I propose a new file type - .umd - Universal Metadata. This file mirrors every regular file stored on your hard drive. If you have a file named Photo1.jpg there is a Photo1.jpg.umd file that is created by the operating system in the same directory (but invisible). The purpose of this file is simply
define the metadata characteristics of the file - with the user being able to define new fields. For example - the .umd file could contain mp3 tags, EXIF data for pictures and other such metadata. It works very simply - defining a field, and the value of that field. For example Artist/Bruce Springstein.
Here is where it gets powerful... Through your GUI file browser (Windows Explorer or Finder or Unix equivalent), when in LIST view, you can opt to view Metadata in columns (perhaps from a pulldown list of all the fields defined by all the .umd files). So you can list all the files by artist or photographer or any of the created fields. This would give file browsers the power that programs like iTunes and iPhoto have, but across platforms and applications.
You could also create a new field and value easily - right in your file browser - say perhaps the date of a photograph - create a new "column" - name it, and start typing (or batch changing). All the added metadata would be saved in .umd files. Having a universal format would eliminate the need for embedded tags that are limited to pre-selected fields. The amount of fields would be limitless (although I understand that in theory, only a given number of different fields could be displayed per folder, but most people place similar files together in folders.
Let's say you had a folder full of pictures of your DVD covers (you either scanned them or got them off the internet). You LIST the folder and have a whole bunch of movie names - add a column for Rating and you can start adding stars to your movies. Add a column for Lent To, and you can keep track of whom you have leant movies out to. The filesystem becomes a database itself.