The current implementation uses KXML as its parser (as it is small and we probably will not get any version interference with other XML parsers).