I wish there where 48 hours in a day (and me being able to actually use the extra 24 hours :), I could use them to hack up a proof of concept: tag oriented file storage is something I've been telling people about for a couple of years. I'm absolutely sure it would be very useful, for power users and everyone else alike.
The whole directory/file mantra is an outdated concept, it works yes, but tags can offer much more flexibility. And after all, using tags isn't that much different from creating a directory tree, in a sense one is already supplying meta data for the files stored in it: work related -> tenders -> customerName for instance.
One has to get used to the idea of describing the data we're working on and retrieving them by describing them ('tenders oct 2008' or 'tenders customerName'.)
Some tags can be auto generated like the date/time created, last consulted, filetype (text document, picture, ...). When for instance camera's get equipped with GPS by default, an automatic tag with the location can be added to every picture we store in the fs.
And it wouldn't be static, if I can only think of a limited set of tags at first and sometime later can think of more and better tags, I can always add them at that time.
Plus, my view on the data probably doesn't reflect someone else's view on it. Tags allow us to have one user describe data with one set of tags and another user describe it with another set of tags. Finally, at least digitally, no arguments on the illogical place where stuff was put by someone else, e.g. your SO :)
From using tags one can go to creating relations between objects stored and their tags.
Most modern file systems already incorporate meta data functionality. What's needed is a way to efficiently index and search the tags.
Plus of course the user interfaces to them.