> Solution 2 - The second solution requires utilization of the preceding-sibling 
> axis in XPath.  The problem with using preceding-sibling in any sort of test is 
> as the processor continues through the tree another preceding-sibling is added 
> to the list of preceding-siblings.  This test will evaluate each and every 
> preceding-sibling so it doesnt take long for performance to become a BIG issue. 


this is essentially why "muenchiam grouping" using keys is a big win.
The original posters (rephrased) problem sounds like the canonical use
case for the grouping techniques (as explaied at
www.jenitennison.com/xslt/grouping) 

David

________________________________________________________________________
This e-mail has been scanned for all viruses by Star Internet. The
service is powered by MessageLabs. For more information on a proactive
anti-virus service working around the clock, around the globe, visit:
http://www.star.net.uk
________________________________________________________________________