> Would it be a good idea to focus on finding an algorithm first, and then
thinking about how to code it in XPath?


Of course, it would! :-)

Thanks so far for everone's ideas and pointers, which are food for thought for
the weekend (and maybe longer).

- Michael

PS: I was pretty sure that I would have some of the terminology wrong, due to
my almost little knowledge of scientific terms for mathematical problems.
Therefore, I am especially grateful for your efforts to interpret my post.