Great - thanks for that. One more question that will help my understanding - is the set of discrete partitions created when generating the search tree for a given graph unique? And so, the min/max in this ordering (however we define it e.g., half/full adj matrix) can be taken as the canonical label?