"A unified theory of garbage collection"?

https://dl.acm.org/doi/10.1145/1028976.1028982

Nice, will put this paper in my bookmarks for later read.

I feel like once you stop and think about it, the distinction between a 'tracing' GC and a 'non-tracing' GC is specious: To do its work at acquire and release, RC collector must in fact trace a tree of objects. It's just that the semantics of how this is declared might differ, or, more specifically the time(s) in which it does this work.

I think a fair part of the unified theory is that a tracing GC traces live objects, whereas a refcounting GC traces dead objects.

Upcounts trace live objects though. Just more incrementally.

By "trace" I mean the act of consulting the state of an object, then recursively tracing pointers in the object. A tracing GC propagates a "known to be live" state through an object graph while doing a collection cycle, and a refcounting GC propagates a "known to be dead" state through an object graph, when RC drops to zero.

Umm... but isn't the refcounting GC more deterministic? If I create a chain of 10 object and then release the head, I run through 10 allocations and then 10 release cycles on those 10 objects. Period.

Yes! Thank you, I couldn't think of the terms to search for to find this.