RE: [Groff] man problem: Latin1 characters in nion-Latin1 locale

[Tomohiro KUBOTA]

Hi,

At Mon, 20 Nov 2000 10:45:28 +0300,
Andrej Borsenkow <address@hidden> wrote:

> > Not yet.  It is still hard-coded, unfortunately.  I've got a patch
> > which adds an `ascii8' device to avoid this dependency, and I will
> > eventually include that (or probably find a different solution).
> >
> 
> Yes, I came to the same idea finally.

I am glad my patch is so diffusing :-)  (I at first wrote the ascii8
patch for 'groff' package of Debian GNU/Linux distribution).

> But speaking about man pages and i18n - there is more about.

Ok, I am interested in i18n.

(BTW, please read my document on i18n in Debian Documentation Project
http://www.debian.org/doc/manuals/intro-i18n/ which is in great 
rewriting now.)


> Russian words in average are longer than English (the same applies to German,
> BTW :-) and without hyphenation display loks ugly, with several words and
> large inter-words gaps. I assume, we could find hypehnation table in TeX, just
> as for english - but here are some problems.
> 
> Currently, hyphenation table for US is uncondtitionally loaded. Is it possible
> to load more than one table, like Russian?

I can imagine a mechanism, which switches hyphenation rules and other
orthography rules by 'language tag' in the roff source.  (Please note
I am not familiar with the source code of GNU roff!  I am just talking
about a general algorithm.)


> Second, there are three widely used russian charsets and one "standard" one.
> Is it possible to access current locale in troffrc? Given possible variety of
> locale names, something like locale alias table looks reasonable. I may try to
> implement something like this.

I am now working on this problem.  Werner and I discussed on this problem
and concluded:
 - troff, the main typesetting engine, recognizes only UTF-8.
 - preprocessor and postprocessor will be supplied which convert
   encoding between encodings determined by the locale and UTF-8.

The encoding can be written in the roff source or by command line option.

However, I think the idea of 'troffrc' is not nice.  If all softwares
you are using would follow this idea, you will have to edit thousands
of 'rc' files for all softwares.  ISO C determines a nice way: 
LC_ALL, LC_CTYPE, and LANG environmental variables determine the 
encoding.

The discussion is found at:
http://ffii.org/archive/mails/groff/2000/Oct/0056.html

You can find the first version of preprocessor at:
http://ffii.org/archive/mails/groff/2000/Nov/0013.html

---
Tomohiro KUBOTA <address@hidden>
http://surfchem0.riken.go.jp/~kubota/