makes perfect sense. If there's the opportunity to take advantage of Lucene's internal faceting engine, that'd probably be the best solution ... less code and all, and I doubt we can get better performance than the native engine, though I don't know if that's an option for infinispan.