Shared Libs.

[physh at berkeley]

   	 I find the idea fairly simple to implement on a VAX.  You could
make them system calls,  or have the C compiler map them to a jump table.
This undoubtably adds a little overhead, but if you have a 750 with little
or no disk space most of the time, and no core left, and everything swapping
like mad, you try and find ways of conserving.  I think this can be
implemented straitforwardly on machines that have more then two protection
modes (Vaxen have 4) provided that you can always read the data-segment.

	As for the problem of people using strange "features" and "bugs"
to an advantage, well, I always found that you fix bugs and features
you find, not use them to an advantage.  Some one will get gotten, probably
you. 

	Everything that was created before the shared library will be
compiled, therefore wont use it.   If it needs to be recompiled to fix
the bugs that probably showed up from using "features" to advantages,
they can dam*-well be fixed so that they are portable.

	If its too old to be fixed (poor code with no particular structure
in mind, like UUCP) you can use the old compiler, linker, and libs.  This
is the way we currently retrofit V6 trash.

	Anyway, I'm going ahead with shared librarys.  If anyone has any
real good implementation ideas for me, send to to addresses listed below.

					Jon Foreman
					College of Marin
					ucbvax!amd70!notavax!physh
					physh at LBL-Csam