yeah, you could just pull the parser out of any open source browser and voila a parser not only battle-tested, but probably the one the page was developed against