[OLSR-users] Non-OLSR ad-hoc client

[Henning Wangerin]

On Wed, 2006-06-07 at 18:19 +0200, Joe Bush wrote:
> > freifunk uses an approach where every node has a pool of ip's handed out
> > via dhcp to non-olsr-clients. These clients are then nat'ed to the ip of
> > the node.
> 
> Oh, Freifunk has the radio working simultaneous in adhoc and ap mode,
> routing traffic between the two. This saves hardware, and is an
> interesting solution.

No it still runs ad-hoc, and there is no problems running dhcp on that
link.

My dads windows-laptop (non-olsr) is just happy to get an ip from my
open network, and getting natted by the router that hands out the ip.

I'm able to walk around the area with my linux-laptop (with olsr),
hopping from on to another node. He has only connection as long as he is
within reach of the original node.

> I was more interested in the idea of having clients connect in adhoc
> mode and functioning as 'end-nodes', where olsr nodes would put these
> nodes in the routing table. The problem is that, since these end-nodes
> don't have olsr running they are essentially ignored.

Yes they are. The best and simpliest solution (with very limited
mobility) is in my view the way freifunk works.

Nodes run dhcpd, and hands out ip's that are natted to the adress of the
box.

> Don't get me wrong. I am all for the idea of having all clients
> running olsr. I even have my wired PC running it. It is just not
> always practical (PocketPC, etc). I have been thinking about a pluging
> solution, but not sure it is possible. It seems like the main source
> would have to be changed so that all connected adhoc clients are
> automatically updated in the routing table, and the olsr protocol
> takes care hops, hna and such. But I see now that it would break with
> the specification which seems to require using HELLO messages to
> determine connectability.

The HELLO messages tells everybody who is near. There is no other way to
do that.

I don't se any other way than either natting the end-clients, or sening
HNA for the entire subnet of each node.

I'm running each node with a 10.128.x.1/24 address, so I have room for
additional services. Some of my nodes rund dhcp on 10.128.192/26, and
making these ip's available to the rest of the network requeires either
natting or HNA

There is no other way to do it.

-- 
Henning Wangerin <(spam-protected)>