On Tue, Feb 19, 2008 at 07:02:20PM +0000, Darren Kenny wrote:
> On Tue, 19 Feb 2008 11:42:03 -0600
> Nicolas Williams <Nicolas.Williams at Sun.COM> wrote:
> > > In the case of XDG user dirs, it is not a GNOME applications, but desktop
> > > independent, hence it doesn't use glib, so we need a way to configure the
> > > charset to use - that is the purpose of setting the charset at the system
> > > level.
> > 
> > I don't get this.  No matter what login program and desktop environment
> > is used there always is a user-selected locale.  Therefore, why not
> > honor the user's locale choice by taking the charset from the locale?
> 
> Because that's what causes the confusion when switching locales in the first
> place. By a desktop env consistently using the same charset for the file names
> then this eliminates one of the causes of issues when people using the same
> locale but different charset, usualy without realising it see wierd chars in
> their filenames.

Picking different locales with different charsets will lead to confusion
here no matter what.

> > > Saying that, we actually wouldn't expect this to be changed much, and 
> > > should
> > > always be utf-8.
> > 
> > Someone's bound to set this and a user is going to login using a locale
> > with a different charset, and things won't look right.
> 
> Well, I coukd put in a warning or comment against doing it, would that help?

Or guidance on when to use it.

> > 
> > That the sysadmin can set the charset here is obnoxious, but if the only
> > way to use a charset other than UTF-8 when the user selects a non-UTF-8
> > locale, then that'd be a serious bug.
> 
> How so? GNOME or KDE wont acknowledge that encoding either at the low levels 
> of
> glib, but will re-encode for presentation to the user. To use utf-8 like this
> is consistent with the current behaviour of these environments.

How will GNOME/KDE re-encode non-UTF-8 filenames?  How can they know
what charset was used for them?  Ah, I think I get it -- see below.

> > At the very least we need to support non-UTF-8 locale selections by the
> > user without having to update this file, and if that's done then why
> > bother with this system-wide setting?
>  
> But we do, it's only the file-system representation that's utf-8, the UI will
> be, if anything, more consistent in it's presentation to the user in the 
> locale
> of their choosing  -it's much easier to convert from utf-8 to a non-utf-8
> locale than from one unknown locale to the current locale.

Now I get the impression that xdg-user-dirs-update will default to using
UTF-8 no matter what the locale is.

If so, that should be in the case materials and docs, no?  And the
dependence on GUIs to a) assume UTF-8 is used for filenames, b) convert
to the user's locale's codeset, should be documented as well.

Is there precedent for GUIs/desktop environments assuming that
filesystem object names are encoded in UTF-8?  Not that I think there's
a better way to do this, but that this needs to be documented.

Nico
-- 

Reply via email to