Subject: Re: Path MTU discovery
To: None <tech-net@NetBSD.ORG>
From: Wolfgang Rupprecht <wolfgang@wsrcc.com>
List: tech-net
Date: 07/15/1997 10:05:06
kml@nas.nasa.gov (Kevin M. Lahey) writes:
> In private e-mail, Perry pointed out that the BSDI stack supports it, 
> and mentioned something about it being publicly available.
> We have source for BSDI at work, so I didn't pursue the publicly
> available part, but the code is definitely there.

I've used BSDI 2.1 for a year now and the MTU path discovery code,
without fail, will "discover" 1500 byte MTU, 1460 byte tcp_mss.  This
is the same answer as my SunOS machine that has been running a fixed
1460 byte tcp_mss for > 5 years now.

So unless one is on a network where the MTU is larger than 1500 (FDDI,
100b-tx w. hacked packet size etc.)  there is very little interesting
for the MTU discovery code to discover.  The answer is almost
certainly "1500".

Has anyone played around with the "big packet" code for the DEC TULIP
chips?  (eg. for packets of > 1536 bytes).  Does it work?  Can these
packets pass through any hubs, or does one need to use a crossed-over
ethernet cable between two 'big-packet' boxes?

-wolfgang
-- 
Wolfgang Rupprecht    <wolfgang@wsrcc.com>         http://www.wsrcc.com/