I think what Alan's getting at is that the hierarchial filesystem in today's
world is no longer good enough FOR the purpose of classification and
organization.
That doesn't mean it's not USEFUL - it is very useful. The current
filesystem model is great for well-defined containers and well-defined files
that belong to one, and only one, container.
It works great, as you pointed out, for separating data files from
executable files. In fact, I can't think of a good permissions model that is
NOT based on a hierarchial filesystem.
However, things get tricky when the user has hundrends of thousands to
millions of data files. No longer can you categorize a file into a single
folder and have any expectancy of easily finding that one specific file by
browsing the filesystem tree.
The need arises for files to belong to more than one folder (in essense; I
am of course talking about metadata, or tagging files with as much info
about them as possible).
In this model, files are still part of a low-level hierarchial file system
based, which is based primarily on the need for security permissions, so
things are organized by who should have access to them.
However, once you get into a part of the filesystem that contains data
files, it is much more efficient to be able to go into "metadata mode" and
browse files by metadata, instead of forcing the users to use a
no-longer-useful "folders and files" abstraction.
I was RELLY looking forward to seeing MS implement this, and I was very
disappointed when they dumped it
. We're left with just a small fraction
of the UI that would have been.
- JB
Vista Support FAQ
http://www.jimmah.com/vista/