(For the sake of argument, if all precomposed glyphs were abolished,
leaving NFC==NFD, then how would we store composition specializations
inside fonts...)
You have to distinguish between characters and glyphs here. The number of Unicode
characters representable with a font is different from the number of glyphs in the font.
Because as you wrote, diacritic marks for Latin/Greek/Cyrillic and other
combining characters take different shapes and different positions depending
on where they're used. The same is true of base characters The shape of
a base char. is different whether it's used alone or combined with combining
characters and how many and which combining characters it combine with.
In modern intelligent fonts like opentype fonts, char to glyph mapping is not
1 to 1 but m to n where m and n >= 1. The way this m to n mapping is
stored in fonts and accessed by rendering/layout engines varies.
(there's even a proposal to add this intelligence to old X11 BDF.)
Opentype fonts have layout tables like gsub and gpos that have to be
accessed and activated by rendering engines like Uniscribe and Pango.
The amount of intelligence in embedded opentype fonts is smaller than
that in AAT (Apple's intelligent font format) in that in the former
Uniscribe and Pango should more work than necessary for AAT fonts.
Graphite is another font format(? it uses opentype format, but
its layout tables are different from gsub/gpos and so forth used by
Pango/Uniscrbe) and rendering library pair.
For details, see http://www.microsoft.com/typography http://developers.apple.com/fonts http://www.pango.org http://graphite.sil.org and Adobe's page
Jungshik
-- Linux-UTF8: i18n of Linux on all levels Archive: http://mail.nl.linux.org/linux-utf8/
