[gclist] Finalization and death notices

Nick Barnes Nick.Barnes@pobox.com
Tue, 09 Oct 2001 16:56:06 +0100


At 2001-10-08 14:44:30+0000, Denys Duchier writes:
> Nick Barnes <Nick.Barnes@pobox.com> writes:
> 
> > All this talk about finalization prompts me to describe the
> > finalization system we have in our Memory Pool System (MPS).  Here it
> > is:
> 
> That's pretty much the approach described in "Guardians in a
> Generation-Based Garbage Collector" (R. Kent Dybvig, Carl Bruggeman,
> David Eby, June 1993), right?  That's also the approach I implemented
> in Mozart/Oz.

I've just conferred with David Jones, who designed and implemented the
MPS finalizers, and he confirms that it is partly based on that paper.
The MPS system does not provide some of the "guardians" functionality
to clients (e.g. the guardian objects themselves), because there was
no good argument for doing so.

We believe that the MPS finalizer design documentation includes a
citation of that paper, but cannot confirm this until we have finished
extracting the documents from the Lotus Notes database which is still
holding them captive (the fact that the documents themselves are in
plain ASCII just makes this more irritating).

For those who are interested, here's a link to the paper itself:
<http://citeseer.nj.nec.com/10581.html>.  A lot of the discussion in
this thread is a rehash of the arguments presented there.

Nick Barnes
Ravenbrook Limited