We should probably do the opposite. Build a markdown parser and then have a way to give the output an understanding of ddoc features. It will also be usable in other projects ;)