[XeTeX] XETEX cannot access OpenType features in PUA?

[mskala at ansuz.sooke.bc.ca]

On Sun, 27 Nov 2011, Aleksandr Andreev wrote:
> It appears that XeTeX colors are handled by inserting pdfliteral nodes
> around colored items, which breaks the access to GPOS.
>
> Unless there's been some work on this issue since 2007, it appears
> that I will need to look for a different way of typesetting my
> document.

I bet you'll see similar effects on other systems, too.  I'm actually
surprised that Firefox seems to work the way you want it to; I've seen
(for instance, in the page at
http://www.uni-graz.at/~katzer/korean_hangul_unicode.html - right now I'm
working on implementing conjoining behaviour of Korean letters via GSUB)
HTML markup being used explicitly for the purpose of breaking up glyph
sequences that would otherwise be subject to Unicode conjoining
behaviour, and I would expect it to usually have that effect.  If there's
a "style" of any kind applied to one glyph and not to the previous glyph,
there's good chance that software in general, not just XeTeX, will break
the glyph sequence and not see substitution rules that would apply to the
glyphs if they were really adjacent.  If it's important to be able to
style one glyph and not another, then I think you're asking for trouble to
also depend on a substitution rule matching them both.

Many people have had similar problems with the fact that software (in
general, not just XeTeX) usually doesn't treat "space" as a glyph even if
you define such a glyph in your font; as a result, substitution rules that
try to match a "space" glyph (for instance, to recognize starts and ends
of words, or to carry state-machine information across word boundaries for
things like pseudorandom alternate-form substitution) rarely work.  You
can find some discussions of that by searching on the Typophile forums;
some places where someone would want a space glyph for substitution can be
implemented using "ignore" rules instead, but others seem not to be
reliably doable in OpenType at all.

Part of the OpenType base/mark concept seems to be that a mark becomes
part of the glyph is it marking (it can be seen as a special way of
implementing, internally to the font, what would otherwise be a single
precomposed glyph), so if it is part of the semantics of your musical
"language" that the marks are separate objects and can be individually
selected for the purposes of things like coloring, having them be marks
may not be ideal.

Can what you're trying to do be implemented by zero-width negative-bearing
glyphs instead?  Folks on Typophile seem to turn up their noses at those,
but I've had good results with them and they seem to be more reliable than
mark-to-base.

-- 
Matthew Skala
mskala at ansuz.sooke.bc.ca                 People before principles.
http://ansuz.sooke.bc.ca/