The intent is certainly to improve performance, its just that this wasn't the driving force. Right now ResultTreeHandler does extra work to collect all arguments before emitting a true SAX call to the serializer, such as with startElement(). Often however the particular flavour of handler isn't interested in certain things, and the extra work is wasted.
By delegating that work to the serializer, the serializer will do a better job because it knows what the output method is. I'm hoping that the ResultTreeHandler will go away, but there are certain things in it that will probably have to survive. Brian Minchau XSLT Development, IBM Toronto Phone: (780) 431-2633 e-mail: [EMAIL PROTECTED]