[femto]

What I meant by a native implementation is one with no underlying infrastructure. Freenet (protocol) would be running directly on the hardware. Set up a mesh of (media agnostic) point-to-point links and use the Freenet protocol to route traffic and turn them into a network. A big caveat is that Freenet might not scale as required.

One problem you are going to face if you stick to an IP based network is: who is going to administer the address space? The same is true of any centralised function. In the "glory" days of community WiFi networks, most fell by the wayside because of this issue.

[wmf]

who is going to administer the address space?

How about IPv6 geographic addresses? http://tools.ietf.org/html/draft-hain-ipv6-geo-addr-02

[nextparadigms]

I think the ultimate goal should still be maximum decentralization. Even if there will be some "super nodes", they should still be as decentralized as possible, and without being tied to some corporation or Government. Otherwise, what's the point in even doing this.

The ultimate goal should be a close to 100% P2P "Internet". All of that may not be possible within this decade, but maybe in the future. For example the hardest part to decentralize this will be when trying to link 2 continents to each other through a true mesh network. But again, I think future technology will solve this. Until then, the best strategy is to build ultra-local mesh networks, then larger local ones, then city-wide, state-wide, and so on. Grow it gradually, like any movement.