Circular references and python

[Michael Hudson]

"Tim Peters" <tim_one at email.msn.com> writes:

> [posted & mailed]
> 
> [nascheme at enme.ucalgary.ca]
> > I have been looking at Toby Kelsey's code [a gc patch posted to
> > c.l.py a while ago].  It seems quite interesting.  Do you think
> > something like it would have a chance to be accepted by Guido?
> 
> Yes.  I've studied the code (at Guido's suggestion, actually), and
> corresponded with Toby about it.  Toby probably invented this approach on
> his own, but Rafael Lins has published a great deal about it under the name
> "lazy cyclic reference counting".  It has many attractions!  Like it's
> (well, can be made) 100% portable in strictly std C, builds on the current
> rc (which many people like) rather than replacing it, is theoretically
> sound, and extension modules that don't play along can't cause incorrect
> behavior (this is a real problem for Python! any scheme has to work well
> with what's already out in the field).

Where might one find this code? A moderately serious deja & web search
failed to turn it up...

Aha! It's in the pipermail archive. It doesn't seem to have made it to
USENET; at least it isn't on deja, and I don't recall ever seeing it.

Is it on the web anywhere? I'm currently downloading 4 megs of archive
so I can rip it out - pipermail has HTML-ized the uuencoded attachment
- which seems a bit unnecessary.

Cheers,
Michael