On Tue, Sep 4, 2012 at 4:00 PM, James B. Byrne <byrnejb at harte-lyne.ca> wrote: > >> That should happen directly without C's involvement if the netmask is >> 255.255.0.0 on A and B's eth1 interfaces. > > It is not. The netmask on those interfaces is 255.255.255.0. Netmasks apply to (and describe) connected subnets, not individual interfaces. Linux will sort-of sometimes work with mismatched subnet masks but some things won't see arp broadcasts with the wrong broadcast address (which again is for the whole subnet). >> >>> Instead it goes to Eth0 on C where it dies as one would >>> expect. >> >> Why does C have both internet and LAN addresses on the same >> interfaces? >> > > I am experimenting to see if this arrangement is workable. I want to > know if it is possible to have two separate 192.168.x subnets on the > same network. Some things might work sometimes. You can overlay separate subnets on the same wire, each with a correct subnet mask, and a designated router between them, but random things will happen with mixed netmasks. > Why? I do not have a purpose in mind. I am just > checking out whether it can work or not. You would probably be better off using VLANs than overlays in any case. > If it is impossible then then I will discover why that is so, which I > think will be useful in itself. The broadcast address for a subnet is tied to the bits in the subnet mask, and ethernets need arp broadcasts to work. -- Les Mikesell lesmikesell at gmail.com