[gclist] John Carter's one-place/one-object approach

Jon S Anthony jsa@alexandria.organon.com
Mon, 13 Jan 1997 14:03:02 -0500

> There's a major problem with your proposal, from my point of
> view. All you have done is push the gc problem off a level and back
> into the programmer's lap. To see this, suppose I have a container
> into which I start inserting objects, and I allow objects in the
> container to "name" other objects in the container, using your
> integer indexes. How do I know when a specific object in the
> container is no longer needed?

Well, the scheme (which I agree sounds a lot like linear types) does
not allow that sort of aliasing.  Recall that the objects can have at
most one reference.  That's what the "handover" stuff is about.

> whereas you make no such distinction. Hence there is no way to know which
> stored integers represent index values (and into which containers), or to

This would be very easy to solve.  Just make sure the index types are
separate types.