jmarin@pyy.jmp.fi (Jukka Marin) writes:
> How come no-one else is doing this [path MTU discovery]?  I have
> only had this problem when the remote system has been running m$
> crapware.

In this case MS is doing it better than a lot of systems out there.
Sending 3 small packets when 1 big one will do isn't polite these
days.  The cpu load on routers is to a large part dependent on the
number of packets they are asked to route.  Tripling the interrupt and
cpu load on the core routers just isn't nice.  Sad isn't it when folks
have to use MS as the show-and-tell example?

BTW, NetBSD could do path-MTU discovery, but it is off by default.
Avoiding problems with misconfigured firewalls is no doubt one reason
its still off.  (Routers that don't send back icmp's for packets that
they need to fragment but aren't allowed to might be another more
minor problem.  Are there still any out there that have this
lossage???)

-wolfgang