>This system keeps the categories more straightforward, and pretty well avoids the sort of subtle bias Wikipedia >has >been caught with here. Defining the precise intersection of interest is up to the user.

But the corresponding weakness is that it depends on the editors hitting all the right categories to work properly (as well as the tool itself, which as heavy toolserver users know is not always the case). Someone may categorize in two of three but not the third (guess which one might get forgotten?)

Compare it to the weaknesses of the current category system. 98% of editors don't know what they are doing. Categories and subcategories are applied inconsistently all the time. Nobody has an overview of the entire tree structure, or even a major branch of it. Something that is a subcategory of American novelists today may stop being one tomorrow, just by dint of a single edit, and no one would be the wiser (unless they keep hundreds of categories on their watchlist). The category tree (or weave, as categories can have several parents) changes daily, with categories created, renamed, recategorised, and deleted. There are incessant arguments about how to name, categorise and diffuse categories, and about perceived iniquities. Wiki-gnomes spend days working and undoing each other's work. It's insane.

Using a defined set of basic tags in combination with something like CatScan – ported across to the Foundation server if you like, and given a friendly front-end with shortcuts to the most common searches – would do away with that.