New, more realistic multi-hop network testbed

[Polychronis Ypodimatopoulos]

Wad,


John Watlington wrote:
> I would like to point out that both Marvell and Nortel have mesh testbeds
> (100 and 50 nodes, respectedly).    Our problem has been that no-one has
> a collaboration testbed.

A collaboration test is all I did in the past 
(http://wiki.laptop.org/go/Simple_mesh_test_(Cerebro) ) and all I 
proposed for the future. If you want to do a collaboration test the 
right way you _must_ measure your stack's overhead against the number of 
participating nodes, improve your design and implementation and start 
over. And this should be done not with 2 or 3, but with 10s of laptops. 
To the best of my knowledge, this has never been done; OLPC is only 
passively testing (with 10s of laptops but does not contribute to the 
implementation of telepathy) what Collabora is "guessing" that will work 
(no real world tests done by Collabora).


>> I must admit that it is rather hard for OLPC to act on such results. 
>> It may be for lack of resources, but I'm speaking for myself when I 
>> say that OLPC has a hard time trusting developers unless they're on 
>> its payroll, especially for core parts of its software (with the 
>> exception of Marco? ;-). I think commitment, communication and 
>> roadmaps is the answer to this problem.
>
> Lack of resources and QA testing.
> As Martin put it, we can't take Linus' approach of putting stuff
> into our build and letting the customer test it.    If we put it in
> our build, it should have been tested, tested, tested.

Wad, no matter how much you test the current collaboration stack, it 
will just never work like you'd like it to. Why is this so hard to 
understand? And you've already taken Linus' approach - we've seen 
reports from Uruguay from teachers failing to initiate collaboration, 
even in 6 groups of 2 laptops 
(http://lists.laptop.org/pipermail/devel/2008-May/013921.html).

p.