And with SPT I can sling together barn door SQL that runs like billy-o without having to intricately assess the optimizer in case of surprises. We both know that KISS is the optimizer's friend. That is a rather dangerous assumption - not unlike other assumptions (i.e. 'deduced it based on my knowledge of the optimizer').
Now, simply stating that the subsequent filtering of a column derived from a RANK function will yield a table-scan...that in itself can merely be an innocent mis-statement. However, given the other times that the great 'T-S' argument has been invoked, one does wonder if the invocaton is more pretext than presumption.