So, in case any of you are interested in how this played out, the IT
group has decided that we are, in fact, worthy of our own subnet.
However, before I found this out, I got halfway to solving the
problem:

gwy<-------->(de0)heimdall(de1)<-------->hideo
1.2.3.254    1.2.3.170  1.2.3.233       1.2.3.41

The netmask is supposed to be 255.255.255.0.

In order to get heimdall to talk to both gwy and hideo, I had to do
the following:

# ifconfig de0 inet 1.2.3.170 netmask 255.255.255.0
# route add default gwy
# ping -c 1 gwy ; arp gwy
# route add -host gwy -link <ether of gwy> -llinfo -iface
# ifconfig de0 delete 1.2.3.170
# ifconfig de1 inet 1.2.3.233 netmask 255.255.255.0
# ping -c 1 hideo ; arp hideo
# route add -host hideo -link <ether of hideo> -llinfo -iface
# ifconfig de1 delete 1.2.3.233
# ifconfig de0 inet 1.2.3.170 netmask 255.255.255.255
# ifconfig de1 inet 1.2.3.233 netmask 255.255.255.255

...I think that's all right.  Anyway, I could only have both
interfaces up at the same time if they had very small subnet masks --
I suppose that's very logical, in retrospect.  The trick was that, if
you have 1.2.3.233/32 configured, you can't add a route to 1.2.3.41
through that interface, even if the route includes a link-level
address.  But, if you put the route in there while the interface is
configured as 1.2.3.233/24, then it's legal.  If you tell it to be a
permanent route, it'll stay there, even if you delete that address
from that interface.

Really a gross hack, and I'm glad I'll be able to set up a real
subnet.

The next step would have been to get the arpd that a couple of people
mentioned, which should be able to do sufficiently sophisticated
proxy-ARP stuff to let these machines ARP each other.  The final step
would have been to get ipf installed on heimdall, using the fastroute
stuff to forward between the two interfaces.

Thanks for all your help, everybody!

Chris

-- 
-----------------------------------------------------cjones@math.montana.edu
Chris Jones                                          cjones@honors.montana.edu
           Mad scientist at large                    cjones@nervana.montana.edu
"Is this going to be a stand-up programming session, sir, or another bug hunt?"