[Fontinst] Kerning

Hilmar Schlegel hschlegel at ubcom.de
Tue Aug 26 11:17:21 CEST 2003


Indeed it would be nice to automate kerning a bit for those cases when
synthetic kerns are sufficient. 
The problem with \setleftkerning type of create kern pairs are as
o typically also aacute atilde, atilde adieresis, adieresis acaron &c
are generated, which simply causes a mess of practically useless pairs
o the T a vs. T acaron problem is still there (and must be cured as
Walter suggests)

The suggestion would be to have \setleftkerning restricted to
upper/lower case.
Actually some \setHleftkerning and \setxleftkerning would be needed
which work only on a restricted set of kern pairs from the template
character, e.g. \setxleftkerning would take over only the kerns with
lower case characters.

To make this work one would either build in a test (for certain
parameters) or set up some character classes (which are then selected). 

No idea if that would be too complex for an implementation in TeX. I
think low performance would be less critical than faulty kern pairs.

Anyway, some restriction to the total number of mechanically created
kern pairs would be very sensible for fonts which already provide lots
of kerns, esp. OT. To throw out kerns between accented characters has
proved to help considerably for practical purposes but this may be
language dependent.

Hilmar Schlegel

Walter Schmidt wrote:
> On Mon, 25 Aug 2003 14:21:31 +0200, Ulrich Dirr wrote:
> >Easiest would be to have some kind of table like below
> >
> >%left char(s) right char(s) value
> >T [a aacute atilde aring ...] -80
> >T [adieresis agrave acaron acircumflex ...] -30
> >-- yielding -->
> >T a -80, T aacute -80, T atilde -80, T aring -80, ...
> >T adieresis -30, T agrave -30, T acaron -30, T acircumflex -30, ...
> (1)
> Choose a "model glyph" for each class, say "a" and "adieresis"
> (2)
> Put the following commands for the "model glpyphs" in an
>   \setkern{T}{adieresis}{-30}
> (3)
> Write a "general" mtx file, which defines the kerning of
>   \setleftkerning{agrave}{adieresis}{1000}
> Walter
> _______________________________________________
> fontinst mailing list
> postmaster at tug.org
> http://tug.org/mailman/listinfo/fontinst

More information about the fontinst mailing list