On Sun, 11 Apr 1999, Paul Newhouse wrote:

> /etc/ifconfig.de0:
>    inet 205.219.89.45 netmask 255.255.255.248 broadcast 205.219.89.47
> 
> /etc/ifconfig.de1:
>    inet 205.219.89.46 205.219.89.41 netmask 255.255.255.255
:
> As soon as I do:
> 
>    ping <some remote address>
> 
> I get the route for .41 changed to:
> 
>    205.219.89.41      link#1             UHL         0        0      -  de0                <<<<<<<<
> 
> I can't get to any address on the DSL link beyond .46.

That netmask means that all of 205.219.89.41-6 are on the same subnet.
That what you have... and 8 ip _subnet_, not simply 8 ips. You may be
able to force it if you leave of the netmask, and enter the route for
each machine by hand. A more natural way would be to buy another ip
from your isp.

If you intend to let the NetBSD box do nat, you still need to be aware
that de0 and de1 define two different subnets. So each ip of the
205.219.89.40/29 would be aliased to de0, and each box on de1 gets a
private ip, including the router (yecch). Since you have enough ips,
you could then map each external ip to a single external one, without
portmaping. IMO, this is all wrong for you.

OTH, why not just leave the isp's set-up alone, and put the netbsd box
on the 205.219.89.40/29 subnet. Then it would only need two
interfaces, ipforwarding turned on, and no nat. You only have to do
something with the routing metric to make sure everyone in the office
isn't going through your cable modem link. Say de0 gets
205.219.89.46/29, which is also the default route. Then try `route
change en0 metric 100' and see if that takes.