Hi Kevin,

Yes this is exactly what I'm doing with it and it does seem strange to
me that on a transformation only basis the new code is slower but I
wonder whether the MSIL that gets created during the compilation stage
in the new class is optimised for a certain "style" or method of
building XSLT and our stuff isn't in line with that so the raw
interpreter works better than the compiled instructions which are
perhaps adding redundant steps...

What would be interesting would be to see if there is a way we can see
the compiled instructions and see if there is anything obvious. When
I've run this across different stylesheets it is interesting that a
couple of the templates I personally thought were quite complex
actually performed better than the test sample indicates which I think
of as less complex XSL. This is what leads me to the "style" idea
above...

I'm going to try and construct a stripped down example of my code and
run the tests again and then post the code up for additional scruitny.
At the moment there is too much code and too many dependencies to get
my application running to post it up here but I'll get a test case
working that uses these different XSL documents so we can do comparison
of style on the two transform classes.

Cheers
AndrewF