Mesh Potatos and OLPCs?

[C. Scott Ananian]

On Wed, Mar 23, 2011 at 9:30 PM, Peter Robinson <pbrobinson at gmail.com> wrote:
> On Wed, Mar 23, 2011 at 11:38 PM, John Gilmore <gnu at toad.com> wrote:
> Meraki are also doing mesh related things with the APs etc.
>
> Its my understanding (not that I've had much time to play) that mesh
> has improved greatly over the last couple of years with developments
> like BATMAN which is now in the 2.6.38 mainline kernel. I suspect pure
> IPv6 might also assist with some of the problems seen with broadcast
> on mesh networks as they get larger simply due to its lack of it.

IPv6 isn't a magic bullet for mesh.  If anything, IPv6's reliance on
multicast for basic ARP-type stuff adds additional difficulties.

That's not to say that IPv6+mesh is impossible.  I'm just saying it
doesn't make things significantly *easier*.

I agree with Ed's contention in so far as I also haven't seen anyone
doing exactly what OLPC originally wanted to do.  I also agree with
Peter and Charles that Mesh has matured greatly in the past few years
(although the mature deployements are not exactly OLPC's use case).

I think the opportunity is there for either (a) someone to take the
mature meshes and push them the extra distance to work in OLPC's case
(mobile nodes, ideally IPv6, dynamic mesh configuration), or (b) OLPC
to rethink its deployments such that the mature mesh technologies are
more immediately applicable.  As one strawman proposal for (b), you
might consider making the school servers the mesh nodes, not the XOs.
As another strawman proposal for (b), you could "flip a switch" to
make an XO into a dedicated stationary mesh node -- and put it on a
pole in the middle of the neighborhood with a solar panel, say.
Either of these would be significantly different from the original
"every child's machine also a mobile mesh node" vision -- but might be
much closer to the mature deployed mesh technologies.
 --scott

-- 
                         ( http://cscott.net/ )