Received: from ATHENA-AS-WELL.MIT.EDU by po7.MIT.EDU (5.61/4.7) id AA00746; Thu, 4 Feb 93 10:44:17 EST
Received: from CECI.MIT.EDU by Athena.MIT.EDU with SMTP
	id AA02356; Thu, 4 Feb 93 10:44:16 EST
Message-Id: <9302041544.AA02356@Athena.MIT.EDU>
Received: from ITHAKE.MIT.EDU by ceci.mit.edu id AA13476g; Thu, 4 Feb 93 10:44:23 EST
To: aybee@Athena.MIT.EDU
Cc: jud@ceci.mit.edu
Subject: application classes and scope
Date: Thu, 04 Feb 93 10:47:44 -0500
From: Judson Harward <jud@ceci.mit.edu>

	I had one of those flashes @ 3 am, but now in the morning's
light, I'm not sure that it adds that much.

	The kernel of the idea is that each application object should
specify a separate process.  That implies a single ADL file could generate
multiple processes.  Classes defined outside of the application class
are visible to both applications but no objects or data members can be defined
outside of an application object.  A corollary to this is that libraries should
be filled with module, not application definitions.  Application objects
defined by files read by a single parser are automatically register with each
other and will run on the same workstation (although in the multiprocessor
case, not on the same CPU).

	What do you think?  Does this clarify and give the application object a
clearer role?

						Jud
