The same comment I leave every time this subject comes up again somewhere: it's already been built, it already has lots of users, it's already secure and anonymous: http://www.i2p2.de/
I've built mesh networks. I know they can be done. I also know that, even if you know which corners to cut and which equipment to use, they can be expensive and challenging to build out to any reasonable size. Plus, you still have the limited-hop issue, so any network resources that you want to access at any reasonable amount of bandwidth have to be cached at most just a couple of hops away.
From what I understand, Tor is an anonymizing service for web browsing (and other web activities). I2P is an anonymizing layer above the regular internet. I.e., you can use Tor to view pages/use services on the regular internet, while with I2P you can only use services that are part of I2P. On the other hand, I2P is less vulnerable - it is fully decentralized & distributed, with everyone being an equal peer in the network, while Tor relies on exit nodes to function, and normal clients don't help the network at all.
I've never used I2P, but Tor also has hidden services, which can only be accessed through the network and do not require exit nodes. I imagine both Tor and I2P use similar routing schemes to achieve "anonymity".
http://www.torproject.org/docs/hidden-services.html.en
I am not too trusting of the architecture though, since there has been (at least) one significant practical break before. Also the hidden services are susceptible to DoS, despite the claims that to DoS one node you have to DoS the entire network.
Sorry for the long comment. I've been involved in several "anonymous data routing" projects ranging from wireless mesh networks to route-over-internet to sneakernet projects, in areas where authorities opposed deployments either moderately (vandalism / hacking the network) or severely (you get shot).
The tl;dr is that security/anonymity is hard and it comes at the cost of: money, speed, configuration, user experience. Only in really rare circumstances (oppressive regime or similar) will people choose to use a system that costs more, is slower, and requires programming experience to configure and install.
As I understand it, their plan is to build a wireless mesh network, which sounds great, but it flat-out isn't feasible for this scale. Unless things have changed significantly since I was working with mesh, the cost to route a packet across a largish , dense (best-case) city is millions of times more expensive than the existing infrastructure, and dozens or hundreds of times slower. maybe you could do a little bit better using non-FCC-licensed tech, but obviously the FCC won't like that, and now you're way outside what Newegg will ship to your door to hack with.
Cost and speed are irrelevant if the mesh network is competing against no internet or shoot-you-in-the-head-if-you-read-bbc internet. If it's competing against Comcast 20Mbit, though, it's dead in the water.
If you insist on not relying on the existing infrastructure a much better approach would just be to run cable everywhere; that's worked out great in developing countries. But here you have zoning boards and HOAs and such, so good luck. Successful projects in developed countries are generally neighborhoods or municipalities, and even then, good luck.
The next option is a sneakernet; physical geographical dispersion of flash drives or similar; routing software to instruct people to physically move packets along routes; mail can be used for longer routes. They're cheap to build: flash drives and lockboxes cost nothing, they can be totally invisible to others. The bandwidth rivals current fiber installations; I've built sneakernets that routinely move 50TB per day. The trouble is latency, nobody wants to wait a week for their packet to arrive. If the incentives are there (get in big trouble for accessing certain information), it can work, but you have to offer a COMPELLING reason for a user to routinely walk away from the computer and route a packet.
If you don't like any of those, you're stuck with something like tor/WASTE/I2P2, i.e. leveraging the existing infrastructure. Even then, ordinary people are never going to use any of those projects, the pressure isn't there.
The only thing I can imagine that would be feasible on a wide scale in the US (barring a huge Orwellian apocalypse that would make the Patriot Act look like a birthday present) would be along the lines of replacing DNS and using HTTPS more. But people have been trying to do that for ages. You or I with the benefit of hindsight can sit down and write a better DNS in ten minutes. In fact, lots of people have, but we still use DNS and HTTP just the same. That's because a more-secure internet isn't at all a matter of technical will--just a matter of will.
