Personally, I think the solution that could get some kind of market traction would be to sell WiFi routers that can organize themselves into Metropolitan Area Networks. A certain portion of the units is instructed to become 'long-distance' nodes, connecting to other nodes a mile or two away, and shifting large amounts of data between them. Another portion of the units is instructed to become 'medium-distance' nodes, distributing the traffic from the long-distance nodes to devices within a few thousand foot radius, and all devices are simultaneously capable of being 'short-distance' nodes, connecting directly to end-user devices. Since it takes more resources and a better antenna to run a long-distance node, routers capable of acting as long-distance could possibly be distinct from medium-distance-capable ones, and sold at a premium, with an assurance that local devices connecting to it will get better network performance. Apart from that, devices become either long, medium or short-range nodes depending on network and geographical needs.
The market for these devices will of course get its start with piracy and porn, just like any good burgeoning internet protocol, but you can bet local TV stations, radio stations and advertisers would be all over them to get their content out without having to pay the traditional network gatekeepers. Content producers would get free access to the network if they just run a node or two.
Eventually, as the metropolitan area networks become dense enough and spread out widely enough, it will be possible to link the different cities to each-other and start to create another wide-scale network. Until that happens, though, there's enough going on in most metropolitan areas to give more than enough value to a decentralized network such as this.
I assume you haven't spent much time using mesh radio solutions - they have their uses for sure but their performance and reliability are very poor compared to traditional solutions. And even then most of them rely on backending on wired plant for longhaul.
I have not, though I am aware of it. Perhaps the planned white-space open licensing would give us the bandwidth to address this problem, and the performance of such decentralized systems could be improved. If such an improvement is possible, then something like mesh would have a good shot at market success.
In a given channel, 802.11 mesh networking provides roughly 1/7th the performance of non-mesh (and an ideal protocol only achieves 1/3) [1]; using more spectrum increases throughput but the waste is still there. Many people (especially telco lobbyists) would argue that spectrum shouldn't be wasted on such inefficient protocols.
The market for these devices will of course get its start with piracy and porn, just like any good burgeoning internet protocol, but you can bet local TV stations, radio stations and advertisers would be all over them to get their content out without having to pay the traditional network gatekeepers. Content producers would get free access to the network if they just run a node or two.
Eventually, as the metropolitan area networks become dense enough and spread out widely enough, it will be possible to link the different cities to each-other and start to create another wide-scale network. Until that happens, though, there's enough going on in most metropolitan areas to give more than enough value to a decentralized network such as this.