Re: Monday conference call

>I see the W3C Math group as being at a crossroads right now and
>look forward to seeing Bruce's proposal:
>Route One:
>We stay with SGML tags for marking up appropriate parts of an expression
>together with operator precedence for expressions and SGML entities for
>special characters like the differential `d'. In going down this road
>I would prefer to leverage the HTML 3.0 design as much as possible, e.g.
>by adding new tags such as <op>..</op>, <var>..</var>, <fn>..</fn>,
><plex>..<from>..<to>..</plex> etc.  This will minimize the impact on
>people who have invested in the HTML 3.0 way of doing things.

A full implementation of a full design will make what people have done
so far be what they really are -- prototypes.  As anyone knows that has
written software that real people use, it is a long way from a prototype
to a product -- much longer than from idea to the prototype.
You may hear some implementers complain, but it is the users who are
the important parties, and we should have a clean design for their sakes,
not a compromise.

>Route Two:
>Make a clean break to an entirely new notation that is completely
>independent of SGML.  This gives us the opportunity for extensible
>notations that are much easier to read. Indeed the linear syntax can
>be designed around how people actually say math expressions, e.g. over
>the phone. I believe that I can come up with a simple proposal for
>this quite quickly and thence to a prototype based upon Arena or
>WebEq. This road is less familar to the Web community but one which
>I feel is well worth exploring.

I think it is a very bad idea to design the linear syntax around
how people "say" math expressions for several reasons, the most important
of which is the "anglo-centric" view of math.  Written mathematical notation
has a universality well beyond any spoken language.  It is used
and recognized in the Americas, Europe, Asia, Africa, India, etc, independent
of the user's ability to speak, write, or read English (or American English).
The closer we can make the linear syntax correspond to how it will look on 
piece of paper, the more univerasal it will be.

Obviously, 2-d operators will have to be made up, but that is not
anglo-centric.  I think the only place where the language will have
a anglo-centric design is in the name of the sgml-tags and character
entity names.  In these two cases, we are simply following established
practices (tag names) and international standards (assuming we use SGML
ISO standard names).