X-VM-v5-Data: ([nil nil nil nil nil nil nil t nil]
	[nil nil nil nil nil nil nil nil nil nil nil nil nil nil nil nil])
Date:         Fri, 23 Feb 90 21:54:00 MET
Reply-To: LaTeX-L Mailing list <LATEX-L@DHDURZ1.bitnet>
Sender: LaTeX-L Mailing list <LATEX-L@DHDURZ1.bitnet>
From: "Nico (Poppelier@Hutruu51.Bitnet)" <POPPELIER@HUTRUU51.bitnet>
Subject:      Suggestions from Utrecht (I)
To: Rainer Schoepf <BK4@DHDURZ1.uucp>
Status: R
X-Status: 
X-Keywords:                  
X-UID: 27

Sectional units; table of contents


Before I start a short explanation. The reason for sending my previous
message only to Frank was simple: I just wanted to inform him about a
reaction in more detail at a later stage. Never mind: here's what I
had in mind.

In process of digesting a LaTeX document 3 layers can be
distinguished: the document itself, the macros defined in the document
style, and the LaTeX macros, what I call the kernel. In the case of
the sectional units and the toc entries, information goes from left to
right in the following diagram

     document | style | kernel

     text    -->     -->

The kernel should pass things like numbers back to the style, so that
the macros defined there can really put the components of the text on
the page in whatever wild arrangement the designer comes up with.

In the current version information passes from left to right and
doesn't come back: the kernel decides about the sectional units and
the table of contents. In other words: commands that handle this
should be moved to the document style.


1. I think that it is clear to everyone on the list that LaTeX's
\@startsection macro is not generic enough to allow all possible wishes
a designer might have. I've got two cases to prove my point

- LaTeX standard styles: \part and \chapter are defined
  without \@startsection
- NTG styles: Victor had to re-define some latex.tex parts to
  achieve the desired result

My own work for ESP's journals and book series does not serve as
proof, since I can't show these to you, but let me assure you that I
run into the same problems as Victor did with the NTG styles. For the
last few days I have been thinking about redefining \@startsection in
the Elsevier styles. But suppose I do that: wouldn't it be better if
this forum can decide on a better design, one that can be used in the
styles of both (i) the standard set AND (ii) the NTG AND (iii) Elsevier?

My idea was the following: \@startsection takes care of two things

- embed the sectional unit heading in the vertical list, i.e. add
  spacing and appropriate penalties
- generate an entry in the table of contents (I will come back to the
  table of contents in part 2. of this message)

My version of \@startsection would have the following syntax:

\@startsection{NAME}{LEVEL}{BEFORESKIP}{UNSTAR}{STAR}{AFTERSKIP}
               *[ARG1]{ARG2}

NAME       as before
LEVEL      as before
BEFORESKIP as before, including sign convention
AFTERSKIP  as before, including sign convention
UNSTAR     a macro that creates the heading for the unstarred case
           \def\UNSTAR#1#2{ ... }
             #1: number of the sectional unit
             #2: text of sectional unit heading

STAR       a macro that creates the heading for the starred case
           \def\STAR#1{ ... }
             #1: number of the sectional unit

The designer sees to it that STAR and UNSTAR create appropriate
\hbox'es if the heading is run-in, and is a construction such as
\hangfrom for the other case.
For the time being, until someone can convince me of the opposite, I
am convinced that it is

- impossible to foresee what designers want `to do' with sectional
  unit headings -- Victor gave a few nice examples
- it is just as impossible to incorporate all possible degrees of
  freedom in one super-generic definition.

If you would try the latter though, I think TeX's limitation on the
number of parameters would become a handicap.

I find the idea outlined above attractive because of its simplicity,
and because of the similarity with the current situation for table
captions, figure captions, chapter headings and part headings, where
in all cases separate macros are defined.


2. Table of contents

Similar arguments can be given for the table of contents, list of
figures etc. Again, I think that the macros defined in the file should
receive NUMBER, TITLE and PAGENUM as arguments and do with these
components what the designer chooses. In the current approach these
three are not clearly separated, which makes it hard to typeset each
in a different font for instance.
I propose that in the new approach the kernel routines write these
components to the appropriate files, where they can be `picked up' by
the macros defined in the document style.

While I'm on the subject, I would like to suggest the addition of a
few new tools. In some books, Elsevier books but also Martin Bryan's
SGML book, you found tables of contents per chapter AS WELL AS a table
of contents in the front of the book. In the current version
\addcontentsline{toc} means: `write to \jobname.toc'. If you divide
a document in sub-documents with \include{...}, you could also write
toc entries to .toc files of the sub-documents.

O.K., that's it: end of part I.


Nico