Le mer, 31/03/2004 à 14:47 -0500, Gary L Greene Jr a écrit :
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Saturday 27 March 2004 06:11 pm, Nicolas Mailhot wrote:
> Le sam, 27/03/2004 à 17:34 -0500, Gary L Greene Jr a écrit :
> > This is a proposal for a standard to accommodate the accessibility of the
> > filesystem by end-users. We request discussion on this as a new standard.
> > The URL to get to the document is:
> >
> >
http://www.csis.gvsu.edu/~abreschm/uafhs/
>
> From a usability POW hidden files (and directories) are very bad.
> I must admit I'm a bit horrified by the number of new hidden roots
> you're suggesting to create (not only you duplicate classic / topdirs
> but you also enshrine stuff like fonts that doesnt exist on the system
> root).
I was under the impression that ~/.fonts was added as an X standard, from font
config tools.
The fontconfig people seem way too smart to keep a legacy hidden dir if
all the other dotfiles are moved to a clean subtree.
> My proposal would be simple :
> 1. use a single or at most two-three top-level dirs in ~ (etc and the
> rest)
Which others would you remove? The only one that I thought was not neccesary
was config, and that deals with your third point as it helps clear the
current clutter.
You do not have to put everything under $HOME directly. Just use a
single (hidden or not) root and put the mirriads of dirs you want to
create under it.
Hidden files are sometimes, but not always, bad. One of the
suggestions was
that a distribution should create an interface to this, just as they do to
the current installation directories.
Please do not. The reason the FHS is great is one can navigate it
without special tools or "interfaces". If you need an interface to
access your layout, that means it's an hopeless mess.
The difference between fontconfig xml and gconf xml is one was designed
to be used by humans the other hidden behind an "interface". Guess which
one can actually be used by anyone ?
The reasoning is, and please feel free
to critique, that interfaces to program installation change day by day and
are constantly improving, so a long-term standard should not be involved in
them.
File layout should not be a moving target. There will always be enough
overlooked problems for the layout to evolve over time - if the core is
not sane people won't bother with it.
Also, when do you actually look in /bin/ ? It's really not a
place to be
looked at directly.
Actually, I end up doing it pretty often. And I'm glad I can - it's much
better than a black box I could not inspect at all.
The problem is that users only have access to their home
directories, so they can't put it somewhere they won't be looking. Hence, the
hidden nature of the directories.
there would not be any hidden directory problem if all the program
data/config in $home was properly layed out instead of using a flat-tree
layout with hundreds of files and directories competting with each
other. The GNUStep, public_html, etc experiments show users are ready to
accept non-hidden data roots as long as there are not scores of them.
Cheers,
--
Nicolas Mailhot