[nodoodles]

One fun mechanism I've been using for reducing html size is diffing (with some leniency) pages from same domain to exclude common parts (ie headers/footers). That preprocessing can be useful for any parsing mechanism..