[GRASS translations] Re: [GRASS5] french translation for GRASS-5.7 CVS

[Glynn Clements]

Markus Neteler wrote:

> > 3. UTF-8 isn't a good choice for message catalogs. Much of GRASS won't
> > handle multi-byte encodings correctly (e.g. curses assumes 1 byte == 1
> > character == 1 cell). I suggest using single-byte encodings such as
> > ISO-8859-1 instead (FWIW, Alex' Russian translation is in KOI8-R).
> > 
> > E.g. regarding the French translation for the message in point 2:
> > 
> > 	#: ../lib/gis/history.c:154
> > 	#, c-format
> > 	msgid "generated by %s"
> > 	msgstr "généré par %s"
> > 
> > If this text is written to a terminal which doesn't use UTF-8 (and
> > UTF-8 terminals are still in the minority), it will appear as garbage;
> > e.g. if the terminal uses ISO-8859-1, "généré" will appear as
> > "généré".
> 
> What's the recommendation for the upcoming Asian support?

Whatever we do, support for "wide" character sets (more than 256
"characters") is likely to be a lot of work. In particular, anything
which uses curses will need significant changes. Similarly, other code
which tries to "align" output (e.g. r.info) may require changes.

In that situation, there isn't much to choose between the different
encodings; they all require multiple bytes per character, and you
can't assume that all characters have the same width on the terminal.

It's really down to local conventions. Except where a particular
program insists on a specific encoding (a lot of free/open-source
software forces the use of UTF-8), Japanese uses EUC-JP or Shift-JIS,
Chinese uses GB-2312 or Big5 and Korean uses EUC-KR. Any of these may
be used directly or wrapped in ISO-2022.

-- 
Glynn Clements <glynn.clements at virgin.net>