Message-Id: <9305030423.AA09419@sc.zib-berlin.dbp.de>
Reply-To: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
Date:         Sun, 2 May 93 23:12:51 -0500
From: Cameron Smith <cameron@SYMCOM.MATH.UIUC.EDU>
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: documentstyle option versions
Status: R


I've been away from email for a bit and just got a chance to see some
discussions that took place a few days ago.  Sorry for the belated response.

Michael Downes asks how one may guard against "subtly wrong results"
when A sends B a LaTeX document that uses a non-standard but widely
popular foobar.sty and B uses an older version of foobar than A used
in creating the document.

David Carlisle suggests testing \foobarversion (or some other macro
known to have changed between versions); Kresten Krab Thorup
suggests incorporating version control into LaTeX by allowing
\documentstyle options to specify what version they expect;
Nelson Beebe points out that languages "like Lisp and Bliss" have
this functionality, but he recommends packaging all but the most
stable and "major" style files with your documents when you send
them out into the world.

I ask a counter-question: what kind of "subtly wrong results" are we
worried about?  If they're that "subtle", what difference do they make?

LaTeX is supposed to be a structural rather than a visual markup system.
(It's not perfect, but that is and always was the ideal that LaTeX
was supposed to be about.)  If a change in a style file merely causes
a paragraph to re-flow or a figure to be re-positioned or some other
"subtle" *visual* incompatibility, then I claim that that is not a
problem.  It's not even an error, really -- after all, the theory is
and always was that in principle you should be able to completely re-format
a document by simply changing the \documentstyle from (say) "article"
to "report".  (I know that this is not practical in general, and that
there was some discussion on this list a few weeks ago about defining
families of interchangeable styles for this very reason.  But again,
the fact that reality isn't perfect doesn't change the ideal.)  As long
as the text of the document isn't mangled (and what kind of style file
change would actually cause text to be lost?) we should not (I claim)
worry too much if "subtly wrong results" mean only changes of appearance
and not changes in content.  (I'm assuming -- based on Michael's use
of "subtle" -- that we agree that an incompatibility that causes major
hunks of a document to be mangled would be so obvious that no warning
messages were needed.)

I think the point I'm trying to make is this: if what you want is to
be able to cause a colleague's printer 3000 miles away to spit out
identical pages to the ones your printer produced, then you don't really
need a content markup language like LaTeX; you need a page description
language like PostScript.  Conversely a structural markup language like
LaTeX should *not* bog itself too deeply in questions of visual markup.
I would not consider it a great advantage to "enhance" LaTeX with
a lot of features whose underlying design goal is to turn it into a
system that can ensure that identical pages are always produced from
the same source file -- I think the fact that LaTeX is somewhat "aloof"
from such concerns is one of its strengths.

I know that if you're doing professional-quality publishing (books,
journals -- as opposed to classroom handouts and seminar announcements)
you have to be able to exert fine control over the appearance of the
finished document, and I don't mean to say that I don't want LaTeX
to be *able* to do that sort of thing.  But that kind of work necessarily
involves exact knowledge of the fonts being used (sometimes even to
the point of knowing which vendor's version of the font is being used),
what output device is being used, what kind of paper and ink are being
used, etc. etc.  This level of production detail is necessary when you're
producing finished work -- but it is inherently site-specific, and it
is emphatically *not* the kind of thing that LaTeX should concern itself
with when it is being used as a document exchange language.

Part of the problem is that LaTeX wears several hats.  When it is used
to produce (say) pages for a book, fine control over output appearance
is needed -- and the document becomes inherently non-portable.
When LaTeX is used as a medium of exchange among colleagues who are
using different computers and different output devices and different
previewers and different fonts, portability is necessary -- and quibbles
over "subtly wrong results" are unimportant.

Please don't let either of these two very different applications of
LaTeX outweigh the other, and please don't mix them up!  As I see it,
there is no "right" answer to the question Michael asked because the
question itself calls for incompatible premises (that a document at
once be portable from site to site but also exert rigid control over
the appearance of the output).  I don't really object to the idea that
a document should be able to declare what version of each component
it expects to use, and should be able to complain if it gets the wrong
version.  What I do object to is the (as I understand it) underlying
idea that LaTeX should be able to wear both its hats at once.  I think
that is an un-achievable goal, and I think we could do major harm to
LaTeX (and possibly to ourselves!) if we set our hearts on achieving it
and try to re-design LaTeX accordingly.

All IMHO, of course.  Just an off-the-cuff reaction.

Finally, let's remember that many posters to comp.text.tex (including
some members of this list) seem to consider the right answer to all such
questions as Michael's to be something along the lines of:

  "You have been found guilty of using outdated software.  The problem
   you reported was fixed in the September 1943 release of foobar.sty.
   Please don't burden the net with complaints about incompatibility
   if you can't be bothered to keep your style files up to date."

This is meant with only 1/2 of a :-)
I really am somewhat in sympathy with the position that if you're going
to use foobar.sty and you want to be able to receive documents that use
foobar.sty and have them work right, then it's your responsibility to
maintain your local copy of foobar.sty.  Nobody else can do it for you.
Even if we work out a way to automatically complain because you haven't
done it, you *still* won't be able to get the right output until you fix
your problem.  So get off your arse, B, and update your foobar.sty!  :-)

--Cameron Smith