# Some comments to Re: Behaviour of ...

Lars Hellström Lars.Hellstrom@math.umu.se
Tue, 2 Jun 1998 18:38:33 +0200 (MET DST)

<greeting of your choise> everyone.

During the last week the fontinst mailing list has certainly shown a new
side of itself; months of silence---and then dozens of posts each day! For
one thing, it's probably a good way to scare off people that are new to the
list, although I suppose there's not much that can be done about it.

After having spent some hours browsing through what I found in my mailbox
today, it seems hard not to get a few things of your own to say, but as
they are not really part of that discussion, I preferred not using the same
subject line.

1. (Primarily to Rowland): It was nice to hear that someone else is using
the same hard- and software combination as I am. Sometimes it has seemed as
if the TeX world on the net is entirely populated by UNIX-users who
couldn't care less of the occational non-UNIX-user who appears, but that is
in my darker moments. Most of the time I just go ahead and write something
of my own instead of trying to adapt something someone else has written.

The trouble of the many encodings seems familiar, but I suppose I managed
to avoid most of it simply by refraining from using \latinfamily;
personally I believe that is a too high level to use fontinst from, but on
the other hand I am probably one of those people who know fontinst too good
to be able to write a good documentation.

2. Adobe Garamond was taken as an example of the problems of creating both
a basic and and expert font set:

>If you add the expert set, you not only add 6 expert fonts (8x),
>but also 2 SC and 4 OsF fonts.  Since the oldstyle digits are taken
>from the expert fonts, anyway, the OsF fonts may be ignored, but the
>SC fonts will be used.
(Ulrik Vieth at Fri, 29 May 1998 13:52:02 +0200)

As this is a font family I am quite familiar with, I should perhaps point
out that in a way there is no point in using SC fonts either, as all the
glyphs they contain can be found in the corresponding base or expert font.
(FIsmall, FLsmall, and SSsmall are exceptions, but on the other hand these
are identical to what fontinst would fake them with.) Hence, IMHO, a person
who generates metrics for CTAN for this font should take the time to
generate the VPLs so that they do not require the user to have the
AGaramond-RegularSC and AGaramond-SemiboldSC fonts. Originally, the expert
set did not include these, so there's a chance some users do not have them.
(I had the AFM for over a year before I got the LWFN.)

There is however one piece of information in the SC fonts that is useful,
namely the kerning pairs between upper and lower case letters. The
extraction of this information is a non-trivial problem since Adobe seems
to name their glyphs according to encoding, not according to what they
really are, but it ought to be easy enough to solve for the average
fontinst expert; even I managed to solve it.

3. There seems to have been a lot of minor updates to the unofficial
fontinst these last few days. Here's a suggested point of another: The
definition of \resetint is

\def\resetint#1#2{
\eval_expr{#2}
\ifnum\result<\max_mathchardef
\ifnum\result<0
\x_cs\edef{i-#1}{\the\result}
\else
\x_cs\mathchardef{i-#1}=\result
\fi
\else
\x_cs\edef{i-#1}{\the\result}
\fi
}

Since the token following \ifnum\result<0 is \x_cs, and \x_cs is
expandable, TeX will expand it looking for a possible digit. This work is
normally in vain since most numbers in fontinst are nonnegative,
furthermore \resetint is pretty heavily used. The suggested fix, however,
is simple: Change \ifnum\result<0 to \ifnum0>\result.