Is there anything in the configuration that would make it think
"*.*.*.255" is in a /24 network?
We're grasping at straws here, you haven't provided enough information
to get any concise help.
I'll grasp anyway though...
To my knowledge Unix and Windows do not stop you from pinging the
broadcast ip address unless that machine (Unix or windows) knows
specifically that the ip you are trying to ping is a broadcast ip. The
only way that would really be is if the server were physically on that
network. For example, I have a redhat server on 10.0.0.0 /24. The
broadcast on that network is 10.0.0.255 and if I try to ping it I must
put in the "-b" flag to indicate that yes I do want to ping the
broadcast ( I don't think windows does this or cares if you ping the
broadcast). In such a case I would get a response back from every
device on that subnet. Well, lets look at another example. 10.0.0.0
/23, the broadcast there would be 10.0.1.255 not 10.0.0.255.
10.0.0.255 would be a perfectly valid ip address to use as it is in
the middle of the subnet. As would be the case if you used 10.0.0.0/32
as a loopback address - there should be no problem with that as it is
a perfectly valid single ip subnet per the VLSM rules. Just the same,
as stated earlier by someone else - *.*.*.255/32 is a perfectly valid
single ip subnet that would be just fine to use and route as a
loopback address.
Are the UNIX servers you're speaking of on the same subnet as your
windows 2k servers?
What is the exact configuration of that loopback and it's
corresponding routing entry?
for example
loopback0
ip address *.*.*.255 255.255.255.255 ?
router ospf 1 <--assuming ospf
network *.*.*.255 0.0.0.0 area 0?
Also, showing a routing entry from a neighbor router might help too -
maybe you have a routing issue.
Also, if the UNIX and W2K servers are on the same subnet - compare
their configurations - make sure they are itentical (with the
exception of the actual IP assigned to each of course).
Good luck.
The Mikester