It requires tremendous resources compared to html parser. It's possible, but very costly if you want to crawl the entire Internet.