The paper itself is an overview that doesn't discuss the cache-friendly optimizations in the original 1986 PhD thesis so it feels like a hash table all the way down.