RE: [Axiom-developer] Re: hyperdoc

[Page, Bill]

On Monday, January 17, 2005 12:18 PM Martin Rubey wrote:
> 
> there are some things I really dislike about MathML (and 
> html, in fact):
> 
> * I cannot read it without browser.
> * I find it very hard to edit.
> * I cannot convert it to readable LaTeX, if at all.
> 
> there are some things I really like about LaTeX:
> 
> * I can read it without any tools, if necessary.
> * I can edit it easily. (even without AucTeX, if necessary)
> * I can convert it to html.
> * I can send it to www.arxiv.org.
> 

Martin, I think you are comparing two things that are inherently
different. It is sort of like comparing postscript to LaTeX or
even dvi to LaTeX. Of course postscript is:

* cannot be read it without a program (e.g. ghostscript).
* very hard to edit.
* cannot convert it to readable LaTeX, if at all.

Like dvi and postscript, HTML and MathML are primarily *presentation*
formats. They are (arguably) optimized for portability and ease of
communication between machines. They are not really intended to be
easy to handle for a human being, although some people do code
directly and a experienced coder can easily read postscript, HTML
and MathML if they really need to with additional tools.

LaTeX on the other hand is an input format. It is intended to be
directly coded by a human being. But doing this inevitably does
require some significant training, just ask a Microsoft WORD user.
(No, I don't mean to compare WORD and LaTeX, my point is only
that ease of use can be quite subjective.) dvi is the native
TeX presentation format and it definitely is not intended to
be manipulated directly by a user.

LaTeX is optimized for specifying how mathematical text looks,
but it does not do well as an input language for computation. For
that most computer algebra systems have developed their own input
language that carries the essential semantics of what the user
wants to compute. Usually this can be converted to LaTeX for
display but most systems (e.g. Maple) don't bother unless asked.

OpenMath (which is closely related to MathML) is potentially a
way to provide a common input language or even an exchange
language between systems with enough semantics to allow unambiguous
computation. But OpenMath is again a language more like HTML
except that it is an "input presentation" language rather than
an "output presentation" language. So it is likely that most
systems will still use their own human writable input language.
It is just that OpenMath fits as a standard somewhere in the
middle so that in principle one could use the same human writable
input language with more than one system. This would have some
obvious applications on MathAction even now between Axiom and
Reduce.

Regards,
Bill Page.