Subject: Re: MTU discovery buglet
To: None <explorer@flame.org>
From: Kevin M. Lahey <kml@nas.nasa.gov>
List: tech-net
Date: 11/23/1997 17:47:21
In message <19971123225506.24656.qmail@kechara.flame.org>,
explorer@flame.org writes:
>
>When I turn on PMTU discovery, I get lots of clutter:
>
>Destination        Gateway            Flags     Refs     Use    Mtu  Interface
>pm300-36.tr.net-li near-gw.cygnus.com UGHD        0      247   1006  de0
>pm300-39.tr.net-li near-gw.cygnus.com UGHD        0      224   1006  de0
>pm301-14.tr.net-li near-gw.cygnus.com UGHD        1     2148   1006  de0
>pm301-29.tr.net-li near-gw.cygnus.com UGHD        0     4226   1006  de0
>pm301-46.tr.net-li near-gw.cygnus.com UGHD        0      405   1006  de0
>
>Note that only one of these hosts is really in use.  Shouldn't
>the others expire when the reference count reaches zero, or
>at least after a timeout after the ref count reaches zero?

Yup, it should.  

That part's not quite done yet.  It doesn't look to me like
it'll be done in time for 1.3, either.  My apologies, I've been
dawdling on this for quite awhile.  The other thing we need is
some sort of black hole detection, to figure out when routers
are failing to emit (or blocking) the fragmentation necessary/
packet too big ICMP messages necessary to do MTU discovery.
That's a bit tougher, and there are plenty of stacks that
don't support it yet, but it sure would be nice.

Kevin