Re: Unrecovered memory leak thoughts.

> The bigger issue is that extension writers would have to rewrite their
> extensions to increment the reference count when they do an assignment.

And possibly decrement with their finalizers, and provide a traverse
method if they desire.  I guess the good news is that today's
extension's 'finalizer' methods could be rewritten to use dec instead
of mark, and could also basically double as traverse methods, so, as
you pointed out, the biggest issue would be that extension writers
would need to learn to increment when they do assignment.

My latest thought is that some type of hybrid approach might be
possible, as well--retain mark and sweep to get rid of cycles, and use
reference counting for common, trivial cases.  But, like you pointed
out, it would be a change that requires extensions that retain
pointers to the heap (not their own internal pointers, just pointers
to ruby objects on the heap) to be rewritten.