[Battlemesh] Battlemesh v5 tests

[Markus Kittenberger]

btw i agree (with marek) that features not available in all protocols can
still make sense to be tested,..

but if only available in one protocol, hmmm its more like a presentation,
not a comparition/battle,..

but there is nothing bad about presenting/testing unique features,..

>
> > I didn't talk about moving NODES, but I talked about moving CLIENTS.
> >
> > E.g. in your ninux network could be a normal user which, for some
> reason, moves
> > from an Access Point to the next one (imagine two close houses with two
> > different ninux APs).
>
> AFAIK the other protocols (or at least their open source
> implementations) don't have (yet) this feature, so I don't think is the
> case to include it in the "official" tests.
>

hmm about what feature are we exactly talking?

as "usually" (as for for olsrd (or likely all layer 3) meshes ) APs run
somewhat outside of the mesh network,.. (and clients connected to them are
usually not running mesh-software,..)

i.e. the APs are just attached to a meshnode (or if they are physicall part
of it they are still logically seperated from the mesh)
and typically the AP NATs its clients.
-> so if a client roams from one to another AP his connenctions are for
sure dead,..
(and he usually also needs to get a new dhcp lease, and so on)

so i guess we are talking about clients NOT getting nated, and the protocol
immediately announcing such moved clients,..
(which imho is something i would really NOT consider something a layer3
mesh/network should deal with,..)

but in case of a layer2 mesh protocol things are different/easier (-;

btw. there are for sure several methods to allow such client-roaming
between APs connected to a layer 3 mesh network,..
(and maybe some could benefit from integrating/interfacing with the routing
protocol)

Markus

p.s. i recently help(ed) designing/building/operating a somewhat
compareable AP network with currently about 300 APs. (and designed for
thousands, but still imho nothing fancy)
It uses layer2 vpn and a central server, so not really very elegant (from a
mesh network perspective)
But it roams "perfecty/instantly" and as clients are only interested in
reaching internet, the central server=uplink matches this quite well.

and hmm if there is interest, i guess i could setup such a AP network on
top of any layer 3 mesh, fairly fast.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ml.ninux.org/pipermail/battlemesh/attachments/20120312/f49e646d/attachment-0001.html>