Re: Agenda for Monday April 1st

I've read the minutes of the last few meetings and Dave's 
proposal. Here are some comments.

1. If you want to have a model for large or extendible characters,
I would like to suggest that we consider (again) the model of TeX
and the Computer Modern fonts. Barbara Beeton of the AMS has also
done interesting work on math font layout in some ISO workgroup.

2. Automatic numbering, or better: automatic generation of numbers
(labels) from keys used in symbolic referencing, a la \label/\ref
in LaTeX or ID/IDREF in SGML, is fraught with difficulties.
Between =-=-=-= lines I'll include a piece I wrote about that in
relation to the Elsevier Science DTD's.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
The problem of symbolic referencing has never been one of the DTDs,
but one of finding an algorithm for deriving (printed) numbers/labels
from symbolic labels in the SGML source document. 


Earlier summary of the problem
------------------------------
A generic mechanism for counters, including some rules for
presentation of counters, is as follows.

- Every type of numbered object, table, figure, equation,
  reference, has its own counter.
- All top-level counters start at the value 0 (zero) at the
  beginning of the document.
- If nesting is possible, every nesting level has its own counter. 
- Counters of levels below top-level start at the value 0 (zero) at the
  start tag of the next higher (parent) level.
- From the values of the counter(s) a label is calculated. This
  label is e.g. printed in the article as it appears in a journal
  issue. The relation between labels and counters is
  application-dependent.
  Example: |<sec id="99">| could be printed as ``3.2.1''.
- The values of the |id| attributes are unique within the
  document instance.
- The |id| attribute is not identical to the counter value.
- If the |id| attribute is missing, the counter is not
  incremented, and the element is not numbered. This allows for e.g.
  un-numbered equations. This means that tables and figures always have
  an |id| attribute even when they are not referenced. (Or are
  there also cases where tables and figures are not numbered?)

There are (at least) the following exceptions to the algorithm
described above:

1. The contents of a figure form one (physical) entity, but
inspection of the contents show that there are logical sub-figures
(hidden) inside the figure. Example: the contents of figure~2 contains
markers `a', `b' and `c', which indicate logical sub-figures 2a, 2b and
2c.

2. Normal equation numbering (1, 2, 3, ...) is interrupted by an
equation number that is `derived from' an earlier one. Example: (1),
(2), (3), (1'), (4), (5), ...

3. Numbering in lists, or sub-numbering in equations, is
interrupted. Example: a numbered list with items 1, 2, 3, is followed
by a paragraph of text (the |<p>| is on the same level as the
|<l>|), and then the list continues 4, 5, 6. The intention of the
author appears to be {\em one\/} list, but this list is
interrupted. A similar example can be found in equation sub-numbering:
equations (4a) and (4b) are followed by a paragraph of text, followed
by equations (4c) and (4d). 

Below I will discuss each case separately.

Hidden sub-figures
------------------

When a figure (as far as we know, hidden sub-components occur only
with figures.) that exists as one physical entity consists of logical
sub-entities, the only solution lies in adding this information, which
lies *outside* the document, to the document. One possibility for this
is to give the |<fig>| start tag a new attribute that specifies the
labels of the individual components. This enables generation of the
labels of the sub-figures.

First an example of a regular case:

  <fig id=fig4>
  <fig id=fig4a><figlnk file="fig4a"></fig>
  <fig id=fig4b><figlnk file="fig4b"></fig>
  </fig>

Here the outer figure and the two nested figures each have their own
unique key for referencing (fig4, fig4a, fig4b), and we have the
following (id, label) pairs:

  (fig4, 4)  (fig4a, (4 1))  (fig4b, (4 2))

where the counter strings (4 1) and (4 2) are represented as `4a' and `4b'
(but this presentation is application-dependent of course).

Second an example of hidden sub-figures:

  <fig id=fig4>
  <figlnk file="fig4">
  </fig>

the picture in |fig4| contains two labels, `a' and `b', that identify
sub-figures. Both the figure as a whole and the individual sub-figures
are referenced. If we add an attribute to the start tag we can achieve
this: |<fig id=fig4 comp="a b">|. The keys `fig4a' and `fig4b' are
implied (automatically generated); these implied keys act as aliases
for `fig4', since they refer to the same physical object.
Unfortunately, the attribute |comp="a b"| also suggests a
presentation. Now we have the following (id, label) pairs:

  (fig4, 4)  (fig4a, (4 1))  (fig4b, (4 2))

where |4|, |(4 1)| and |(4 2)| refer to the same object
(physical figure) in the document instance.

