groff vs jgroff (Re: man-db locale support)

[Fumitoshi UKAI <ukai@debian.or.jp>]

At 01 Oct 1999 00:14:21 +0900,
Changwoo Ryu <cwryu@debian.org> wrote:

> > > As for groff, here we usually use latin1.  I do not know whether it's nice or
> > > not: it just works.
> > 
> > I should have misse dsomething, as they are competely different
> > charsets. maybe you mean that you just excenge fonts and that's all?
> 
> If his terminal program can display koi8-r encoded text, there's no
> problem with latin1; latin1 device is 8-bit clean.  But printing them
> is another game.  The Japanese support before was for printing (and
> perhaps for SJIS encoding?) AFAIK.

latin1 might be 8-bit clean, but it's not multibyte friendly.

For examples, output of `LANG=ja man ls | od -c' (with jgroff) looks like
followings, and it looks fine.

0000000  \n  \n  \n   L   S   (   1   L   )                            
0000020                                                                
*
0000060                                                           L   S
0000100   (   1   L   )  \n  \n  \n 314 276  \b  \b 314 276 301 260  \b
0000120  \b 301 260  \n 
 (snip)

Bold characters are represented as 
 [1st byte of char] [2nd byte of char] \b \b [1st byte of char] [2nd byte of char]

However, in case of groff, it looks like as follows, and we see broken chars...

0000000  \n  \n  \n   L   S   (   1   L   )                            
0000020                                                                
*
0000060                                                           L   S
0000100   (   1   L   )  \n  \n  \n 314  \b 314 276  \b 276 301  \b 301
0000120 260  \b 260  \n   
 (snip)

Bold characters are represented as
 [1st byte of char] \b [1st byte of char] [2nd byte of char] \b [2nd byte of char]
and it's bad way for EUC-JP.

AFAIK, jgroff only supports EUC-JP and doesn't works with Shift-JIS encodings.

Regards,
Fumitoshi UKAI