my garbage collector sucks

When I was testing the new WeakRef callback feature I set up a read
notification callback on an object to print out "read".  When I started the
test my screen my screen started scrolling "read" as fast as it could.  I
quickly realized my garbage collector was scanning through all the objects
reading them as it did it's mark-sweep algorithm.  This was happening so
fast my one test object was being read repeatedly.  I "fixed" this by using
a different read-lock routine in the garbage collector that doesn't trigger
the notify feature.

This gave me a graphic demonstration that my garbage collector is visiting
all the objects all the time which is going to be hell on the cpu cache when
I start to worry about performance.  Even when I optimize the collector to
run less often it will still have this obnoxious behaviour of visiting each
object up the three times per cycle.

Unless anyone can come up with any other ideas I will go to the Python
reference counter scheme, even though I hate having to keep track of the
reference count in my C code.  Someone please save me!

At the rate I'm going I'm going to end up reinventing Python!

Seriously though, finding out which parts of Python are really needed and
which parts aren't by trying the hard way is not such a bad idea.  When we
are finished with this process we will be confident we have the correct
choice for each decision.