Message-Id: <9308061708.AA16168@sc.ZIB-Berlin.DE>
Reply-To: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
Date:         Fri, 6 Aug 93 17:55:07 BST
From: Paul Taylor <pt@DOC.IC.AC.UK>
Sender: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
To: Multiple Recipients of <LATEX-L@vm.urz.Uni-Heidelberg.de>
Subject:      \mathsf etc
Status: R


A colleague with above average LaTeX experience has just come in to my
office to ask
"how can I define a macro for a maths keyword which will appear in sans serif?".

My answer was
   \ifx\mathsf\undefined\def\mathsf#1{{\sf{#1}}}\fi
   \def\id{{\mathsf{id}}}
to which he made a gesture like a goldfish.

Why is this complicated piece of code necessary?

If his document is run through traditional LaTeX, \mathsf is not defined,
so we'd better define it. Arguably one of both of the pairs of braces
can be omitted.

Using traditional LaTeX, or NFSS with \unbracefonts@ alias nomargid.sty
alias oldlfont.sty, \mathsf behaves like traditional \sf and so needs to
be surrounded by braces.

Using AMS-LaTeX, ie NFSS with \bracefonts@ alias margid alias newlfont,
it takes an argument, so the text must be braced instead.

Either way, if he types A^\id (as may well occur, this being category theory)
the outer braces are needed to ensure that the whole thing is treated as
the superscript, as intended, and not just the \mathsf macro itself.

Now, to anticipate the arguments that the "margid" interpretation, inherited
from amstex, is more logical, let me say that I agree entirely. I also want
to state for the record, lest there be any doubt about this, that I have
a very high opinion of all of the Mainz code and design principles, and
look forward to the day when I can delete lfonts.tex latex 2.09 et al from
my disk.

However I do wish to re-iterate the serious criticism that the principles
of reverse compatibility are very important and have been disregarded.
It's all very well to say "this is a better design", but if it conflicts
with the existing documents and code a barrier is erected which effectively
prevents people from adopting the better design. I know from my own
experience (and remarks have been made on this list by others) how
reluctant many (most) users are to update their software even when it is
completely reverse compatible.  They prefer to keep the old version ---
even when it has obvious bugs like 10pt bold or AMS symbols in subscripts
and on slides --- rather than risk any incompatibilty at all. All possible
benefit of the new design or version is then completely lost, and the poor sod
who has to maintain the local software and deal with users' questions
is left with a big headache.

PLEASE Frank, Rainer, Chris, and the rest of the LaTeX 3 team, PLEASE
take note of this criticism, and PLEASE make your code reverse compatible.

You must be well aware of this problem in Germany at the moment, as you are
adopting new telephone and postal codes.  I don't know the details there,
but in Britain the practice is: implement the new code in virgin namespace,
advertise the change on a particular date, intercept calls to the old numbers
with recorded messages telling people the new numbers, and after than make
them "unobtainable". For example the long-standing 01 code for London was
replaced with 071 and 071 three years ago, and the 01 prefix is being reserved
for a further two years (until we all get a superfluous 1 inserted in our
numbers). The Germans are supposed to be better organised than the British,
so I'm sure you know how to do it!

The particular problem I mention is even now not insuperable. (The official
solution, margid/nomargid or oldlfont/newlfont is *NOT* reverse compatible,
but it is *VERY CONFUSING*.)

For compatibility,
    \mathsf and friends should have their AMS-LaTeX (margid) meaning.
    \sf etc should have their traditional LaTeX meaning (nomargid)
    \Bbb and \frak should have their amstex meaning (margid)

The only way I found to implement this without replacing huge parts of NFSS
was as follows:

\def\math@ungroup#1{%
  \let\save@math@bgroup\math@bgroup
  \let\save@math@egroup\math@egroup
  \let\math@bgroup\@empty
  \let\math@egroup\@empty
  #1%
  \let\math@bgroup\save@math@bgroup
  \let\math@egroup\save@math@egroup
}%

% do a LaTeX font change in the old way
\def\oldLaTeX@fontchange#1#2#3#4{\relax
  \ifmmode
     \math@ungroup#1%
  \else
     \fontfamily{#2}\fontseries{#3}\fontshape{#4}\selectfont
  \fi}

\def\psf{\oldLaTeX@fontchange\mathsf\sfdefault\default@series\default@shape}%
etc

You'll find a complete replacement for oldlfont.sty on these principles
in /tex/contrib/Taylor/tex/oldlfont-pt.sty
at theory.doc.ic.ac.uk (146.169.2.27)

With this change and a file full of definitions of \tenrm etc using NFSS,
version 2.10 of LaTeX using NFSS, as promised in "The New Font Family
Selection - User Interface to Standard LaTeX" could be released without
causing much trouble to users.

It would also make the lives of the people in charge of TeX at their sites
a good deal easier.

Paul Taylor

PS Chris Rowley asked me a few weeks ago to air on this list the issues which
arise from the "volonteer tasks" concerned with graphics applications.
I haven't forgotten this, but I do try to ration the amount of my time
I spend doing TeX things!