MD>That's why ZWNL is simpler to implement than ZWL.
I understand: you were talking about a difference in terms of
changes to existing fonts and software.
MD>Of course, this is only the font support. Every other
process: searching, sorting, etc. needs to be modified. Note
that with these simple characters there is no way to
distinguish the few semantic ligatures (that actually should
not be ignored) from the bulk of the decorative ligatures
(which must be ignored), unless one uses markup or adds yet
more characters!
The semantic/non-semantic distinction seems to be pretty
significant. One option is to say that a ZWL is used *only* for
semantic ligatures, but that other ligatures are controlled
using out-of-band style information. So a font might contain
"s" + ZWL + "t" -> "st" ligature
"s" + "t" -> "st" ligature
with the former always applicable but with the latter only
applying if certain font features are enabled. In this
scenario, processes like spell checking would not ignore ZWL.
But, this has potential to end up in an ugly mess given no way
to control users from forcing ligation using ZWL in cases of
aesthetic, non-semantic ligation, with the result that spell
checks, etc. don't work as they're supposed to.
Given that you've clarified, Mark, what differences between use
of ZWL and ZWNL you had in mind, it seems to me there is still
a open question along the lines that Lloyd was meaning: Why is
there any difference between ZWL and ZWNL in terms of what
mechanisms in software are needed or in how complicated the
mechanisms are?
On 1999-12-27, John Jenkins wrote:
>There are two issues here. One is getting system software
support. The other is getting applications to take advantage
of the system software support. The latter can be an enormous
uphill battle, as our experience getting people to support
ATSUI shows.
>The former is also enormously problematical. The problem is
that the TrueType spec doesn't offer any direct support for
mapping multiple characters to single glyphs?the presumption is
that this is handled in the AAT/OpenType tables. I don't know
how OpenType libraries like UniScribe or CoolType work, but I
know enough about the guts of ATSUI to say that it would be
fairly difficult to get it to handle ZWL; ZWNL would be
comparatively simple. I would imagine that OpenType would have
similar problems. Basically ZWL would be useless except for a
plain-text exchange mechanism, and even there it would be
problematical.
As Lloyd observed, it seems that *both* ligation via ZWL and
ZWNL can be implemented using existing rendering mechanisms
(some changes may still be needed for other processes).
John?
Peter
This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:57 EDT