Mamoulis and Papadias motivate the work presented in this paper with the observation that complex queries for spatial databases may involve multiple inputs, and hence multiway joins. They adapt pairwise join algorithms for multiple inputs, discuss the role of synchronous traversal and query plan optimization using combinations of synchronous traversal and pairwise methods, and address query optimization using dynamic programming and randomized search algorithms.
Evaluation is done with both synthetic data sets, and with actual geodata from selected datasets. The work and its exposition are detailed and solid, and the results facilitate useful advances in implementations of geodata processing. Skew in datasets is also addressed quite competently, though I would be interested in seeing more experiments with--and analysis of--the effects of skew on the performance of the algorithms.