Derived equation numbers
------------------------
We illustrate this case with an example. Suppose an article contains
the following sequence of equations: (1), (2), (3), (1'), (4), (5),
... In that case we have the following (id, label) pairs for the first
three equations:

  (eq1, 1)  (eq2, 2)  (eq3, 3)

The next equation should have an (id, label) pair like this: 

  (eq$1, label(eq1)') 

where |eq$1| is the unique |id| assigned to equation~(1'), and
|label(eq1)| stands for the label of the equation with id=eq1.

And the next equation, (4), should have an (id, label) pair like this: 

  (eq4, 4) 

The question is whether appending material to an existing label is the
only form of derivation, or whether other forms of derivation exist.

Interrupted sequences
---------------------
This case occurs with list items within a numbered list, or with
numbered sub-equations. (Lists look like a special case of nesting:
numbering of list items normally starts with 1, but every list item
must have an \verb|id| that is unique within the document. This could
be achieved if we also numbered the parent element \verb|<l>|.)

Suppose we have 

  <l>
  <li id="liA">a footnote is ...
  <li id="liB">footnotes should be ...
  </l>
  A paragraph of text comes between the two numbered lists.
  <l>
  <li id="liC">a figure is ...
  <li id="liD">a table is ...
  </l>

and the author has numbered the list items as 1, 2, 3, 4. We agreed
earlier that one of the guiding principles is to follow the author
unless there are errors in the manuscript.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

3. "HTML-Math is intended to allow ... as well as rendering to
graphical bitmapped displays." What about rendering to paper??

4. "There are hundreds of symbols in use in mathematics." For those
of you who think this is covered by Unicode, Computer Modern fonts and
the AMS fonts (although the latter *are* very useful), should study
pages 47-50 of the following document

    ftp://ftp.elsevier.nl/pub/sgml/artdoc.ps.gz

This is the documentation of our article DTD, and it contains 
tables with *all* symbols and characters we use in typesetting our
products. There is redundancy in it, I know. But more important is
that to cover this set 100% you need more than the font sets I 
mentioned earlier. Elsevier Science is in discussion with several
societies about the possibility of completing a font set that
covers all this, and perhaps more, and make it freely available.
This will also solve legal problems with embedding fonts or parts
of fonts in PDF files.

5. Superiors/inferiors: I think we should add something about
vertical position of inferiors in mathematics and chemistry (these
differ, as some of you may know).

6. In the list of layout idioms I believe there is at least
one missing, namely an arrow that adjusts its width to the text
appearing above or below it. This is used in chemical formulas, and
is different from an arrow or brace AS ORNAMENT, that adjusts its width
to the text underneath it. Examples of both:

    \widebrace{a b c ... x y z} (the brace is the ornament)

           catalyst 
      A  -----------> B         (the text above the arrow is the ornament)

7. In example 2 a lot of reserved words occur: from, to, of. How
do you get these words as normal words in the text (in roman)?

8. What about coding fractions: are we going to follow LaTeX and most
SGML DTD's and use something like (fraction (numerator ...)
(denominator ...)) or are we copying the horrible TeX construct
(... over ...)? I've been asked by people on the LaTeX3 development
team to express a strong preference for the former, and a strong
dislike for \over!

That's it. I'll try to monitor the corner designated to our group
on www.w3.org, although the password Dave gave me doesn't work!  :-)

If Dave can let me know several days in advance when a telephone
conference occurs (on Monday evenings) I can try to attend. But
there will be Mondays on which I am not at home from 7 pm to midnight
(that's 12 noon to 5 pm EST, since we're in daylight saving time
as of last weekend).

Regards,

Nico

------------------------------------------------------------------------
Dr. Nico A.F.M. Poppelier
Elsevier Science, APD, ITD               Email: n.poppelier@elsevier.nl.
Molenwerf 1, 1014 AG Amsterdam           Phone: +31-20-4853482.   
The Netherlands                          Fax:   +31-20-4853706.   
------------------------------------------------------------------------
                  The truth, the whole truth, and nothing but the truth.
                                             And maybe some compromises.