change subscribe address

Andrew Valencia vandys@cisco.com
Tue, 06 Dec 1994 12:40:49 -0800


[rideau@mathp7.jussieu.fr (Francois Rideau) writes:]

>"rideau@frmap711.mathp7.jussieu.fr"
>Could I be just subscribed as "rideau@clipper.ens.fr" (that is myself) ?

Sure.

>... What I'm looking for is a completely decentralized system,
>where inter-object communication would be automatically inlined as much as
>possible when possible (and wishable), at compile-time link-time, load-time
>or run-time, that is, when wishable (sometime between the moment it is
>made possible and the moment it is made necessary).

I looked at the deeply object-oriented OS designs.  Having to drop C is a
bitter pill for a from-scratch OS effort, since it guarantees you have to
write a mind-numbing amount of code from scratch.  Even if you preserve
POSIX interfaces with some other language binding you're paddling upstream.

Even more alarming was the track record of such efforts.  The pattern always
seemed to be early successes followed by efforts increasingly bogged in the
contradictory requirements of coherency, scalability, transparency,
security, fault tolerance, and so forth.

To even further generalize, the systems would tend to grab a great deal of
the abstraction, and then let it dribble back piecemeal as deeply difficult
problems with a system-wide policy interacted with individual application
needs.

Not to nay-say your effort.  Clearly, I believe that experimentation with
operating systems is always a Good Thing.  But since you posted to the VSTa
list, I felt I should offer an explanation for the relatively nuts-and-bolts
type of microkernel I ultimately implemented.

Finally, I believed that the VSTa microkernel can provide the basic engine
for certain implementations of such distributed operating environments.
Thus, I was not ruling out more powerful implementation technologies for the
future, although admittedly this is not an active area currently.

						Regards,
						Andy Valencia