Re: Extent of localisation customisation

[Danilo Segan <danilo gnome org>]

Today at 17:49, Jody Goldberg wrote:

> The notion of creating a locale on the fly scares the bejezus out of
> me on several levels.

I was not thinking of that -- that's very sucky for many reasons
besides those you mention.  I was thinking of helping users get their
changed locales included in GNU libc (eg. if there's a large group of
Gnome users who use ru_RU locales, yet a dot for decimal separator
instead of comma, with Gnome being part of GNU, perhaps we could help
push for ru_RU decimaldot;  of course, I don't presuppose that
ru_RU decimaldot would really have that many users, it's just an
example).

>     - There is code that uses the locale name (eg mapping to MS
>       Office formats, or manually looking at translations in xml)
>       As a result there would be trouble naming the modified locales

Well, since XML uses incompatible "locale" naming (IANA registry,
RFC3066 as well methinks), it's really hard to map that to POSIX
locale names.

>     - There are so many configurable elements in a locale that the
>       ui would either be a mess or have the potential for over
>       simplifiying.  Is the user specifying the number format for
>       LC_NUMERIC or LC_MONETARY ?

Most "authorities" (i.e. everyone who manages locale-based data) keep
concepts of "regular numbers" (LC_NUMERIC) and currencies
(LC_MONETARY) separate.  I don't know if that's good or bad, but if
we had a mechanism to adjust it, it should handle both.

> Before entering this swamp I'd like to see some convincing arguments
> that users realllly want/need to create their own locale.

I think there're no arguments for that.  I only meant to say that we
might encourage users to get their locales into our "default" platform
(and with that being GNU, into GNU libc).


The other option is to create several "generic" locales which contain
all the usual variants of date/time, number and currency formats, and
put that in GNU libc.  But I suppose that's how l10n capplet works
now, right (except it uses some locales which are fairly stable)?

Cheers,
Danilo