[Mesh] From libre-mesh

Mitar mitar at tnode.com
Wed Jul 31 19:50:59 PDT 2013


Hi!

> Also, probably most of these links will be WiFi so having a protocol
> which bases its metric in the packet loss would be a good idea.

No, they will not be WiFi. They will be some abstracted away L2
LAN-look-alike. They will have no idea about underlying WiFi links.

> I don't know how deep is your knowledge about BMX6 but it is *NOT*
> just designed for WiFi links. In our networks with bmx6 we are mixing
> wired and wireless.

What I am saying is that it is designed with assumptions of direct
possibly lossy links (wireless or wired) and not with assumption of some
strange blob of multiple nodes which behave like one giant L2 LAN blob
where you send a packet and it just works and is once routed over one
path over physical links and another time over another packet.

> During the WCW Axel, Simon, Guido, Nico, Giohack and me were talking
> about a small daemon named two-faces which will share metrics between
> Batman-adv and BMX6 to be able to get more profit from this
> dual-protocol setup. But for the moment it is not implemented so we
> have to deal which what we have now.

Yes. I think more we will peer between our networks over wireless in
general, more a need for such a daemon will be. We are getting to the
point where this is really needed. So not just for peering smaller
clouds, but for peering whole networks together.

> I'm going in vacations for some weeks, so I don't think I will be able
> to reply more mails.

No problem. You were really helpful with explanations and making me
better understand your ideas and plans. Thank you very much!


Mitar

-- 
http://mitar.tnode.com/
https://twitter.com/mitar_m



More information about the mesh mailing list