The asymptotic efficiency of the Kaplan-Meier product-limit estimator, relative to the maximum likelihood estimator of a parametric survival function, is examined under a random-censoring model.