PyRXP claims to be the fastest Python XML parser around… but this has a price. Once XML is parsed, you get a tree structured as tuples, lists and dictionnaries. While this is very Pythonic, it’s not so trivial to write/read code handling such a tree.
People at Reportlab argue about that for speed. Dealing with classes in Python is mem hungry, so they use only basic data structures (tuples,list,…). I can understand that but i believe it’s an extremist point. Finding a balance between the DOM API (which is extremist too :) and such structures could turn to be more popular, maybe a Python xml_node builtin type …