> we will still have the problem of interactions between different macro
> (or whatever they may be called) definitions.

Or will we?  Sebastian, in his less anarchic moments, and The XML
Brigade, do not want You to have such nasty things as macros to make
the language extensible; at least that is my understanding.

[Noisy aside: no author-macros would make producing LaTeX3 infinitely
simpler too: we can see ways around that (but not sensible ones using
only the current TeX).]

> And doing any sort of extensive writing in mathematics *without* some
> sort of macro facility is just too awful to contemplate.

I completely agree, and I do not think that editors would like it
either.  So I hate to say it yet again, but research maths notation
*is* different from natural languages (and, he added hastily, even
more different from formal languages, such as mathenmatica, maple etc

I am not at all against XML/MathML as a useful lanaguage, but it must
fit into authoring/editing systems for all types of maths that fully
supports all the different types of people who need to use them.