TODO (Most important items are listed first)

- Use only PyXML http://pyxml.sourceforge.net for the parsing and processing.

- Get rid of ElementTree because ElementTree serialization, even if valid,
  causes problems to XSLT processors like xsltproc.

- Get rid of minidom parsing, because minidom is reported to have serious memory
  leaks.

- Rework the way we gather metadata from the OOo document.
