A file in a typical file system resides within a hierarchy of directories, a tree with directories as branches and files as leaves.  Each directory has a name unique within its parent directory, so a file's location can be specified with a sequence of directory names.

To get a web of files, we can turn directory names into keywords and associate them with files.  Deferring issues related to the significance of the keyword order, we can imagine a file system UI that looks like this:

keywords: [Incoming, Win32, Blog]
========================================
  Name …
—————————————-
  blogx2.zip
> newscrawler13.zip
  Syndirella_20030216-src.zip
========================================
related: [commercial, Win32, Blog, …]
website: http://www.newzcrawler.com
…

Top section is used to view and edit a list of keywords defining the view.  Middle section list items related to all the keywords listed above.  Bottom section shows information about items selected in the middle section.  Keywords related to the selected item is also displayed.  Much of information about an item can be retrieved lazily from the web.  Note that listed files may be stored anywhere on the disk as long as they have ancestor directories named 'Incoming', 'Win32', and 'Blog' in any order.

At this point, its already useful.  However, it would be nice if the same UI can simulate traditional file system browsers, allowing the user to navigate using hierarchically.  More on this later.