[OLPC Networking] MANET

[Krishna Sankar]

Folks,

	Have been reading the mesh networking posts with interest. Some random observations.

	a)	Like Ivan points out, we should treat the OLPC mesh as a special case of sensor networking, as power awareness would be the main priority. This was my first thought as I was reading thru the archives.

	b)	There are also time sync algorithms in the sensor networking world that might be applicable - but I do not think we need very accurate, monotonic time. 

	c)	I have been looking at the MANET space for sometime. At least in the case of OLPC, the problem is beter bound than many of the use cases I have encountered, due to the various constraints. 

	d)	On-demand looks more promising than the proactive ones ; I know Aaron has proposed OLSR and that is good. Would be happy to assist anyway with the Google Summer Code project.

	e)	In addition to OLSR, we also should try an on-demand protocol - like DYMO (DYMO looks promising in many ways, let us see how it develops) and then figure out the characteristics most optimum for us. 

	f)	Eventhough I think the on-demand might be less resource intensive, but it depends on the usage and the rate of change of topology. Kids being kids, I assume they will be moving a lot ... We really have to figure this one out with field testing and then abstracting the metrics.

	g)	Another thought is to use a hybrid and dynamicall change from on-demand to proactive depending on the inferred usage; depends on how much intelligent inference we can make.

	h)	Am waiting for a few cards to work on this ....

Cheers
<k/>