er, someone's comment made me realize that I forgot critical
information. let me try again...

My network has several parts. There's an internal network
(10.34.0.0/16), and a routable network (66.93.137.160/27). Connecting
these is a netbsd 1.5.2 box. There is also a small (/30) network
connecting the netbsd box to a cisco. The cisco routes the routable
block over to the rest of the world, and the netbsd box does generic
nat and routing. It looks something like:

   (internet cloud)
          |
          |
   [public address]
     cisco router
 [private 192 address]
          |
          |
 [fxp1 private 192 address]
     netbsd box [fxp0 66.93.137.161] ---- 66.93.137.160/27
   [ex0 10.34.0.1]
          |
          |
 internal net10 network
 

Right now, I'm routing between 66.93.137.160/27 and 10.34.0.0/16, and
nat'ing everything that goes out fxp1. I'd like to not nat things that
go to the cisco's private 192 address. I don't see how I can do this,
as packets that go to it, go out the same interface as packets to the
rest of the net. Is there a clean way to do what I want?

for reference, here's my ipnat.conf:

bash-2.05# more /etc/ipnat.conf 
map fxp1 from 10.34.0.0/16 to any -> 66.93.137.161/32 proxy port ftp ftp/tcp
map fxp1 from 10.34.0.0/16 to any -> 66.93.137.161/32 portmap tcp/udp auto
map fxp1 from 10.34.0.0/16 to any -> 66.93.137.161/32


As I'm not subscribed to netbsd-help, please cc me on any
replies. thanks

seph