I guess I misstated my query last time. I understand the implementations as defined in the spec; either create a data result cache and hit that or regenerate the query each time. What I was wondering was which people were choosing to implement and why.