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

Re: Miscellany

Jeremy writes:

>   \reusefam\frakfam\myfam

I think this would be a nice extension to the NFSS2 family allocation
system, although this is slightly off the point of encodings!

>This must surely be an argument for (as I presume is the intention)
>regrouping all the symbols so that they are more logically arranged.

This is indeed one of the intentions.

>For
>example, when I use CM I use the text italic for maths identifiers, but I
>can't wheel out the math italic fonts because they contain odd symbols that
>I want to use. Knuth's encoding is not at all modular.

For multi-letter identifiers, we forsee two possibilities:

 * A document style specifies an additional T1-encoded font for use
   with multi-letter identifiers, which can then be used with
   $\mathmyfont{foo}$.

 * A document style specifies an MC-encoded font containing glyphs
   with the correct kerning, italic correction and horizontal
   escapement to form multi-letter identifiers, which can then be used
   with $foo$.

The first option is easiest on the implementor, since you just slot
in another T1 font.  The second option is much harder, since
generating a good math font is very very difficult.  It took the AMS
about one person month to produce an acceptable italic for use in
mathematics out of Autologic Times.  This hasn't got much to do with
encodings, it's about finding the optical center of the glyph top, the
appropriate super- and sub-script positions, and making sure that
$|f|$ produces a horizontally centered glyph.

Mind you, if you can live with the poor accent and script positioning,
there's nothing stopping you building a VF with cmti glyphs in the MC
encoding!

>Dispensible historical reasons, or still-relevant technical ones?

The latter.  Try
\begin{theorem}
   For infinitely many values of $a$ a pile of spaghetti can be
   $a$-automatically generated.
\end{theorem}

>This may be off topic, but anyone's observations on how to turn a text
>italic font into a suitable math italic one would be most welcome. How does
>App G's usage of fontdimens differ from the "expected" usage?

It's hard.  If I've got some spare time at some point (ho ho ho ho ho)
I'd like to try producing a tool to help with the task, that will
print appropriate test documents and ask the user to read off numbers
>from a scale.  This would produce a first run that could then be
tweaked by hand.

Appendix G uses fontdimens 1--7 in the standard way (OK, except for
\fontdimen2 which it uses as a binary flag to find out if the font is
a math font or not, see rule 17).  The main difference is that it uses
the charwd and charic values in a very odd way.  As a rough summary,
charwd is used as the horizontal escapement to the subscript position,
and charic is used as the horizontal escapement from the subscript
position to the superscript position.  For example, in the formula
$T_0^1$, there is no horizontal kerning between the T and the 0, but
ther is a charic kern between the 0 and the 1.  All very weird.

>(The right answer is presumably to spend half a
>day playing with them, but that isn't on the cards at the moment... :-( so
>if anyone has already done so, please let me know!)

The right answer is to spend closer to a month playing with them!

Alan.