Wow, this is really amazing. Can you give us a rundown on how it works? Does it do all of its own parsing, and then dump a more well-formed version into the iframe, or does it just send the text verbatim to the iframe and utilize the browser's parsers? It must use some sort of tag recognition to autoindent the input, but the output seems extremely generous (tolerant of malformed or partially completed input) to the extent of what a browser would do.