On Monday, July 1, 2002, at 05:31 AM, Michael Everson wrote:
>> I must point out that for English (and a lot of other languages), the
>> use of ZWJ to control ligation is considered improper. The ZWJ
>> technique for requesting ligatures is intended to be limited to cases
>> where the word is spelled incorrectly if *not* ligated
> What! No! Look at my paper and the examples of Runic and Old Hungarian
> and Irish. There are examples where ligation is used on a nonce-basis,
> not having anything to do with global ligation or "correctness".
Michael, I was very careful to say "English (and a lot of other languages)
." And, by and large, the software which supports ligation doesn't compel
global on-or-off, so that nonce-bases are supported.
What's frustrating for me about this never-ending discussion is that it
always seems to come down to the stupid ct-ligature in English. I have a
book that uses it *everywhere* and it gets *really* annoying. :-(
I have sitting in front of me a reprint of a nineteenth century
reproduction of the 1611 King James Version of the Bible. It uses the "ct"
ligature in the headers, but not in the text. (It also uses the long-s,
by the way.) But if someone were to come to me and ask how you would use
plain text to reproduce this text, I'd tell them you can't, or shouldn't—t
rying to reproduce precisely the visual appearance of a text isn't a job
for plain text. Period.
I also have a font on my machine based on the handwriting of Herman Zapf.
It's a gorgeous font with a huge, idiosyncratic set of ligatures. It
doesn't make sense to have the user (or software) insert ZWJ all over the
place on the off-chance that the text will end up being set with Zapfino
to make sure that these ligatures form correctly.
Our system fonts are set, moreover, to do fi- and fl-ligature formation
automatically (well, most of them are). That's because it's the
appropriate default behavior for most Latin-based languages. (Not all. I
know that.) Where this behavior is *not* appropriate, there are
mechanisms, including the ZWJ/ZWNJ one, which can override the default
behavior. This means that file names, menus, dialog boxes, email, and so
on all do the most-nearly-correct thing without having to be told to.
>> (and similarly ZWNJ is intended to prevent ligature formation where that
>> would make the word spelled incorrectly). The kind and degree of
>> ligation in English is generally considered a sylistic issue and is best
>> left to higher-level protocols. Thus saith Unicode 3.2.
> It doesn't go so far as to say what you did. Maybe Book needs to check
> the text some on this point. We should have consensus.
No, the bit about spelling is simply my attempt to state informally the
idea that Unicode 3.2 is attempting to convey.
Let's just have a quick survey here to see if there's consensus:
1) In Latin typography, ligature formation is generally a stylistic choice.
There are exceptions, and these exceptions are more or less common
depending on the precise language being represented.
2) Where ligature formation *is* a stylistic choice, it should not be
controlled in plain text but by some sort of higher-level mechanism. Such
a mechanism should allow the default formation of ligatures with the
ability for the user to override the default behavior.
3) Where ligature formation is *not* a stylistic choice, the ZWJ/ZWNJ
mechanism is an appropriate one to provide ligation control.
4) The precise set of ligatures in a Latin typeface is design-specific. A
typeface should not be required to include a set of ligatures which do not
make aesthetic sense for the overall design.
This last point, by the way, is the one which is the big sticking point
for the large type foundries that I've spoken to.
John H. Jenkins
This archive was generated by hypermail 2.1.2 : Mon Jul 01 2002 - 10:25:44 EDT