Query optimization in modern database management systems faces in many cases large access plan spaces. This papers addresses the case of recursive queries where the optimization space is very large and requires special treatment. I particularly liked this paper because it introduced for the first time advanced optimization techniques, in this case simulated annealing, to the problem of query optimization. Simulated annealing is a probabilistic hill climbing algorithm. The authors study in details the basic factors influencing the performance of the algorithm and provide in a neat and "educational" manner implementation issues. In my opinion this is a model paper for people writing query optimization algorithm papers, where sometime details are not presented and readers are left with lots of questions.