Edit: And for Haskell, achieving needed performance can be quite tricky with regard to its laziness. But Snap is driven by enumerators/iteratees (via the enume

Edit: And for Haskell, achieving needed performance can be quite tricky with regard to its laziness. But Snap is driven by enumerators/iteratees (via the enumerator package), which make it much easier to do fast near-constant space IO. I have built a web application that uses Snap to query large (parse) treebanks, and it was easy to make it performant.