Peter T. Daniels wrote:

>Mark E. Shoulson wrote:
>>For some nice Hebrew with TeX, see Yannis Haralambous' Tiqwah system for
>>typesetting Biblical Hebrew in TeX, and the version of it in use at
>> (see their Bibles Repository and the
>>"" version, tnk.pdf. It's a big file, though). Generated by
>>Tiqwah through TeX, with every last vowel-point and accent accounted
>>for--and when there do appear conflicts, they twiddle the code a little
>>and fix it. Not much fault to be found with their typesetting.
>You're talking about typesetting. I'm talking about typography. (Type
Well, the type design of the Tiqwah font is also pretty impressive.
Haralambous wrote some interesting stuff about Hebrew type design in
passing in
(check the section about the fonts, starting on p. 4), about how some
features of the alphabet seem to morph in one direction and then *back*
as the point-size increases, etc... I'm not too wild about the
hataf-patah, though, myself.

>What you said in your last message about making all the thin and thick
>strokes _exactly_ the same would seem to be a major part of the program.
>That's not how eyes and lettershapes interact, even if engineers would
>like it to be.
Certainly, you have to take optical effects into account. Thick straight
strokes and thick (parts of) curved strokes need to be different
weights. However, a normal alphabet has enough similar parts that it can
benefit from enforcing conformity among *similar* shapes. All the
vertical stems, for example, probably ought to be the same thickness (i,
l, k, d, b, p, h, etc... possibly also m, n, and others, or at least
those should be similar to each other; depends on the particular font's
design). All *similar* curves should look the same... Curved lines need
to overshoot the baseline; similar-looking curves (o, c, d, ...) should
overshoot the baseline by the same amount, and so forth.

It makes a lot of sense... just not quite so much sense as Knuth would
wish it to make. His Computer Modern can be generated in many different
vastly distinct styles (sans, serif, bold...) by changing a bunch of
parameters in the header file, and leaving the actual drawing parts
unchanged. It's pretty neat, but doesn't (in my experience) generalize
terribly well. Maybe I just don't have the patience to put in the effort