This version fixes the points a couple of people raised about the documentation, picks up on a few more well-formedness violations, provides options for either passing entities through unchanged or for rejecting unknown entities, adds support for numeric entities, adds support for > characters in attributes, is more careful about ignoring processing instructions, and provides an option to make <!ENTITY...> and <!DOCTYPE...> declarations fatal instead of just ignoring them. This means that with the right options it now passes all of the w3c's not-well-formed tests apart from those which rely on being able to parse processing instructions and those which require funny characters.
Good work. Thank you.
