On Dec 8, Dr. Bill Studenmund wrote
> I've been following this thread off and on, and I've been wondering what
> is up.
> 
> As I understand it, the original poster has a local, non-routed network. 
> There are at least two machines involved. Machine A is the server, where
> the routing problems are happening. Machine B is on the other end of the
> ppp link.
> 
> The problem is that the IP address given to machine B is within the
> netmask of the ethernet, and pppd is having trouble setting up the routes
> to the other end of the link.
> 
> I don't think this is a bug in pppd, I think its a bug in the network
> configuration. I don't understand why the ppp link addresses were chosen
> within the netmask of the ethernet - they shouldn't be there! Every IP
> address within the ethernet's netmask should be reachable via the
> ethernet. That's the definition (I thought).
> 
> What I'd suggest is that the original poster pick two IP addresses outside
> the ethernet netmask and give them to the ppp link. That way there's no
> such question about routing spaces.
> [...]

What about proxy arp ? This is a well supported configuration by most
network hardware vendors, and works very well. This kind of config has many
advantages: if you have a laptop, it can be on the ethernet or connected
via ppp with the same IP adress. Also, you don't need to worry with
the routes to you remote machine. 

In any case, if we support proxy arp we have to deal with the routing tables
properly. If we stop to support proxy arp ... well, this would be bad.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--