> I should have made that clear. My times are not from 
> beginning to end of command line execution. I am performing the
transformation 
> inside my own Java code. I don't start the clock until immediately before 
> the call to perform the transformation and I stop the clock upon return
from that
> method.

Doing this you've cut out a lot of the cold-start overhead from your timing
but not all. I generally run the thing three times, the second and third
times round are about 70% of the first time cost. I think this is because of
Java class loading and incremental compilation.

Mike Kay


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list