Run a sorting algorithm on all n! orderings of its input and merge the runs as long as they
get the same answers from less: the result is its comparison (decision) tree.
Nodes name the compared elements by their original position in the input, since the tree
does not show how the algorithm moves elements around.