On 19 Oct 2001, Juliusz Chroboczek wrote:

Hi Juliusz,

Thank you for your answer.

> JS>   When used under MS-Windows, glyphs for US-ASCII part (or ISO-8859-1
> JS> part) are about twice as narrow as glyphs for Japanese
> JS> characters. However, the width of glyphs for US-ASCII part is the same
> JS> as that of glyphs for Japanese characters when it is presented to X11
> JS> clients by freetype backend of XFree86 4.x. That is, the width of US-ASCII
> JS> glyphs  is abnormally wide making text rendered with them very ugly.

JC> This is the expected behaviour.

JC> If you put a `-c-' entry in your fonts.dir, the freetype backend will
JC> forcibly cast all your glyphs to the same width.

 No doubt this is expected when the last two fields of XLFD
is iso10646-1. That's why I wrote about subsetting support  so that users
can remove some code range in iso10646-1 fonts derived from CJK TTFs.
However, if the last two fields of XLFD is something like iso8859-1,
wouldn't it better to follow some kind of 'wcswidth convention' per
'character set+encoding' basis instead of 'ind. chars basis'?


  BTW, for this reason, I had to remove iso10646-1 entry in XLC_LOCALE
definition of ko_KR.UTF-8. Otherwise, all the gnome applications requiring
fixed-width fonts  running under ko_KR.UTF-8 pick up iso10646-1 fonts with
twice-as-wide-as-expected glyphs for Latin characters. However, this means
that I cannot use Hangul precomposed syllables outside the repertoire of
KS X 1001 (KS C 5601) *even* in ko_KR.UTF-8 locale unless ksc5601.1992-3
is supported by Xlib which will certainly bloat Xlib, which I don't
think is desirable unless there's no way around. On the other hand,
support for Big5 is already there so that adding ksc5601.1992-3 support
might not be such a big deal... (I have little idea about this last point)

JC> If you want to use a variable width font, you should be using `-p-'
JC> instead.

  However, there are cases where '-c-' is necessary and users
expect this dual-width convention. Of course, I'm not saying this has to
be done *within* a single (as seen by X11 clients) font which is against
the very definition of character cell font, but it appears to make sense
that *two or more* 'logical' fonts with different values in the last
two fields of XLFD presented to X11 clients derived from a *single* CJK
TTF  have *fixed* but different widths.

  Suppose we have the following entries in fonts.dir

watanabe-mincho.ttf -watanabe-mincho-medium-r-normal--0-0-0-0-c-0-iso8859-1
watanabe-mincho.ttf -watanabe-mincho-medium-r-normal--0-0-0-0-c-0-jisx0208.1983-0
watanabe-mincho.ttf -watanabe-mincho-medium-r-normal--0-0-0-0-c-0-iso10646-1

Now there's a html rendering engine which uses '-c-' fonts to render <pre>
or <tt> and a user feed it with the following html snippet.

------------
<div style="font-family: mincho" lang="ja">
<pre>
Japanese and English mixed text
</pre>
</div>
----------

Wouldn't it be natural to expect Latin characters to have half the width
of Japanese characters?

One may ask: why don't I just remove the first entry above if I don't
like the glyphs coming from that entry? I could, but more often than not,
Latin glyphs included in CJK TTFs are designed to go along better with CJK
glyphs in CJK TTFs than Latin glyphs in Latin-only TTFs (Times-Roman,
Courier, Helvetica) and users may wish those glyphs to be used.
Moreoever, users migrating from MS-Windows are used to this behavior
(when they specify 'mincho', 'song', 'ming', 'batang' as a font to use,
they get every letter/character rendered with glyphs drawn from that font)
and expect the same from X11.

  Jungshik Shin


_______________________________________________
I18n mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/i18n

Reply via email to