[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#160284: ITP: po-debconf -- Manage translated Debconf templates files with gettext



Hi,

At Fri, 13 Sep 2002 00:50:36 +0200,
Marcin Owsiany wrote:

> $ du -sh /usr/lib/gconv
> 4.4M	/usr/lib/gconv
> $ dpkg -S /usr/lib/gconv 
> libc6: /usr/lib/gconv
> 
> Aah! I always thought this was in locales package.

A good news!  Recently I tested using iconv(1) with renaming
/usr/share/i18n/charmaps (in "locales" package) to something
else and I found iconv(1) works well.


> :  @specials = qw(C en_US); # anything else?
> :  $locale = $ENV{'LC_CTYPE'} || $ENV{'LANG'};

Exactly speaking, what we need to know is not locale name but
encoding name.  For example, if we assume that debconf templates
will be stored in UTF-8, encoding conversion will be needed for
all non-UTF-8 locales.  Current version of debconf stores templates
in popular encoding for each language (EUC-JP for Japanese, KOI8-R
for Russian,...) and encoding conversion will be needed when
the language is used with different encodings (UTF-8 for Japanese,
ISO-8859-5 for Russian,...).

In short, we need encoding information (ISO-8859-1, EUC-JP, KOI8-R,
UTF-8,...) in the locale, not language (en, ja, fr, ru,...) nor
locale (en_US, ja_JP.eucJP, fr_FR@euro,...).

This is achieved by nl_langinfo(CODESET) in C language or
"/usr/bin/locale charmap" in shell script.  Now, since debconf
is written in Perl, the following code will work well.

    open(HANDLE, "/usr/bin/locale charmap|");
    $localeencoding = <HANDLE>;
    close(HANDLE);


Next problem is to determine whether we need encoding conversion
or not.

> :  if (grep $locale =~ /^$_/ @specials) {

If we will use UTF-8 for storing all debconf templates, then
the test will "$localeencoding eq 'UTF-8'".  However, if in the
current situation, we will need a table of encodings for languages.


    %table = {
         "ja" => "EUC-JP",
         "ru" => "KOI8-R",
         ....
    };
    $language = $ENV{'LC_ALL'} || $ENV{'LC_MESSAGES'} || $ENV{'LANG'};
    $language =~ s/_.*//;
    $templateencoding = $table{$language};
    if ($localeencoding ne $templateencoding) {
         NEED_CONVERSION;
    }

However, if we migrate to gettext, then we don't need all of such
codes because gettext does this automatically.  Is it possible?

---
Tomohiro KUBOTA <kubota@debian.org>
http://www.debian.or.jp/~kubota/
"Introduction to I18N"  http://www.debian.org/doc/manuals/intro-i18n/



Reply to: