Nice to see they finally found out about full text indexing or using lucene (storage system dependent!). Although I did think "I wonder if they built their own binary tree of all word variances and store it in memory to do this"....horrible but I can see it.