I am full of admiration for Jeni's ingenuity, but I wonder whether a
simpler approach might not be to build  a lookup table of character
positions and their expansions (using XSL keys), and then cycle over
the text letter by letter seeing if there is a replacement?

If you work by cycling through the replacements, surely

<foo:char>$</foo:char>
and
<foo:replace>$\mathbb{P}$</foo:replace>

will fight? the $ in the second fragment might end up escaped

sebastian


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list