Are these properties of the XML language (vocabulary) that you are using, or properties of the information that you are modelling?
I would have thought if you are modelling the structure of an aeroplane, or the structure of a natural language text, or the history of first class county cricket, then the complexity of your XML is likely to be determined largely by the complexity of aeroplanes, natural language, and cricket respectively.
I guess you have design choices in modelling that will substitute one kind of complexity for another, for example you may use a process of abstraction to derive more generic concepts, which will reduce the number of types but increase the level of polymorphism.
Michael Kay
Saxonica
