To everyone who responded, thanks.  I'm going to short-
circuit answering all of the questions people asked because other 
problems have crept up.

	First off, I had the broadcast address and netmask reversed.
I was working from memory and wasn't thinking much about what I was
doing.  A broadcast address doesn't look much like what a netmask
address should look like.

	On one machine (ceres), when I ifconfig ep0 to set the
address, etc., I get an error message: "/netbsd: stray interrupt 7".
The card is at IRQ 10, and there's nothing listed at IRQ 7.  (I guess
that's to be expected if it's a stray interrupt.)

	On both machines, when the interfaces are listed as UP and
RUNNING, ping will get me the error message: "arpresolve: can't 
allocate llinfo".  I looked into the code (/usr/src/sys/net/if_ether.c,
if I remember correctly), but I'm pretty inexperienced at looking at 
'real' code.  I believe the code snippet from arpresolve() was:

if (la == 0 || rt == 0) {
	/* Syslog the error and return. */
}

	Am I correct in assuming that rt being NULL indicates that
there's no routing information?

	Here's another interesting bit:  I tried to ping the broadcast 
address from the other machine (callisto).  I got replies from callisto,
but nothing from ceres.  I did notice out of the corner of my eye, 
though, that ceres spit out the arpresolve error message.  Perhaps
something got there, but it had no idea how to route it back.

	For those who are interested, here are the IP addresses, et al.:

	ceres		128.169.94.88
	callisto	128.169.94.6
	netmask		255.255.252.0
	broadcast	128.169.95.255

	If someone has any hints as to what's going on, please let me
know.  All help is greatly appreciated.

Chad Mynhier <mynhier@cs.utk.edu>
Sorcerer's Apprentice, CS Department        
University of Tennessee, Knoxville                   
WWW: http://www.cs.utk.edu/~mynhier/