Re: Reopening jumbo frames in IS-IS

["Stephen Sprunk" <stephen at sprunk.org>]

Thus spake "Bob Hinden" <bob.hinden at nokia.com>
> Radia,
>
> > What are the technical reasons that IEEE does not like large packets?
>
> I can't speak for IEEE, but I have always thought that one of there
> reasons that Ethernet has been so successful is that that IEEE tried
> very hard to insure backward compatibility between the different
> versions.  It made it easier to bridge between 10M/100M/1G/10G/etc.
> versions, new versions didn't break any protocols that ran over Ethernet,
> and it is easier to build NICs that supported a range of variants.
>
> It also avoided having to build things like a FDDI/Ethernet bridge I once 
> heard about that supported IP fragmentation.  I bet you remember that :-)

Wait a minute...  If Ethernet supported jumbo frames, the FDDI-Ethernet 
bridge wouldn't have needed to support fragmentation -- just set the 
Ethernet side's MTU to 4470 and all would be well.

I dealt with this many times when bridging Token Ring and Ethernet, and the 
only solution that worked in every case was to drop the Token Ring's MTU 
(network-wide, since TR implies SRB) down to 1500 -- a horrible kludge. 
Sometimes other tactics worked, including dropping oversized frames with no 
fragmentation, but some SNA apps were really touchy about that.  IP handled 
things a bit better, but still not as well as jumbos would have.

S

Stephen Sprunk      "Those people who think they know everything
CCIE #3723         are a great annoyance to those of us who do."
K5SSS                                             --Isaac Asimov