Message-Id: <9302041907.AA25866@sc.zib-berlin.dbp.de>
Reply-To: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
In-Reply-To:  <01GUAIZ63HCIFN4Y4W@MATH.AMS.ORG>
Date:         Thu, 4 Feb 93 20:08:20 CET
From: Michael Downes <MJD@MATH.AMS.ORG>
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:      Re: document classes & numbering systems
Status: R


>
> ... I don't see any reason to really consider
> these [examples] to be items which _demand_ a new class of document style.
> There's a certain point at which we have to accept that simply
> changing the \documentstyle cannot do everything for us.
>  It won't
> change a text which is written using the US style of
>   ``this.''
> to Viking Penguin's house requirement of
>   `this'.
> (note change in quotes and movement of period). Or consider the
> fact that grammatical statements which make sense with one
> citation format will be different for another. The kind of
> ``plug-and-play'' which we'd like to have is just not possible,
> but I think the correct response is not to say that we will
> multiply the numbers of document classes, but rather to say that
> we will accept some limitations to what is possible.
...
>-[Subclasses weren't what I was arguing for here, (1) was intended only
>-to be taken together with (2).]
>
> Perhaps I misunderstand?

I realize my explanations haven't been clear as vodka. Let me try to
restate.

Suppose we have a document class "journal-article" and documentstyles
"article", "article-a", "article-b", "article-c", ... producing
different visual results for documents in this class but sharing a
standard set of document elements with standard syntax.

Suppose we want to be able to freely switch between documentstyles of
the same class, e.g. "article-a" and "article-b", without modifying
anything in the document body.

Question 1 (submitted for debate): Is it a good idea to
allow documentstyles in the class "journal-article" to have varying
defaults for the numbering of the standard elements?

Your first reaction was "Of course it's a good idea, don't be
ridiculous" [my paraphrase]. Maybe you're right, but I wanted to raise
before the subscribers to this list the idea that numbering changes
can sometimes have bad side-effects, side-effects of a sort that
can easily be overlooked by a publisher, *especially* when the
publisher is striving to meet authors' desire to have their documents
published without any meddling by the publisher in the electronic
files, and striving to save the cost of redundant proofreading.

Question 2 (for debate): Is there any way to reduce the
possibility of these bad effects?

Your answer was "no, this problem is similar in difficulty to the
quotes/punctuation problem" [my paraphrase].

But suppose that it were to be required that all documentstyles for
a given class must have the same default numbering system for the
standard elements. Your response was "you can't do this because
publishers almost always want to customize the numbering to their own
preferences in their documentstyles" [my paraphrase].

Question 3 (for debate): Is there a reasonable way to give publishers
what they want while still requiring a standard default numbering
system among all members of a particular class?

On the face of it this probably sounds like nonsense, but perhaps it
is only nonsense under the prevailing understanding of
documentstyles---and isn't it one of the purposes of the LaTeX-L list
to re-examine existing concepts under the light of LaTeX3 design
goals?

To tell the truth, I don't know any good answer to Question 3. I was
hoping that someone else might come up with one! But here is a bad
answer to give some idea of my line of thought.

Start by assuming that documentstyles of the "journal-article" class
must all adhere to a standard numbering system for the defined
elements of that class.  Next suppose publisher X wants to create a
documentstyle "article-x" and they want their authors to use a default
numbering system that differs from the decreed standard. Let's say,
for the sake of example, that they want sections to be numbered as

  \thesection=\Roman{section}
  \thesubsection=\thesection-\roman{subsection}
  \thesubsubsection=\thesubsection-\alph{subsubsection}
  secnumdepth=3

In other words, a subsubsection number would look something like
"III-iv-a" and sectioning levels from \paragraph on down would be
unnumbered.

Well, if the publisher is not allowed to put these numbering changes
into the main documentstyle, where can they go? If they should go into
the document preamble, then presumably LaTeX3 would provide some sort
of nice interface:

  \renewnumbering{%
    section=Roman,
    subsection[section]{-}=roman,
    subsubsection[subsection]{-}=alph,
    secnumdepth=3
  }

Then, the publisher can get what they want by stating in the
documentation for documentstyle "article-x" that the above
\renewnumbering command must be included in the document preamble.

I can't say that I care very much for this approach, but notice that
documents written in this way *can* be switched to another
documentstyle of the same class without fear of bad side-effects from
numbering changes.

Another possibility might be to allow a main documentstyle such as
"article-x" to set up a different default numbering system, but write
a compact description of that numbering system into the .aux file.
Then require that documents must be submitted together with their .aux
file; and during a switch to a different documentstyle of the same
class, arrange to issue a warning if any numbering-system
discrepancies are detected. This will at least alert the publisher
that side-effects should be checked for. And if the numbering system
specified in the .aux file is not too bizarre, I'd bet that publishers
would generally prefer to preserve that numbering system rather than
enforce the defaults of their own documentstyle, so that once again
the possibility of side-effects related to numbering changes would be
eliminated. (This is certainly the past and present philosophy of the
AMS editorial department.)

Another possible answer to all these questions is to say that
portability between documentstyles of a given class is impractical and
abandon the attempt. (You seem to be leaning that way a bit---the
quotes/punctuation thing, etc.) Then---theoretically speaking---each
documentstyle would be in a class by itself and any interchangeability
would be regarded as a happy coincidence. Hand-editing of documents
when going from one documentstyle to another would be considered
normal, even when both documentstyles have the same set of elements.

Although I myself might be favorably inclined to this last answer, at
the same time it goes against a rather strong tide of opinion from
authors who are gung-ho on the idea that their documents can be
submitted to the publisher and used *without any changes whatsoever*.
In order to turn the tide, I can't just say "that's impractical", I
need to be able to explain why. How fortunate I am that I can turn to
all you LaTeX aficionados on this mail list for help in coming up with
answers!

>->-(2) Any numbering changes are done in the individual documents.
>
> ... numbering is part of the style specification.

/////
|@ @|
|  >|
 \_~) (tongue in cheek): If changes in numbering style are immaterial
to the essential contents of a document, then what persuasive reason
can be given by any publisher for deviating from a standard numbering
system required for a given document class? When creating a document
class, why not poll a few publishers, distill a happy-medium numbering
system from their various preferences, and say to them "Here's the
numbering system you have to use, unless you can give us a good reason
for deviating from it"?

Michael Downes                              mjd@math.ams.org (Internet)