[OLSR-users] Bellman-Ford instead of Dijkstra?

[Benjamin Henrion]

Thomas Lopatic <(spam-protected)> [060621]:
> Hey Benjamin,
> 
> Thanks for your clarification. Now I get it. Looks like there really
> isn't any way around implementing Bellman-Ford for your scenario. Hmmm.
> But then again, maybe we could turn the edges with weight -2 into edges
> with weight 0 and tell Dijkstra to always take the 0-weight edge from a
> given node. To be unambiguous this would restrict us to one 0-weight
> edge per node.

One of the problem with Bellman-Ford is the appearance of loops, but it
seems that you could modify the algo to get rid of those. Google for
Bellman+Ford+loop or Bellman+Ford+loop+free, there are some papers
whitch talks about howto remove loops.

> But then again, moving to Bellman-Ford shouldn't be that hard. We could
> keep most of the code in lq_route.c, I guess. Anyone? :-)

I could consider sponsoring a bit the guy who has time to do it.

I am not an expert in C, in fact I hate it :-)

--
Benjamin Henrion <(spam-protected)>
FFII Brussels - +32-484-566109 - +32-2-4148403