OLPC does end run around IP addresses

[Mikus Grinbergs]

>> What I find interesting is that Neighborhood View at every XO shows
>> *all* other XOs (plus their names) physically attached to the ethernet.
>>  'olpc-xos' shows the non-customized XO with its eth0 (radio) IP
>> address;  the other XOs are shown with their eth1 (ethernet) IP addresses.
>>
>> My conclusion:  The XOs are recognizing each other over the ethernet,
>> despite having "non-pingable" IP address identities activated.
> 
> The 169.x.x.x subnet is reserved for link-local addresses, which is
> what these are.  They are pingable from the local link.  It's all
> standards-compliant and kosher, be not afraid.
>    http://en.wikipedia.org/wiki/Link-local_address

I'm not afraid (nor am I looking for help).  I understand whence the
169.x.x.x subnet comes from.

My point is that ALL the XOs show up in each Neighborhood View, even
though the other XOs (192.168.1..) cannot ping the non-customized XO
(169.254...), nor can the non-customized XO ping the others.

What I see the XOs doing is an "end run" around my concept of how remote
nodes are supposed to be accessed.  I believe 'ping' is behaving the
standards-compliant way (192.168.1.0/24 does not access 169.254.0.0/16,
and vice versa).  Whereas what shows up in the XO Neighborhood View (and
in 'olpc-xos') appears to ignore standards-compliance.


As I said, I am not looking for help.  I am sharing an observation,
which I believe would not occur if I were not using XOs.

mikus