Just thinking out loud here, but are we using our photographs in Commons to maximum advantage?
Reason I ask is that there is more and more geographic tagging of pictures starting to happen (e.g. http://www.panoramio.com/ ). If we take a large repository of images like Commons, and add geographic tags (long + lat) to those images (maybe with a bot based on the that articles that include them, and that in turn have geographic tags included), then you could very quickly build a very large and powerful repository of images, that are tied to specific points on the globe .... which could be, y'know, be kinda cool, because then you could potentially search for images on Commons based on location, and it could come back with a whole list of images, in progressively increasing distance from the point that you just specified.
So using London as an example, say you search for a point near Big Ben (by clicking on a map or something), and you'd get back pictures of Big Ben, then the Houses of Parliament, and the Thames River, then the London Eye, and so forth - so you would get an immersive image search, perhaps in some ways more powerful than what we have currently.
But that's just half of it - because then you could do it the other way around too, and you could add a geographic tag to articles, and then it could suggest some pictures that you might want to add to your article, because they're close to the point you just specified. And the best bit: No language or translation problems - because Longitude and Latitude are universal, language-neutral things - so it would help _every_ Wikipedia.
Of course, to do this stuff, it would be nice to have some sort of structure to specifying Long + Lat, possibly something more low-level than just a template, because you'd want it to be readily and quickly searchable & indexable. And if you're going to do it for images, maybe it should be done for articles too - that way you could say "find me other articles that are about somewhere close to the article that I am currently looking at", and it could work automatically, without having to manually add links. _Maybe_ you'd be looking at extra columns in the database, which gets tricky, because it's extra baggage that might not relevant to some wikis or some articles (although it is _very_ relevant to many articles in an encyclopedia, in my opinion, and it's exceedingly relevant to images because every image was taken somewhere on earth, or looking at something on earth, apart from the 0.0001% taken in space exploration).
This starts to get bogged down in some of the questions related to Semantic wikis, but I'm thinking of something much simpler - longitude + latitude only.
Thoughts?
-- All the best, Nick.