Re: What are your plans for MathML macros?

> Maybe we ought to have a specification for the external preprocessing.

Any such embedding needs to be part of a global picture. It is no use
MathML on its own defining some interface to this. A typical HTML/XML
document is likely to have lots of elements containg markup from
different XML DTD (or possibly even external notations as you suggest).
The controlling browser needs to be able to recognise the parts and call
out to appropriate code. If each XML application defines its own
interfaces the result will be that nothing will get implemented.

So this has to be (and in fact, is) a more central W3C activity rather
than a Math working group one, although clearly one needs to ensure that
any outcome has the functionality required to support math.

At the current time while many of the issues that are likely to be input
to any such system, XSL, XML-namespaces, ... are still at a draft
stage, it is hard to be sure how these things will work out, 

Having said that, personally I think that it would be a shame
if something like this became common practice. 

  <embed syntax="URL_for_my_copy_of_TeX2XML">
  6 \times 9 = 42

There is a real chance that  MathMl can get rendered on the web, and
controlled by stylesheet mechanisms as part of native support, just as
the rest of the (son-of-HTML) document.

Even if the interfaces are finalised that let you do something like the
above, the end result would be that your document is only processable
by a subset of systems, ones that can link in to whatever engine is at
the end of that URL, and the document won't be editable at all in XML
aware editors.  It is not just static rendering that is at issue,
we need systems where you can select sub expressions, cut and paste from
your web browser into Axiom (or mathematica or maple, or tex, or ...)
detecting subexpressions if the parent expression is in a foreign syntax
as above is no doubt technically possible but likely to be over
complicated compared to the more direct option of writing 
 6 \times 9 = 42
in your preprocessor but putting the full MathML XML syntax into the
document. That is the whole point of XML: to make the parse tree of the
expressions simple to obtain and manipulate in real time in a browser.
(well its not the whole point, but that's a major point driving the
simplifications that produced XML from SGML)

Note that while I think it is not a good time to guess what the final
form will look like, it is a _very_ good time to be specifying what the
requirements are, so this message is not in any way supposed to suggest
that you (and other people) should not continue to post ideas to this
list.  I should also say that although I am (now) on the math working
group these replies are personal ones, not `agreed' responses from the
group, although I don't think I've said anything too controversial:-)