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:         Wed, 18 Jul 90 13:38:19 CET
From: Rainer Schoepf <BK4@DHDURZ1.bitnet>
Organization: Inst. f. Theor. Physik d. Univ. Heidelberg
Subject:      Proposal for a section heading concept
To: LaTeX discussion list <LATEX-L@DHDURZ1.uucp>
Status: R
X-Status: 
X-Keywords:                  
X-UID: 199

I have an idea how a section heading should be handled.  At the moment
I'm not so much concerned about implementation or actual syntax, only
about the principal way of doing it.

When the user writes a section-type command (s)he specifies the
following information:
- the text of the heading
- automatic numbering or no numbering
- writing to toc or not, possibly with a special text for the toc.

This is then passed to the kernel, in the form, say,
  \here-starts-a-section{text}{t/f}{t/f}[extra text]
where the two t/f means either true or false (Probably one would
set switches, but I don't care about the method at the moment).

Now the kernel generates the structual (not textual) representation
of the section number (if needed) and writes the necessary information
to the toc file.  With `structual representation' I mean the sequence
of counter values `2 2 1' instead of the textual form `Section 2.2.1'.

Now the style comes into the play: its the style's task to determine
how the section heading is typeset.  It will therefore convert the
structual representation of the number into textual form.  [So far
nothing new, only mentioned to set the scene.]

Now, many headers in books are of what I call the conventional form:
some space going before the text, then the text, possibly typeset
in a slightly special way (such as the last line centered instead
of flushleft), followed again by space, and then the body of the
section, with some additional requirement such as special indentation
for the first paragraph or that there must be at least n lines on
the same page.  For this format Frank's proposal is well-suited, and
I think we can adopt his solution.

There remain `the rest of the cases', e.g., the format I mentioned
in an earlier message: the text of the heading typeset in the margin
next to the first lines of the section's body.  I think that layouts
like this can be handled in one of the following ways:
- The heading is typeset in a box of standard dimensions (0x0 or
  \textwidth x0), and it is the style's responsibility to put the
  right things into it.
- The style returns a box containing the heading, and it specifies
  where to put it (say, relative to the bottom of the last paragraph
  of the preceding section, to the top of the first following paragraph)

Finally, there are certain layouts that cannot be implemented in this
way.  I strongly vote for leaving a `hook' in the following sense:
the style may set a switch that tells the kernel that it has nothing
to do, that instead everything is handled by the style.

In short, I am proposing a two-level scheme: the toplevel for the
`usual' cases, a second level (where much must be done by hand) for
more complicated layouts, and finally an escape to do all and
everything by hand (within the limits of TeX, of course).

I did not say much about implementation, as my ideas have not reached
that state yet.  Maybe they will never, when you decide that this is
not the way to do it.

Eagerly awaiting your comments and flames...

Rainer