               snail://UK/NW1_6LE/London/21_Harewood_Avenue/