[Xorp-users] OLSR test results on Linux.

[Ben Greear]

On 07/09/2009 12:25 PM, Bruce Simpson wrote:
> Ben Greear wrote:
>> ...
>> Normal UDP (unicast) traffic works fine. I can't get mcast working, but
>> that may be some other problem on my end (it works fine with OSPF in the
>> same configuration)
>
> Multicasting isn't supported by XORP's OLSR implementation at the
> moment; that's likely to need kernel patches.

Ahh, but if I run pimsim beside it, it seems to work.

I had a lame bug in my previous xorp conf file that caused pim and the
other multicast daemons to not start.

Thanks,
Ben

>
> Specifically what's needed for the MANET protocols, is a way to capture
> the socket layer memberships, and the way to do that is probably to
> redirect IGMP into one of the userland daemons, and take action
> accordingly. That makes the MANET stuff relatively transparent to
> multicast apps.
>
> FreeBSD 8.0 is about to be released soon, and IGMPv3 is in there -- I
> left placeholders for the hooks needed in comments, but it does mean
> tagging an interface as being exclusively for MANET use.
>
> Tunneling, in some way, with a histogram based duplicate detection, is
> pretty much the only choice, because convergence can't be relied upon
> for loop detection in an ad-hoc network -- the topology is too dynamic.
> Look for the stuff Joe Macker's group at US NRL were doing.
>
> Also, the tie-breaks for multiple links in XORP's OLSR isn't ideal --
> it's strictly Dijkstra and the ETX stuff isn't in there. I would far
> rather the Boost refactoring happened first, before that, because it
> will streamline that code greatly. I think I left some comments in OLSR
> about Boost candidate uses.
>
> cheers,
> BMS


-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com