From: Larry Evans (jcampbell3_at_[hidden])
Date: 2003-05-23 21:28:50
Alexander Terekhov wrote:
> Larry Evans wrote:
>
> [... Weighted Reference Counting... ]
>
> Thanks. Looks intriguing, but has some {probably severe}
> drawbacks with respect to "general purpose" application,
> I think. I'm now reading this paper:
>
> http://cs.nyu.edu/goldberg/pubs/gold89.pdf
> (Generational Reference Counting: A Reduced-Communication
> Distributed Storage Reclamation Scheme)
Thanks, I hadn't seen that. However, it is dated 1989 and was
referenced in Rodriques and Jones "Cyclic Distributed Garbage
Collection with Group Merger". I'm also wondering whether
Bacon's method ( http://citeseer.nj.nec.com/bacon01concurrent.html )
might be better. He provides a comparison with Rodgriques'; however,
he doesn't mention Goldberg. Either way, both require traversal
of the pointer graph. In boost/files/shared_cyclic_ptr/stl_container.cpp
there's a prototype of a method for doing this derived from Detlef's
method ( http://www.cs.kent.ac.uk/people/staff/rej/cgi-bin/searchbib?pattern=detlef+run+time+typing )
I could use some feedback. The html file in that directory is out-of-date; however,
the points it makes about the advantages of the Detlef method versus
others in boost are, IMHO, still valid.