JAnderson Wrote: > My thought was you could use Ref counting for shared memory and mark and > sweep for unshared. That's a really good idea. Mark and sweep (as currently implemented in D) can be very expensive with multiple threads.