Hi!
At this point I believe you understand why using a
mesh dymanic
routing protocol instead of BGP.
Not really. BMX6 cannot really measure any packet loss over Batman
clouds because all the packet loss is abstracted away anyway, no? So it
sees whole Batman cloud as one hop? Still, for best performance, some
information from WiFi layer below should be available to BMX6. This is
why I am not sure if BMX6 is suitable. Probably it would be easier to
adapt Batman itself to be able to do some simple L3 announcements on
border nodes? It would have access to underlying data as well.
But yes, BMX6 can probably work as well and the job. I am just unsure if
it is the best option for the given task. BMX6 was developed with some
other design parameters (running on wireless mesh, measuring packet
loss, finding the best path there) and now you are moving it to some
other problem domain. Maybe BMX7 would be more suitable. :-) BMX7 which
would talk with underlying Batman and collect real link quality
information. Because if it does not talk with it, then maybe something
like running cjdns on top of those Batman clouds to merge them together
would be more suitable? You would get crypto for free. And you would not
care about physical topology anyway.
Mitar
--
http://mitar.tnode.com/
https://twitter.com/mitar_m