The simple fact is that a strict hierarchy is a really lousy way to organize most things, and that's all that a filesystem gives you. (Except in Unix we have links, so we can have cycles in our trees, yay.)
David Weinberger said it far better than I ever could in a lecture some years ago. Video and audio are here:
I refuse to be guilted about my "poor" file organization habits when the system I have available for organizing my files is half-assed and inadequate. Any given file may belong in an arbitrary number of buckets and it's the computer's job to keep track of that, not mine. After all, why do we have DBMS-managed indexes instead of doing it all by hand?
Personally I think a combination of arbitrary metadata and fulltext/metadata search would be the bee's knees, but I don't know of anything that does that yet -- most systems tend to do one or the other. Nepomuk is supposed to do both, if/when it gets into a usable state.