Re: Translations of folder names - two proposals

[Havoc Pennington <hp redhat com>]

On Sat, 2004-12-11 at 16:36 +0000, Alan Cox wrote:
> On Sad, 2004-12-11 at 17:19, Havoc Pennington wrote:
> > You're saying "the right UI decision is to expose the implementation
> > model" and as a categorical guideline that is indefensible (it leads to
> > horribly designed software). Thus, if we have our priorities straight we
> > will not accept this guideline.
> 
> The right UI is to have a working implementation that doesn't play
> games. I've seen so many "fake unix on DOS" "fake unix on VMS" type
> things and they always break in the end. 

OK, so explain how your argument doesn't apply to the following items
which are also inconsistent with shell/motif:
 - file hierarchy is displayed as multi-rooted in file chooser
 - the Computer folder in nautilus
 - making mounted devices appear on the desktop, not in "ls ~/Desktop"
 - any URI not starting with file:/// displayed in nautilus/filechooser
 - .desktop files
 - .directory files

> > The global guideline that would be more defensible is "the
> > implementation model should be made as close as possible to the user
> > model" (rather than vice versa as you are framing it) but in this case
> > we can't change how filesystems work, unfortunately. If we could, then
> > adding a filesystem feature to support this would be ideal.
> 
> I'm still confused why "use this as my Desktop" requires file system
> support. We've managed since the late 1970's with a flat text file
> telling programs and users where to put their personal file store. Its
> translatable, its configurable and programs handle it well. Its called
> "/etc/passwd".

It doesn't require filesystem support if you let us do it in userspace,
as I am advocating. It does require filesystem support if as you claim,
the UI for the desktop has to be the same as the kernel implementation.

Havoc