Carlos Rodrigues wrote:
I don't think this is only a router thing. What I'm talking about is just what I mentioned above. Keep the DHCP info for all interfaces, change resolv.conf et al with the data from the first interface. If that interface *link* goes down (cable unplugged), just pick the data from the next interface and apply it. The first one is always the boss, if its link goes up, its settings get reapplied.
I don't think this is anything esoteric, we already have something keeping an eye on the interface, otherwise we would get no DHCP request when the link comes back up. It just has to be changed so that the whatever daemon that is monitoring the interface that we consider the primary applies the settings for the next interface when it sees its interface going down.
Note that I'm not talking about the default gateway going down or some failure like that. That's the reason I brought the wireless scenario. The fact is our windows users can unplug their laptops from the wired LAN (the "staff" network) and move around the building without having to restart their network interfaces (the wireless network is part of the "students" LAN, a wide-open insecure separate LAN). Our Linux users can't.
We use the combination of ifplugd & waproamd (http://www.stud.uni-hamburg.de/users/lennart/projects/) to make Linux wireless not suck. Check them out.
-- Aaron Bennett UNIX Administrator Franklin W. Olin College of Engineering
