It's been shown that the current approach doesn't scale. This implementation should scale perfectly fine.
