> JIS X 0213 has many characters which are also included in JIS X 0212.
> It is very confusing. I guess JIS people think JIS X 0212 is
> obsolete.
Basically, only Emacs supports JIS X 0212...
> A few characters in JIS X 0213 are not included in the present
> Unicode.
AFAIK, this will be fixed (or have already been fixed?) in the next
Unicode release where more than 10000 CJK characters are added (in the
surrogate area).
> > > Then the 'font definition file' will be irrationally large.
> >
> > Right. I think I've answered this problem in my last mail (regarding
> > a `glyphclass' directive in font description files).
>
> Then all of these glyphs have to have the same width.
Why? It is intended that `glyphclass' can occur multiple times. Say,
one glyphclass command for full-width glyphs, and another one for
half-width glyphs.
> > Indeed, the default behaviour should be that the preprocessor adds
> > a
> >
> > .mso tmac.<charset>
> >
> > line or something similar to the document, but there must be a
> > possibility to override it manually.
>
> Good idea. Thus '#ifdef I18N' part can be restricted in pre/post-
> processors.
Exactly.
> Overriding? Well, the current Groff has '-a' option. I think this
> can be used for this purpose. (Anyway, we can provide substitution
> only for non-letter symbols like soft-hyphen, '(C)', circles,
> squares, and so on. I think this is sufficient.)
The `-a' option is almost useless today IMHO. It will show a tty
approximation of the typeset output:
groff -a -man -Tdvi troff.man | less
It is *not* the right way to quickly select an ASCII device. To
override the used macros for the output character set we need a new
option.
Using `-a' is comparable to dvi2tty or similar converters.
> We have to think about uniting my idea on design of preprocessor
> and Ukai's idea of '.encoding "encoding-name"' in roff source.
>
> - it is the preprocessor that handles the ".encoding" .
> - priority is that
> * --input-encodings wins.
> * .encoding is next.
> * then falling into the default (locale-sensible for i18n OS
> and latin-1 for non-i18n OS).
Exactly. Compare this to the Emacs model of `local variables'.
Note that such an encoding request has to determine the encoding *and*
character set of a document (similar to Emacs).
I suggest that we don't use `.encoding' but
-*- charset-encoding: xxx -*-
in the first comment block (almost similar to Emacs). troff shouldn't
notice encoding issues at all and just accept UTF-8.
If really necessary, we can add two additional commands to select
encoding and character separatedly:
-*- charset: ...; encoding: ... -*-
Examples:
.\" -*- charset: JIS-X-0208; encoding: EUC -*-
.\" -*- charset: JIS-X-0208; encoding: ISO-2022 -*-
Werner
