> Your example is quite clear, but raises the issue of "what is 
> identity?", 

It's a very hard concept to create a definition that isn't circular: you can
use the concept of "the same position in the same document", but that still
leaves the problem of "same document".

I think the way this tends to be done in formal specifications is to start
with an axiom: "there exists a set of nodes", which automatically implies
that each node in the set has identity. But I'm not a formalist.

Michael Kay
http://www.saxonica.com/