Watching this conversation, I found myself interested in how many
experts say there is rarely a need to match text nodes. I do that
all the time. Some examples include:

* transforming straight quotes to curly quotes
* conditionally changing "-" to en-dash or em-dash
* convert PUA characters to<tei:g>  elements
* deal with soft hyphens
* ditching the punctuation that follows a<list>
* finding the character before a footnote

So I'm guessing either
a) there's a better way for me to be doing these things, or
b) those experts who spoke of the rare need for matching text()
    either don't deal with data like mine -- TEI, i.e., mostly mixed
    content --, or deal with so much more of it that my needs for
    matching text() are, in fact, pretty rare from their point of view

I'm hoping someone will post the better way if it's (a). :-)