> >From what I remember (from back when I worked at a bridge manufacturer), it's
> not actually all that tricky.  You just *can't* bridge packets which don't
> fit on the target network, period, but header rewriting is perfectly fair,
> and as I recall FDDI to ethernet only requires bit-flipping the MAC address.
> (I *think* FDDI addresses are assigned from the same "namespace" as Ethernet
> addresses, so there shouldn't be any collisions.)  Of course, I could be
> misremembering, and even if not, it is tricky to *use* such a bridged network
> (you have to tell the FDDI hosts to use smaller packets, which is relatively
> easy for IP but not necessarily easy for any other protocols).

One of the reasons I use FDDI, though, is the large MTU.  It makes NFS
and other protocols that can use large packets work REALLY well.

In order to not have to change the MTU of everything, you'd have to
fragment out 4096 byte FDDI frames into 1500 byte ethernet frames.