[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Summary of Aston

In message <m0oNTtu-000EvOC@csrj.crn.cogs.susx.ac.uk> you write:

>Glyphs are being allocated to math encodings on the grounds of:
>\item[Glyph shape.] All glyphs of a similar design should be in the same
>   encoding.  For example, all the Greek glyphs should live together.
>\item[Kerning.] Any glyphs which may need to have a kern pair should be
>   in the same encoding.  For example, one common request is for kerning
>   between `$f$' and `$($', and so these glyphs should live together.
>   (The situation is somewhat more complex than this, since \TeX\ will
>   only kern or ligature when the first glyph is a math atom consisting
>   only of a single-character mathord.  See Rule~14 of Appendix G of \TB\
>   for more details.)
>\item[Ligaturing.] Any glyphs which may need to ligature should be in
>   the same encoding.  For example, some fonts do not have a separate
>   \psnon{mapstochar} glyph, and will instead use a ligature between
>   \psnon{mapstochar} and \psnon{rightarrow} to produce \psnon{mapsto}.
>\item[Orthogonality.] Each encoding should use as few different glyph
>   styles as possible, to minimize the number of virtual fonts needed.
>   For example, the Computer Modern Symbol encoding includes roman
>   glyphs, geometric symbols, calligraphic letters, and dingbats, and so
>   a different VF is required for each combination of roman, geometric,
>   calligraphic and dingbat font.  A site with 100~text romans, four
>   geometric symbol fonts, three calligraphic fonts, and three dingbat
>   fonts would need $100 \times 4 \times 3 \times 3 = 3600$ VFs.
>\item[Slots.] Some glyphs have preferred slots, for example it would be
>   useful if the letter `A' was always in slot~65.

All we've got there is absolutely correct. But...

All those five reasons are technical.

All that interests the user is that the encoding works. The other main
user interest is that there are very few families used, so that he can
bang in the other families whatever extra stuff he needs.  One of the
things he would like to do is to be able to get rid of the black letters
(old german) if he knows he is not going to use them in a given document,
and use that family for something else. In doing so, he does not want to
loose to much other usefull stuff, or commonly used stuff.

We are not doing this encoding so that it is easy to make VFs. If VFs are
needed, even if they are hard and long to make, people will make them. A
good ilustration of that atitude, is that people use TeX. Why do they use
TeX? because they need TeX, and tex produces good results, but TeX in not
exactly easy to use. For an encoding to produce good results, it must be
efficient, and flexible. It will be flexible if it can be used together
with many other fonts.

Kerning and ligaturing, is something we can't avoid, slot positions are no
problem (most of the time).

Shapes and orthogonality still need discussing to my mind. If we give
these criteria to much importance, we will end up doing something that is
technically good, but not really what the user wants.  Of course the best
would be to combine the two. But I think that if compromises need to be
made, they must be made on the technical side of things, and not to the
user's disadvantage.

It would be usefull if people gave their point of view on this.

Justin Ziegler

PS: does every other user play around with virtual fonts ?

PPS: Any other suggestions for glyph grouping ?