Re: Comments on parsing steps:

About prefix embellishment operators:

First I should point out that we (Wolfram) have not yet described (to
this list) our proposal's parsing algorithm in enough detail to
consider it "specified", but have only described some of its
characteristics. Soon, I hope to correct this by sending a description
of an actual algorithm which has all the features we've described.

If we can figure out a reasonable and reliable way to add prefix
embellishers to this algorithm, we'll do so, before I send the writeup.
But (unlike every parsing feature described in our proposal so far)
this will not have been feasibility-tested in Mathematica.

If we don't figure this out, or it's undesirably limited, of course
anyone else is free to propose an alternative parsing algorithm which
supports this.

In any case, HTML-Math should also include markup for surrounding
things which should be in a different font, or which should be bold or
italic. So, at worst, something like <bold>+</bold> (but perhaps with
a different tag name, perhaps one common to the rest of HTML) could be
used to make an operator bold, even if  &bold; +  did not work.
(Suggestions on making this compatible with the rest of HTML and/or CSS
are welcome. I am not up to date on all the related aspects of HTML
3.0.)