The founder and chief architect of mesh networking firm Tropos adds his two cents to the "ugly truths" about mesh networking thread: Narasimha Chari has added his comments that clarify many of the problems with mesh networking that Francis daCosta raised at Daily Wireless several days ago. I linked to daCosta's original post and posted and linked to two replies on the topic already in this archived item.
Chari says that daCosta has his primary calculation wrong: instead of a bandwidth reduction in a single-radio mesh system where n is the number of hops of (1/2)^n, it's 1/n. This calculation results in an enormously higher throughput than daCosta (click slide from Nitin Vaidya that Chari references at upper right for the dropoff by hops in throughput).
Chari makes a number of additional useful points on this topic that are worth considering. Tropos has deployed through its integrator and resellers the largest number of mesh points in commercial use in the U.S., and potentially worldwide, that I'm aware of. If they don't know what they're talking about, then a lot of networks are operating by hypnosis instead of science.