-- Entered by Zach 5/10/2004 --

ping 10.0.255.255 from a non-atheros node on a network containing an
atheros nodes results in replies from all IPs except the atheros nodes. 

ping 10.0.255.255 from an atheros node on a network containing
non-atheros nodes and at least one other atheros node results in
replies ONLY from IPs asigned to atheros nodes.

Hmm. Further testing suggests that perhaps the problem is simply that
the atheros cards don't respond to ALL the pings on a broadcast
ping. I now suspect that the real problem here is collisions. The
conditions I was observing before seem to have disappeared and are
replaced by a fairly random success rate in getting responses to
broadcast pings from atheros nodes. It still seems like the
non-atheros nodes are more reliable.

--

It seems like the ARP issues attributed to the RADIX/MPATH issue might
be exacerbated by having atheros nodes on the net. Even with antennas
connected to all nodes we seem to be seeing this issue crop up
semi-regularly. Maybe it is related to the above?

May 10 01:57:53 cuw /netbsd: arplookup: unable to enter address for 10.0.9.142@01 on (null) (host is not on local network)
May 10 01:57:53 cuw /netbsd: arpresolve: can't allocate llinfo on ath0 for 10.0.9.142

--
Fri Jul  2 14:33:19 CDT 2004

I think both these issues are explained by broken rate-negotiation.
Our Atheros radios probably detect the Lucent radios speaking DSSS,
and try to speak back to them in OFDM.  I know what needs to be
done to fix this.
--