logical vs physical spaces (was Re: Versioning and persistence)

Fare Rideau rideau@ens.fr
Wed, 21 May 1997 10:04:10 +0200 (MET DST)

> Most people on this mailing list want to have a single address space
> for all users.
I think that specifying such an idea would be basically *wrong*.
There should be one *logical object space*,
whose implementation is not specified.

Implementation could choose to separate objects in
separate physical address spaces or not.
For the sake of simplicity, an initial implementation
would use only one address space and do everything in software.
In more elaborate implementations,
there would certainly be multiple physical address spaces,
for various performance reasons: taking advantage of paging hardware
to trap things done in software; increase the immediate accessible
space of objects while adding some overhead for seldom-referenced objects;
not having to synchronize a large space; not having to inverse abstraction
while implementing additional semantics like real-time, persistence, etc.

We should be sure to make a difference between the official user interface,
that's meant to be unchanged for years, and the implementation hacks,
that ought to continuously adapt to whatever future technology
may improve performance.