I'm not expert in this area; I'm going from memory from some
interactions with Andy Lippman's group and Michial Bletsas here in the
Media Lab.

On Wed, 2005-12-14 at 02:56 -0600, David Young wrote:
> Here are my scattered thoughts about the $100 Laptop, NetBSD, and mesh
> networking.
> 
> I have not been able to pinpoint the goals of the $100 Laptop project's
> "meshes".  Is it for connectivity in a classroom, throughout a town,
> or cross-country?  

Classroom/village/town, not cross country.  There are existing projects
that have successfully deployed a variety of low cost cross country and
satellite links.

> Is it for web browsing, P2P messaging and/or phone?

All of the above.  Voice is currently a serious challenge.  There is
research here about why voice over more than one 802.11 hops has been
losing completely; they've just succeeded in getting it working, but are
still understanding the phenomena of why there is trouble over multiple
(even 2) hops, where horrific packet losses have been observed.

> IP radio broadcast?  I ask because I am convinced that it will be
> unnecessary, and enormously difficult, to deliver the mesh networking
> "ideal" by the $100 Laptop's "due date" in 2006.

Late 2006 at earliest.  Fallback would be hostap.

>   The ideal, to my
> mind, is peer-to-peer broadband IP transport on a metro-scale radio
> relay network.  But that is probably not what the $100 Laptop project
> has in mind, though.  I am interested to hear more specifics.

We certainly want to go PtP as much as possible; there is no guarantee
of infrastructure where these machines will go, and want everything to
auto-configure.  The big issue is service location right now....  Dave
Reed's started poking at Avahi.
> 
> I've worked on a "mesh" network in Urbana, Illinois, for a few years now.
> Our small group's achievements are rather modest, but we have built a
> good foundation for mesh experimentation, both in open-source software
> (NetBSD improvements, routing software), and in the town (~32 rooftop
> wireless routers).  Our project has been funded, in part, by the Open
> Society Institute (OSI), which is interested in meshes as communications
> infrastructure for the developing world.

Yes, there is also a rooftop net around Cambridge Massachusetts these
days.

The interesting challenge is to have "power aware" mesh networking; you
don't want all machines in the mesh to be burning battery.

> 
> For the networking geeks, here is the path from my house to a neighbor's
> a few blocks away; it might be illuminating to enter the street addresses
> into Google Maps:
> 
> traceroute to n308.w.washington (10.0.10.172), 64 hops max, 40 byte packets
>  1  n801.s.walnut (10.0.237.100)  10.298 ms  8.088 ms  5.390 ms
>  2  n113.w.washington (10.0.237.108)  58.029 ms  103.887 ms  7.903 ms
>  3  n308.w.washington (10.0.10.172)  88.963 ms  7.013 ms  7.576 ms
> 
> Incidentally, TCP does not like that path.  I am using it interactively,
> now, and it feels a bit like I have ssh'd to the moon....  The bad
> performance on that path is not difficult to explain, and improving the
> performance is possible, but it takes "warm bodies."

You may be seeing some of the phenomena noticed here (with voice, in
particular).  As I said, students in Andy Lippman's group have gotten
things working much, much better, but why, is not yet fully understood
(or wasn't as of last week).

They were noticing most packets getting dropped as soon as you went over
multiple hops when trying to run VOIP (kphone in particular).

Also note that there is no guarantee that the machines will follow the
standard 802.11 protocols; part of what the Athleros chip lets them do
is play other games on the air.

> 
> I grant remote access to our testbed to trusted developers who
> want to study and improve a real-world multihop ad hoc wireless
> network.  For folks who want to experiment on their own, the
> Urbana testbed is reproducible, indoors or out, with off-the-shelf
> components.  Some instructions for outdoor "nodes" are on-line,
> <http://cuwireless.net/nodeconstruction>.
> 

I gather the folks here are using local networks that have been running
around MIT and Cambridge for a while.

			Best regards,
				- Jim Gettys