[jan-jaap]

Hmmm, I see some issues here:

• I have systems running IRIX 3.2 / 3.3 / 4.0 / 5.3 / 6.2 / 6.5. I'm not so sure older IRIxes even know about IP aliases.
  
• I'd end up with IP aliases on all but the slowest of my systems and significant hosts files which would have to be in sync across all of them. Maintenance nightmare.
  
• The FDDI IP range currently only exists 'behind' the FDDI router appliance, and my central router (I have other IP ranges and VLANs) publishes a static route for the FDDI IP range via the FDDI appliance. The moment 'FDDI range' IP addresses show up on my ethernet this is going to fail.
  
• If I'm going to assign FDDI range IPs to systems on ethernet, why even keep the ethernet address? I mean, sure, you'd save router bandwidth, but most of these systems don't see a lot of action, are 100Mb/s speed while all routing is at gigabit speeds. Without aliases, only the 10Mb/s systems would remain on "SGI ethernet". Effectively, this is not so different as my idea to put 10Mb/s on a separate IP range, right?
  

I have some 25 MIPS/IRIX systems. I try to have as much configuration centralized as possible and it makes sense to modify my network setup to match the out-of-the-box configuration of IRIX (at least on the subnet where the IRIX systems live) and minimize configuration changes on each system. For this reason alone I would prefer to have routed active if only so I don't need to edit config files on 25 systems (and on IRIX < 6.2 you need to edit the actual startup scripts, there's no static-route.options). This is an example of a (long term) goal, I know routed won't solve my FDDI IP problem.