Planned performance enhancements:
	- Better compression of X_Poly* requests by enhancing
	  dxpc to try multiple predictive models for coordinates
	  and use the one that yields the shortest encoding
	- Better color image compression (the current implementation
	  is really primitive, though fairly effective for something
	  so simple; a better algorithm and support for image depths
	  other than 8 would make dxpc useful for a wider range of
	  X applications)
	- More compact encodings for the compressed data stream
	- Profile the code and look for possible speedups

Possible usability improvements include:
	- Better integration with SSH (I suspect that modifying
	  dxpc to allow the use of a UNIX domain pipe between
	  the Client Proxy and Server Proxy would be a good idea...)
