Re: Math on the web without MathML (CSS 2.1 rendering for HTML and XML)

David Carlisle said:
>
> the issue of availability of mathematical symbols is completely separate
> from the issue of what markup to use. Currently most mathematical fonts
> that are available are not in Unicode positions so either your renderer
> needs to have special knowledge of the fonts or the user needs to use
> some font tools to re-arrange the fonts so they do have internal Unicode
> tables. Neither of these situations is optimal, but that's life and it's
> irrelevant to any discussion of the merits or otherwise of MathML
> markup.
>
> CSS (as specified) gets more powerful and CSS (as implemented) gets
> closer to the specification so the answer of what can or can not be
> reasonably displayed just with CSS changes each time you look. As I've
> showed I put quite some effort into seeing how far you could get some
> years ago. It's possible that you can get further now, which is good,
> but I have yet to see a reasonable rendering of any stretchy operator
> (large brackets in particular) just using CSS, similarly I have yet to
> see a good way of positioning superscripts on large operators.

Yes, there is a few limitations ***today***, for instance rendering of
angle brackets is far from optimal. But situation will be improved in
future, specially with a good CSS-Math module. I also am waiting to see
correct non-tricky rendering of K-12 math using MathML, and apparently we
would wait to future MathML 3.

Now well, does CSS current (today) limitation to render angle brackets
mean that MathML is superior for fence-like constructs? No, because MathML
has limitations also. Some authors have analized limitations of MathML to
rendering of generic brackets. I already cited in the past about that.

In a IAP 2006 talk devoted to math on the web

[http://web.mit.edu/violeta/www/IAP2006/]

Ian Hutchinson cited some of rendering problems of MathML. One of them was
"often large braces or integrals are too big".

> (It's
> quite common  to have a superscript on a matrix for example).

There are available 'old' CSS 2.1 techniques for that!

> Also of course it depends what you mean by "with css" You can of course
> use css absolute positioning and place any character (including
> charaters representing fragments of large brackets) anywhere. You can
> get very high quality by doing that but in practical terms the markup
> that would be required is not useable unless it is generated
> mechanically. This is essentially what the jsmath system does (as far as
> I understand what it's doing) using javascript to parse a tex like
> syntax and position everything with css.

Correct, also is the approach taken by SVG mathematical renderings
(curiously cited in the w3c MathML tools page :-). In both cases,
presentational MathML is of no real utility, specially when one remembers
that CSS and SVG techniques can be used in tasks that MathML cannot (e.g.
rendering of chemical formulae via SVG, rendering of headings and quotes
in a HTML page).

JsMath is interesting but uses images and special font metric information,
therefore is not suitable for several activities.

Default cheap CSS stylesheet can be used for basic rendering (good enough
for many users) and advanced stylesheet and/or probably javascript module
for achieving virtually TeX-like quality.

Note that the usage of javascript modules or even server-side plugins is
not a limitation of CSS because it is also usual technique on MathML
community: ASCIIMath and IteX are two known examples ;-).

> Do you have a reasonable xml input syntax (or even better mathml input
> syntax) and a set of css rules that can display large brackets and
> superscripts on lareg expressions? If so that would be very interesting.
>
> David

Apparently you fail to understand still that rendering via CSS does not
need of mathml ugly and limited markup (one could provide CSS thecniques
for the rendering of OpenMath markup for example) not even CSS needs of an
xml approach. That is cause that you can use the *same* CSS tecnhiques in
XML, XHTML1, XHTML 2, HTML 4, HTML 5... without special mimes, specific
DTDs, and all complexity. In those aspects CSS is again very superior to
MathML.


Juan R.

Center for CANONICAL |SCIENCE)

Received on Monday, 10 July 2006 10:50:33 UTC