Mesh networking is completely useless for providing anything other than last-ditch service. It originated in a) academia where it's a nice thing to tinker with and write PhD theses and b) in the military where having a network connection no matter how crappy can make the difference between life and death. It has no place imho in delivering any kind of service that someone would use day-to-day and certainly not one that anyone would pay for.

WISP networks are built with point-to-point links: the appropriate technology for delivering good QoS. Since radios and antennas are extremely cheap this is the way to go.

The rock you'll flounder upon is that you need to buy a pretty beefy connection to serve any number of customers (even 1). The cost of said connection in most places is quite high. This places a lower bound on the number of paying customers you need to break even. That number is somewhere around 100. This is before you factor in the cost of customer support, schlepping around on people's roofs, finding and negotiating leases on repeater locations, etc. The number of places where there is not a better competitive offering available to customers (or a high likelihood one will show up before you recoup your investment) is low.

Current wireless technology (and according to Shannon, any wireless technology we can imagine) has quite a low upper bound to throughput -- much lower than is achievable with coax for example. So you can't complete on speed with anyone except DSL and Satellite.