Heh - a simple sort on the data would be equivalent to that algorithm (if you don't mind being a O log n slower).

Update: Sorry - you say exactly that in the last sentence.