[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

[Freedombox-discuss] We do need mesh networking



On Saturday, 19 February 2011 17:42:45 Luca Dionisi wrote:
> On Sat, Feb 19, 2011 at 7:49 PM, Michael Blizek
> 
> <michi1 at michaelblizek.twilightparadox.com> wrote:
> > I have not seen *any* ready-to-use of meshing so far that is very
> > scaleable. In order to scale, you basically have to limit the depth of
> > routes you discover and the length of routes you take. Otherwise in
> > average, everybody needs to forward more data compared to the amount of
> > data end nodes can send/receive.
> > 
> > How do you intend to do this in netsukuku?
> 
> Good question.
> For a better explanation I invite you to read the docs on the site.
> 
> Netsukuku uses a reactive protocol, and a path vector algorithm [1],
> hence the data you need to "announce" and memorize for each
> destination is even bigger. The key mechanism to reduce this data is
> to use a hierarchical topology of the network. Nodes are grouped in a
> so called g-node, g-nodes are grouped in g-g-nodes, and so on. In
> analogy of IP classes of addresses.
> So the real problem that netsukuku had to solve (we did it) is to
> assign IP addresses that are consistent with this hierarchy and
> dynamically react to changes in the topology.
> 
> --Luca
> 
> [1] http://en.wikipedia.org/wiki/Path_vector_protocol
> 
> _______________________________________________
> Freedombox-discuss mailing list
> Freedombox-discuss at lists.alioth.debian.org
> http://lists.alioth.debian.org/mailman/listinfo/freedombox-discuss

I don't know if it's been mentionned elsewhere on the list yet, but it appears 
there are ongoing efforts to get mesh networking into the kernel. 

Jonathan Corbet has a write up at: 

http://lwn.net/Articles/426396/

Thought it might be useful.

-- 


Philippe

------
The trouble with common sense is that it is so uncommon.
<Anonymous>



Reply to: