Re: Japanese locale on glibc-2.1.1

[Ulrich Drepper <drepper@cygnus.com>]

Hi,

I looked at the patches and have a few problems with them.

> I used These patch in Japanese locale.
> [2 localedata_sjis.patch <application/octet-stream>]
> 
> [3 localedata_jis.patch <application/octet-stream>]
> 
> [4 localedata_ujis.patch <application/octet-stream>]

The problem with these patches is that the locale descriptions are not
abstracted from the charset.

The way the localedef etc is supposed to work is that each
character/glyph gets a unique name.  The name is the same in all
charmaps.  E.g., <j0101> will then always denote U3000 or so.  The
goal is that it is then possible to write a locale description for
*all* character set encodings.

Your approach already provides three different locale desriptions.  It
can only get more (e.g., in a UTF-8 locale).  Instead there should be
only a ja_JP locale which uses generic names.

> [5 ja_JP.ISO-2022-JP_po_mo.tar.gz <application/x-gzip>]
> 
> [6 ja_JP.SJIS_po_mo.tar.gz <application/x-gzip>]
> 
> [7 ja_JP.ujis_po_mo.tar.gz <application/x-gzip>]

It's very good that we finally get translations.  Thank you.

What is necessary to get the translations into glibc, though, is to
send them to

	translation@iro.umontreal.ca

Please do this (send only the .po files).

> [11 gconv_db.patch <application/octet-stream>]

I hope I have fixed the problem already.  The patch itself is exactly
what I want to avoid since strcasecmp is much slower than strcmp.

Thanks,

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------