confused with locale names...

[Andre Charbonneau]

Greetings,

I'm confused with the way locale names are handled on my linux system 
and maybe someone can help me clarify things a bit.
For example, if I use localedef to create a new locale using the 
following command:

localedef -f ISO-8859-15 -i en_CA --force en_CA.ISO-8859-15

the locale directory created is:

/usr/lib/locale/en_CA.iso885915

But if I want the new locale to be located in my home directory, I use 
the following command:

localedef -f ISO-8859-15 -i en_CA --force ~/en_CA.ISO-8859-15

and the following directory is created:

~/en_CA.ISO-8859-15

Note how localedef is doing some kind of normalization on the output 
directory name.  If I set my system to use the newly created system-wide 
locale using the following environment I get no problems.

export LANG=en_CA.iso885915
export LC_ALL=en_CA.iso885915

But if I tell the system to use the user-wide locale using the following 
environement setup, X applications give error messages saying that the 
locale is not supported.

export LOCPATH=$HOME
export LANG=en_CA.ISO-8859-15
export LC_ALL=en_CA.ISO-8859-15

First I thought that the problem was with X not recognizing the 
ISO-8859-15 charset name, so I renamed the user-level locale to 
en_CA.iso885915, changed the environment appropriately but I still get 
the X error messages.

So now I'm very confused, because it seems to work with system wide 
locales (even though en_CA.iso885915 does not appear in X's locale.alias 
and locale.dir files) but does not work for user-wide locales with the 
LOCPATH env variable set.


Now, if I simply compile the new locale and do not explicitely specify 
the character set in the locale name, like the following:

localedef -f ISO-8859-15 -i en_CA --force ~/en_CA

export LOCPATH=$HOME
export LANG=en_CA
export LC_ALL=en_CA

X does not complain anymore.  But is it because X is still using the 
ISO-8859-1 en_CA locale behind the scenes???

I would really like some clarifications on what is happening and what 
can be done about it.

My second question is the following:
Is the charset part of a locale name used by applications in any way or 
is it only used to differentiate it from other locales compiled with 
different charset?  For example, in the last scenario, the en_CA locale 
under my home directory was created using the ISO-8859-15 character set, 
but the locale name is simply en_CA (instead of en_CA.ISO-8859-15). 
Will this cause any problems on my system, or is the charset determined 
by applications at run-time from the actual content of the locale 
instead of by its name?  In other words, when you compile a locale with 
a character set that is not the default charset for that locale, do you 
HAVE to specify the charset in the locale name for the system to be 
properly localized?

Any feedback will be greatly appreciated.


Thanks a lot,
Andre.