WHICH CLUSTERING METHOD IS "BEST"?
	No gold standard of truth exists.  We can assess algorithms according to stability of clusters or do leave-out-one analysis to assess the predictive quality of the clusters produced by an algorithm. This latter test indicated on four datasets a slight edge for CAST, but differences were small (Yeung Haynor Ruzzo) Cluster stability on parametrically bootstrapped data, where the cluster centers were hand-picked, was modest (Kerr) An area for future research, or at least more wisdom.