Fwd: OLPC Laptop - some suggestions for the software

[MBurns]

On 12/20/06, thomasasta at gmx.net <thomasasta at gmx.net> wrote:
> So which Instant Messenger client and which Messenger protocol will be used ?
> There is a full linux implementation if Cspace and it is working perfectly.
>
> http://www.aabdalla.com/cspace/

Last I heard, work was being based off the XMMP/Jabber protocol. See
some of. [1]

> Meshing:
> Hi David und Ivan.
>
> How does the wireless laptop get their IP-Adress?

See the Trac Ticket 19 [2]. IPv4 is able to do self-assignment (if
needed) in the 169.254.x.x space. Ideally, you would use a cleaner
solution, and it will likely be IPv6, which is able to do
self-assignment of addresses [3]. Don't know if IPv6 will make it into
the first production XO machine or not, though.

>You want to build in servers every 500 meters?

Nah, more like 1500 meters [4]. :)

> Why not a decentral auto-ip-signement and then a routing/hopping protocol from laptop antenna
> to laptop antenna until the ISP Node is found?

Yes, this is what will be done, more or less. Please realize, that the
engineers behind this project are, in fact, quite good at what they
do.

> OLSR or the follow up B.A.T.M.A.N. is quite good for that,
> and: there is as well the option to add auto-ip-signment to Batman.

Which I am sure is true, but irrelevant. As the hardware/software
debate that just happens shows, no general-purpose routing daemon,
superhero-named or otherwise, is appropriate or capable of running on
the specialized networking hardware the OLPC requires. Although
testing projects to see how BATMAN compares to the binary blob are
welcome. Competition is a good thing.

> Third, next to Batman and Meshnode, there is the new wireless WIMAX, which will provide DSL over air wireless in a distance of 50 km.

Yes. And as I understand most wireless marketing, that is a big 'under
optimal conditions with the wind to your back' scenario. The
fundamental shift away from 802.11b/g/s isn't something that is
reasonable to consider before the first production round is sent out
in June(?). After that, we can always reevaluate our assumptions.

> Why then not a network of WIMAX Providers?
802.11s itself is a as-of-yet-incomplete spec. OLPC is basing the work
around that spec. Wimax adds another complex layer of beta work on
top. Also, as mentioned above, it seems late in the game to switch
pitchers.

> No, the mesh routing is software. It's software which runs on the ARM
> CPU in the wireless chip rather than on the host CPU, but it's software
> nonetheless.
>
> A proper open-source version of that firmware would be very nice.

No doubt. [5]

[1] http://wiki.laptop.org/go/Instant_messaging_challenges
[2] http://dev.laptop.org/ticket/19
[3] http://en.wikipedia.org/wiki/Ipv6#Stateless_autoconfiguration_of_hosts
[4] http://mailman.laptop.org/pipermail/devel/2006-December/003296.html
[5] http://dev.laptop.org/ticket/46

-- 
Michael Burns * Security Student
NET * Oregon State University