I think this is a great idea, and have a simple suggestion for how to keep the S/N ratio high, while still keeping the system open and community-centric.
If you look at the success of standalone weblog software like Radio or Greymatter, a lot of the power of the system derives from the ad-hoc networks that spring up between related individuals or sites. Because there are a few simple pieces of glue that link the sites to each other (RSS, XML-RPC, HTTP), and every contributor is responsible for creating, categorizing, and linking their content into the larger system, there doesn't have to be a single leader or control system that keeps everything in check.
This makes the tools and information much more dynamic and lets new communities more or less create themselves. People can easily aggregate a number of trusted streams on the fly, and merge them into a single data set, randomly select new sources, etc., all without the designers of the system being required to babysit the entire process.
So, applying this an information space structured around a real-world geography, the primary role for each community "server" would simply be to associate links to externally-hosted and authored content with geographic locations, while the "client" would then select from these streams based on pre-established trust relationships, searches, or any other metric they desired.
Of course, since a web site would be an extremely valuable interface to this system, the client/server distinctions could become somewhat arbitrary, or even follow a hierarchy. Individual users might actually utilize some sort of portal site, which acted as a proxy to the larger network.
Dave Winer and Jon Udell have both written at length about this kind of spontaneous network architecture, and the blogging community is a good example of it being successfully applied.
Oh, and in case it wasn't obvious, I would be extremely interested in working on this, too. I even have some programming experience to back up all this wanking.