I have 2-3 thousand installed programs. Not counting Microsoft and not
counting Mozilla, only 3 of those thousands of programs are using the
"application data" directory. FORTUNATELY.
I suspect - but do not know - that MS in creating that directory was
trying - however incompetently - to create a place to put "read-only
application data" or something a la the /usr directory tree in UNIX.
Problem was, they didn't implement UNIX's ability to put any directory
on any partition or device. And then they compounded the problem by
only allowing Windows to install on the first partition of the first
hard drive.
I think Microsoft designs systems by first making a list of all the
things they WON'T allow you to do - and whatever is left over is
considered the "feature list" - which most people equate with the "bug
list."
And some of those rules I do not agree with. For instance, MSFT tells
developers it is imperative they never use a local ini for user settings.
This is one where I'm very glad so many developers ignore MSFT.
Yes, the Registry is a flawed device. IBM's AIX uses something like
it, but I don't know the details. Config files are so much safer and
easier to use - at least compared to the Windows Registry. I can
envision a more robust and easier to use Registry which might be
better than text-based config files. But the reason Linux is so
highly configurable and much easier to repair or work around when
there is a problem is the use of text config files - at least when you
can find documentation for the format (which in the case of the
standard stuff is easy to find.)
Without even looking, I know I can believe you that MSDN has some
documents sitting out there instructing to use the "application data"
directory. Good reason is another matter. It is /highly inconvenient/
for me not to have a program's data in its directory. It further gives
my whole directory structure a completely unnatural bump and knot, without
justice.
Check out the Linux File System Hierarchy here:
http://www.pathname.com/fhs/. This explains what should be the
"standard" way of doing things on Linux. The reasoning behind the
separation of and content of the various directory trees might be
interesting for you. Data should be separate from a program's
executables and configuration and support files. But not so separate
that you can't even find it and can't specify where it should go -
which is what Windows does.
Heh - and they say WIndows is easier to use than Linux. I guess
"Grandma" could figure this one out in Windows (see my argument in
another thread in this newsgroup). LOL.
