Reading Papyrix's implementation of HTML... I'm starting to question how much I'd really want to reuse.
Autoconversion from HTML5 to XHTML5 makes perfect sense for them as an ereader, but less sense for a web browser. And I'd prefer the styling to be entirely configured via useragent CSS, just seems cleaner to me.
That said this is useful prior art, & does give me some guidance. Maybe I'll like their CSS implementation?
They do bring up the question of how to handle tables...