                     Future plans and vague ideas


* Implement loading of encodings dynamically at run time.  Request for
  an unknown encoding would trigger loading a mapping file.  Ideally,
  this would also be implemented in the Type1 rasteriser.

* Find out whether its possible to do lazy rasterisation under X.

  This would involve switching the scheme for computing font metrics
  from precise information (deriving it from the bitmaps) to
  approximate information (trusting the font metrics).  As some
  programs (notably terminal emulators) are very sensible to this, it
  should not be done systematically.

* Change the Type 1 backend (wait for the FreeType folks to give us a
  decent Type 1 rasteriser).
