open 80211s on XO 1.5

[Sameer Verma]

On Fri, Oct 30, 2009 at 11:50 AM, Ed McNierney <ed at laptop.org> wrote:
> On Oct 30, 2009, at 3:32 PM, Sameer Verma wrote:
>
>>  scenarios of a handful of XOs in the under-a-tree model
>
> Sameer -
>
> Under a tree, using mesh networking is pointless (unless, I suppose, it is
> an extraordinarily large tree).  Mesh networking allows packet forwarding
> from node A to node B, where such nodes cannot normally communicate with one
> another directly.  Packets are forwarded through node C, visible to both A
> and B, or through multiple such intermediate nodes.  If A can communicate
> with B, mesh is neither helpful nor advisable.  It just confuses things,
> which is the problem we see with large numbers of children in a classroom.
>  The mesh efforts to keep track of how to get from A to B can quickly
> saturate the RF spectrum with a lot of unhelpful traffic.
>

Hi Ed,
I've given this some thought and have come up with this. Let me know
if this makes sense. Its more like talking to myself, but it helps to
voice it out.

We have the TCP/IP layer model, somewhat modified, as follows. I've
expanded the subnet layer to Data Link and Physical as in the OSI
model for more detail:

--------------------
Application (5)
--------------------
Transport (4)
--------------------
Network (3)
--------------------
Data Link (2)
--------------------
Physical (1)
--------------------

A) The Sugar business and gabble/salut collaboration happens in the
Application layer, oblivious to how the network is set up. In a
network where clients are somehow visible to each other (same subnet,
vLAN, etc.), we simply use salut, and if not, we use gabble via a
Jabber/XMPP server, also at the Application layer. How the network
layer handles addressing is something that Sugar and Telepathy do not
care about.

B) We could choose to use 802.3 (Ethernet) or 802.11a/b/g/n (Wi-Fi)
and the collaboration would still work. For 802.3 (Ethernet), we would
use a USB dongle. For 802.11a/b/g/n (Wi-Fi) we would use an
appropriate Wireless Access Point. These would network using a star
topology at the Physical layer. We typically use DHCP and DNS for this
setup, where DHCP and DNS servers run on the AP or on the school
server. http://en.wikipedia.org/wiki/Network_topology#Star

With XO-1, the radio would also use 802.11s in mesh topology and in
combination with link local IP addresses and mDNS, where the volume of
mDNS traffic would drown the real traffic as the number of nodes
increase. http://en.wikipedia.org/wiki/Mesh_networking

In the "Under a tree" model, we want a small number of XOs to somehow
connect at layer 2 and address each other at layer 3. What you are
saying is that 802.11s is not the only way to achieve this because
ad-hoc 802.11a/b/g/n will allow for one hop visibility, and in the
"under a tree" model, we really won't be seeing two hops, and hence
don't need any hopping at layer 2 (802.11s). So, if the extent of
hopping is 1 hop, and we still use link local for IP addressing, and
as long as the Sugar guys can solve the issue of icons and picking an
ad-hoc name easily, we can achieve "under a tree" collaboration with
ad-hoc at layer 2.

If all this is correct, then my interest, which was in seeing if the
Marvell chip will do open80211s, was only to see if open80211s could
be enabled some time in the future (even if it is only to showcase
something cool).

cheers,
Sameer
-- 
Dr. Sameer Verma, Ph.D.
Associate Professor, Information Systems
Director, Center for Business Solutions
San Francisco State University
http://verma.sfsu.edu/
http://cbs.sfsu.edu/
http://is.sfsu.edu/


> I can't tell what it is you're doing at your meetings when your users "all
> use mesh".  At a typical in-person meeting, you have a number of people
> using XOs all in the same room.  Any XO in the room can communicate over
> WiFi directly with every other machine in the room (except in extremely
> unusual circumstances, or too many attendees wearing their tinfoil hats).
>  There's no need for or value to mesh network - A doesn't need C to forward
> packets to B because A can see B directly as another ad hoc node.
>
> If there's an AP providing routing to the Internet or other external
> networks, there's no mesh required there, either, presuming that each XO can
> communicate with the AP directly.
>
> I can't answer your question about whether those scenarios use ad hoc
> networking because I don't quite see what it is the users are doing in those
> scenarios.  What (lowercase) activity are users engaged in when you say they
> "all use mesh"?  What do you think they would be unable to do if they all
> stopped using mesh?  Thanks for the info.
>
>        - Ed
>