How much work does it take to go from an engine that can read standard XML to one that can read an ODT document's XML and do something useful with it? At what point of complexity does that engine create a de facto standard?