Small things:
- find code for the missing ligatures.
- check for underscores '_' drawn as a stroke by latex.
- correctly detect and layout list bullets.
- correctly layout superscript characters.
- complete "latexsymbol" conversion table (?)
- complete accent+letter conversion table and detect above/below accents.
- check '...' symbol and replace it with dots
  (but no '...' symbol generated by latex)
- treat lots of "......" as tab with dots filling
- detect horizontal lines

Bigger things:
- fix height offsets (difficult to do for KOffice 1.2 : I need to
  tell kword the height of a paragraph but there is something wrong anyway).
- compute columns of text.
- correctly detect and layout superscript.
- correctly detect and layout underline and strokeline.
- reduce memory usage (but more computing) by only keeping the relevant data
  from first pass and discarding the page as soon as dumped on second pass.

Hard things:
- detect footnotes.
- detect chapter title/numbering.