BEFORE 3.00

  - write more test scripts
  - documentation updates
      - more examples
      - ...
  - run more leak-tests
  - update README

PLAIN BUGS

  - the tokenpos reported for the generated end event for
    empty tags in xml_mode can't work.  Should tokens be empty
    and tagname reported?

ISSUES

  - Should we rename 'process'?     (candidates: 'pi', 'proc',...)
  - Should we rename 'cdata_flag'?  (candidates: 'is_cdata',...)
  - Should token1 go away?
  - Should we report tokens for 'text'?  If all you want is to know
    the length of a text segment, then 'tokenpos' could be used instead
    of asking for 'text' and then applying length($text).
  - Something better than '' as the default argspec?

  - Should some '\n' be part of markup instead of text.  The SGML standard
    has a few things to say about this.  A '\n' following a start tag should
    be considered part of that tag.  A '\n' preceeding an end tag should
    be part of that tag and not part of the text flow. According to
    '../html4.0.1/appendix/notes.html#h-B.3.1'

FEATURES THAT CAN WAIT

 - report 'linenumber'
 - unbroken_text option
 - utf8 mode (where entities expand to utf8 chars instead of latin1 chars)
 - pic attribute (">" or "?>" are defaults)  useful where you want
   to have "?>" as pic, but don't want the rest of xml_mode.  For instance
   to parse PHP files.
 - <![%app1;[...]]> (parameter entities)
 - $p->ignore("script", "style");

POSSIBLE OPTIMIZATIONS
 - none that I can think of right now
 - run the profiler

MINOR "BUGS" (alias FEATURES):
 - no way to clear "bool_attr_val" which gives the name of
   the attribute as value.  Perhaps not really a problem.
 - <plaintext> should not end with </plaintext>; can't be
   escaped.
 - <style> and <script> does not end with the first "</".


