OSDev.org

I think for certain types of file storage this kind of system makes sense. Specifically in cases where the concept of a "unique filename" is not helpful or particularly well-suited to the way data is organized...

For instance, when you have a large collection of photos, often the relevant information about the image is simply the group it belongs to, where it belongs with relation to others in the group, and any information about what's in the photo... there may not be a "title" for the photo - the relevant identifying information about the file is simply that it exists as part of a certain collection (plus any descriptive information the user might be able to add...)

Of course, that's useful, too - having a simple, definite, completely unambiguous way to refer to a file... The only problem is that you then have to rename the file (creating another arbitrary filename) if you move it to another directory where that filename already exists...

So I think for photo albums and other cases where there's no especially useful basis for a "filename", this kind of idea makes sense. Other files on the system (like the OS binaries, installed software, etc.) would probably be better off with a regular hierarchy...

A good idea.

Each file has 1 or many attributes (= tags) that the user can search.
One the attributes is location, like, /home/leptos/Music/Jazz/...
But the extension and the filename are not enough.
The other attributes may be: Marsalis, trumpet, Lincoln Center, and so on. We could then google on the entire machine.