Allen briggs writes:

>couldn't get much change in behavior.  The netmask and broadcast are
>correct (255.255.255.128 / 151.199.66.255).

For a 25-bit netmask (255.255.255.128), you want a broadcast address
of 151.199.66.127.

Otherwise this sounds suspiciously like an ARP botch, ad does your
earlier message:

>> It would be interesting to know what the collision rate was like.

>Nil.  I ran netstat -i 5 on the -current machine and got on the order of
>300 packets in 5 seconds with 6-10 collisions, then nothing for the next
>interval, then another 300 / 6-10, then nothing, etc...


Can you try a arp -an (or netstat -nr) from the console of the
afflicted machine during the freeze? I am especially curious to hear
wthat difference it makes if you leave a

      ping -v <correct broadcast address>
running on (a) the server, (b) the client end of the connection.

I cant place the 5sec period.  Maybe the automedia FSM has a bug, in
which case the 1Hz ping is unlikely to cause any difference.

Are these machiens connected via a switch/router, or a dumb hub?