Re: Icons of program

["Julian Regel" <julian white-tower demon co uk>]

> I know I'll be killed by about 200 messages agains me, but...
> I don't really like the idea of having my hard disk full of .info
> file for every icon I define on my computer.

I'll risk joining you in that statement. The .info file arrangement
also falls over when you're talking about a multi-user OS. Is the
owner of the file the only user capable of changing the .info data? If
so, then this seems to defeat the purpose of allowing users to
customise their own setings.

Does anyone know how OS/2 managed it's extended attributes (up 
to 64k of additional file information including stuff like icons and
comment fields)? IIRC for filesystems that support EAs the data was
written to the filesystem itself, but for running on other FSs (such
as FAT) it stored all information in a hidden file (I'll avoid calling
it a registry because that'll have negative connotations). 

I imagine such a system would look like this under Unix:

1) A GNOME/KDE/Whatever compliant application opens a directory 
window. The system queries the (users own) EA data file (eg 
~/.objects) to see whether there are any special properties for the
files in that directory. If so, then they are read in. 

2) The system checks the EA data file to see whether there is 
redundant data (the result of a command line "rm" for example). If so,
then it tidies itself up.

3) For files which do not have data stored in the EA data file a
standard default could be applied.

This solution would save the system from having hundreds of 
potentially broken links to files and also provide a framework for a
system object database.

Does this make sense to anyone...?  :-)

Julian
--
The Internet was not built by Microsoft. It was built
by the sort of people that are now working on Linux.
The Internet has changed the world...