Many of the requirements for such a system started to sound like solved problems: identifiers describing where data could be found (some sort of Uniform Resource Locator, perhaps?), distributed storage (like 10 million webservers?), and lastly, a Google-esque search over the data. What if the Google for geographic data is…The Google?

It’s understandable and necessary that old school GIS types get worked up about arcane projections and metadata specifications and so on. But geo data isn’t all that special — as you say, what works well for the web will work for geo too.

One concern with having Google index the web’s geo data is that they will never make their index available. That’s true, but perhaps there’s the real opportunity for an OSGeo-backed initiative: a geo-centric crawling effort with a publicly accessible index.