Re: More on MTU discovery...

["Brett D. Watson" <bwatson () mci net>]

> Based on the number of replies, it sounds like there is quite a lot of
> interest in this subject.  I didn't want to call out any sites in
> particular, because I didn't want to pick on any providers.  I sent
> mail to three people with problems this morning at around the same
> time I sent the nanog mail, and have so far only heard a response back
> from one.
>
> The one who responded turned out to be exactly what Matt suggested --
> mismatched ATM MTUs on Cisco routers.  Apparently the Cisco default is
> something like 4470 even though ATM default MTU is supposed to be
> 9180.
>
> The other two sites haven't responded.  The IP hop in question in both
> cases was, I believe, a cisco router (based on the prompt I got when I
> telnetted to them).  However, since I know that in most cases Ciscos
> seem to do the right thing, I suspect that these sites have other
> problems down at layer two.  Lots of people out there make bridging
> products who have never heard of RFC1191...

  i would be one of those two sites, and i responded tonight.  the mtu problem 
you found (and matt found it because he worked for this particular customer 
this weekend) was on the customer end of one of our links. i couldn't quite 
figure out what you wanted me to do about it.  i'll just say for the record 
that it seems to be a problem between the customer's vendors' router and our 
router :)  the customer is aware of the problem and is following up with their 
vendor.  i was a little taken back by your note which basically sounded like:

"you've got an obvious problem, you need to fix it quick as everyone depends 
on this resource"

 that may not be what you meant, but it sounded that way.

-brett


> If I hear back from the providers for the other two sites, I'll post
> the answers to the list.
>
> --Jamshid