for Ruby there's Hpricot and Nokogiri. Now I must admit not understanding what "parsing as close to the browser as possible" would mean. These parsers would not be for displaying the HTML, they're not rendering engines like those in browsers, but will help you navigate the DOM of a (X)(HT)ML document programatically. Surely I'm missing your meaning.
What I'm aiming at would mean "given this lump of (malformed) HTML, what DOM would a browser give me?" Maybe Hpricot, BeautifulSoup, et al are already there, but I don't know. :)
