Re: support for 'macro' formatting languages

Subject: Re: support for 'macro' formatting languages
From: Norman Gray <norman@xxxxxxxxxxxxxxx>
Date: Mon, 20 Dec 1999 18:04:51 +0000 (GMT)

It strikes me that there are two disjoint issues here: (a) what does
an author want to type, and read, in the master document, and (b) what
information is it important to have in the document immediately before
it is rendered.

Sebastian said:

>  > business.  MathML would probably be a better archival format than LaTeX,
>  > so it might be sensible for the publisher to normalise the author's
>  > source, from LaTeX maths notation to SGML, as the very first processing
> and when you have to revise the paper? 2nd edition of your book in 3
> years time? that way madness lies...

Exactly my point: the publisher normalises the LaTeX as the first step
in _their_ processing, but retains the LaTeX-notation document as the
master, so that the author can have a fighting chance of making sense
of it when and if they have to revise it.

>  > as long as the maths they type into the master document were in LaTeX.
> sure, if thats what they want. you could make emacs translate the math 
> from \ { } into < > as you went along

No!  This is senseless!  The point is that I don't want to see MathML
anywhere in my document.  I want to see (some form of) LaTeX in my
document, because I can read that.  You might as well talk about the
convenience of an editor which magically translates into Polish as
I write.

> use TeX in a notation if
>  - you don't ever want to search or analyze it
>  - you don't care about its typographic relationship to the rest of
>    the text
>  - you don't want to mix it with graphics or text 
>  - you dont care about validating it in the way you do for the rest of 
>    the text

You're persuading me I should use TeX in a notation.

>  - you don't ever want to search or analyze it

Neither I nor any of the folk I've asked about this can think of an
advantage in being able to search the maths content of a paper, or of
being able to suck the maths content into a computer-algebra package.

Notation is not consistent enough to make a search on maths sensible,
or make it worth while loading a big equation into Maple from a paper.
Remember that maths is a natural language, with dialects and rhetoric,
and not the formal system that some computer scientists seem to feel
it ought to be.  I would go as far as to wonder if the maths might be
as opaque, structurally, as an included graphic.

Yes, I can imagine having a window in my crystal ball with a paper in
it, and another with an algebra package in it, and pasting an equation
from one to the other.  Yes, that would be a nifty party-trick; no, it
wouldn't be something I'd use much.

>  - you don't care about its typographic relationship to the rest of
>    the text

This isn't my problem.  I can sympathise with the publisher, but print
appearance comes after ease and legibility in the list of priorities.

>  - you don't want to mix it with graphics or text 

I don't see how MathML would be any better or worse than TeX.

>  - you dont care about validating it in the way you do for the rest of 
>    the text

I don't.  Have you ever tried using a grammar-checker on any text
you've written?  That's the level of validation that would be feasible
for maths, producing that level of irritation.

Later, Sebastian again:

> TeX took us to previously unknown heights of automated typography in
> the 80s, but I would argue that the web revolution means that people
> will be willing to sacrifice typography to other concerns. In the case 
> of math, that means being able to search and symbolically manipulate
> their formula.

I think we may be arguing slightly at cross-purposes here -- I'm not
greatly concerned, in this particular context, about good typography.
I'm saying that even if MathML facilitated _better_ typography than TeX,
that wouldn't be enough to justify forcing authors to use it.

Pieter Rijken appeared to agree with Sebastian here, and I know that he
has a similar mathematical history to me, but I know that I've never
met this putative horde of folk wanting to `search and symbolically
manipulate their formula', and the folk to whom I've mentioned the
advantages of authoring in MathML have recoiled in horror.

Moving on to the second point, both Pieter and David Carlisle mentioned
TeX's complete unwillingness to break lines of maths, requiring the
author to insert breaks by hand, which become completely wrong, of course,
when you try to use a text in a different sized line.  This is a serious
problem, and I'm sure the structuring MathML allows would indeed allow you
to generate a half-decent heuristic to do the line-breaking automatically.
However, I'm sure that exactly the same heuristic could be driven,
if necessary, by subtly extended syntax in the TeX maths.  Besides if,
as David said:

> [...] tex-to-mathml
> converters are not currently reliable enough to be used in production,
> but they will be. You will be able to author in tex (or tex-like syntax) 
> and have this appear as mathml in your document.

then it seems there is enough information in the TeX (or `TeX-like
syntax', ominous phrase) to supply the structure, with or without an
explicit conversion to MathML.

David again:

> If you have ever taken a paper with large formulae broken by hand
> and aligned using (say) the AMS LaTeX packages for mathematics,
> and then tried to set that material to a different page width
> you'll know that the TeX model is something you _don't_ want to copy
> as author syntax for documents that are going to be used for multiple
> purposes. 

There are indeed weaknesses in (La)TeX syntax from this point of view,
but it does not follow that MathML is the answer to them.  I can imagine
the amsmath facilities would be nightmarish to make sense of, but that's
at least in part because amsmath gives authors more alignment flexibility
than is really necessary (though this is the American _Mathematical_
Society, who might be expected to be fussier about layout than their
more applied cousins, and thus even less enthusiastic about MathML).

An answer might be for publishers to support LaTeX maths environments
which differ from the standard ones in supporting more structure.  Such a
development would cause much muttering and grumbling amongst authors,
but would be sellable because there actually is a problem to be solved
here, significant enough that authors will put up with some inconvenience,
in the expectation that it will make their papers more usable.

MathML as author syntax is _not_ a solution here because, alignment and line-breaking
issues aside, it seems to provide solutions to non-existent problems,
putting authors to considerable inconvenience for no apparent benefit
to themselves or anyone else.

I'd love to be wrong here, but MathML seems to have acquired a
horrible momentum.

All the best,


Norman Gray              
Physics and Astronomy, University of Glasgow, UK     norman@xxxxxxxxxxxxxxx

 DSSSList info and archive:

Current Thread