Re: gettext-0.10.36 Japanese PO file

[Bruno Haible]

Iida Yosiaki writes:

>  "Content-Type: text/plain; charset=sjis\n"

> * I see that the codeset part in file name
> "ja_JP.Shift_JIS.po", IE Shift_JIS, is up to the implementor.
> So it is OK to use "Shift_JIS" for Shift JIS, but I see many
> implementations have been using "SJIS", instead of
> "Shift_JIS".
> * The value of the charset parameter "sjis" cannot be found
> in the correction of the assigned names of MIME charset:
>   ftp://ftp.isi.edu/in-notes/iana/assignments/character-sets

Indeed IANA registers this encodings under the names Shift_JIS,
MS_KANJI and csShiftJIS. but the name SJIS is

  1) more easy to spell correctly by users (no hyphen/underscore
     confusion),
  2) supported by both glibc and GNU libiconv,
  3) more customary in other Unixes (HP-UX, OSF/1).

This is why this name is allowed in PO files.

> I see that this looks just like a MIME header and is *NOT*
> exactly a MIME header.  But, as everything but "sjis" in
> that line conforms MIME, it will make many people really get
> confused, especially those who are familiar with MIME.

Good point. "Shift_JIS" is a MIME name, "SJIS" not. I think I'll
change gettext to expect "Shift_JIS" instead.

> * Name of the language in Language-Team header is missing.
> And, of cource, it must be `Japanese <address@hidden>'.

This is the responsibility of the translator.

> * In the Japanese team of the Free Translation Project, we
> already have had Japanese PO file for gettext in EUC-JP
> charset.

This is unfortunate; I didn't ask the translator whether he was part
of the team. One translation is better than none, but two are not
better than one.

> * It is mutually convertible between characters in EUC-JP
> and Shift_JIS charsets.  When the maintainer really insist
> on to support Shift_JIS PO file

Noone insists on using Shift_JIS. I took what the translator gave me,
and there is no technical reason to convert it to EUC-JP, now that
glibc-2.2 and libintl can convert on the fly.

Bruno