KPComm is a library for communicating with KPlot.  RLab will use
patched to use it.  We are not using pipes because we don't want to be
slow.  (Pipes have been tried, and -- yes -- they're too slow.)  We
use this instead of simply hacking into RLab because:
	- That's messy:  This way rlab only needs a patch.  If I'm lucky
the patch will become part of the Rlab distribution.
	- This is more flexible:  Other interfaces to KPlot can be built
using KPComm.  In particular, I'll probably start out by building a simple
file (matrix) loader to test KPComm then do the RLab hacking.  A Perl
wrapper for KPComm might be nice, too.  Then one could write Perl scripts
to process and plot data files.

	

A pipe _will_ be used for control messages and shared memory will be
used to data.
All KPlot commands should be available to KPComm: plot, print,
 plot options, export files, show help, etc.
KPLot's UI should be made rather configurable (menu bar entries, tool
 bar buttons, etc.) and these options should be available to KPComm.
KPlot should have installable features.  That means that, for example,
 a click of a toolbar button sends a message back to the KPComm client
 so that KPlot's GUI can be used to perform arbitrary functions.  A good
 example would be to create a toolbar button which sends the command to
 RLab to generate the plot of choice.  Then, of course, RLab would do
 so by sending the matrix to KPlot through KPComm.
