[j-pb]

This is sad news, he invented not only the actor model, but also planner and did a lot of groundwork for deductive and parallel systems.

One of the more obscure things he pursued as a result of his early actor model work was unbounded nondeterminism as a potential source for hyper-computation.[1]

I'll always remember fondly that he pointed me to some papers via twitter when I asked him some questions about unbounded nondeterminism as an undergrad.

[1]: https://programme.hypotheses.org/files/2019/06/cardone_dayli...

[082349872349872]

If I am reading the footnoted slides properly, do they mean that we are willing to use unbounded determinism when we can take the property that "actions eventually complete" as an axiom (so we prove the rest of the system based on that hypothesis, and make the statistical argument that for those almost never cases where actions don't complete, the process is so unlucky that it "has already been run over by a bus"* anyway, to justify admissability of the hypothesis)?

* thank you Jim Gray (1944-?2007). compare the difference in liveness provability between token ring and ethernet. (but also compare their pragmatic adoption)