Todo pre 1.0.18:
    . Make Markdown.pl and MultiMarkdown.pl warn if unexpected options are 
      used, and also have docs on what options are available. Expose the other
      module options in the scripts.
    . Remove the nasty hackery of all the locals in the markdown routine.
    . Fix tests for the email address encoding. (Maybe call srand in the 
      test?)
    . Fix copy and pasted regexes (nested brackets and StripLinkDefinitions)
    . Make the merging of MDTest more obvious, and get back to the markdown
      mailing list about which tests a) I pass, b) I think should be added to 
      MDTest.
    . Re-merge Movable Type and Bloxom compatibility to the scripts, and add 
      tests.
    . Need a dingus, write a small / demo app then ask the catalyst guys to 
      host?

Todo pre 1.0.19:
    . http://rt.cpan.org/Ticket/Display.html?id=33443
    . More tests in t/15inlinehtmldoesnoturnoffmarkdown.t, I bet turning this 
      feature on breaks things.
    . Document the markdown (and multimarkdown) syntaxes in the module 
      properly.

Todo pre 1.0.20:
    . Stop perl shitting itself due to regex backreference when doing lists
      and fix the list bug
      (see t/17ulolreportedonlist.t and Markdown mailing list)
      This probably involves the first draught of the new parser...
    . Text::Balanced is the slowest thing ever. 
      The syntax guide looks like this:
        49.4   0.569  0.568    113   0.0050 0.0050  Text::Balanced::_match_tagged
        18.1   0.209  0.791      3   0.0696 0.2637  Text::MultiMarkdown::_HashHTMLBlocks

Todo pre 1.0.21:
   . Have a look at extra features in 
     http://code.google.com/p/python-markdown2/wiki/Extras and put anything 
     we consider useful on the roadmap.
   . Integrate maruku (ruby port) test cases properly, and go looking for 
     more.
   . Have a look at the php markdown extra tests to see if we want to steal 
     any of those features.
     
Todo pre 1.1:
   . Build a parse tree when parsing (which can be returned), rather than just 
     using strings. This would make the output format much more flexible, and
     would remove a load of nasty hacks which escape, then un-escape things
     again whist processing..