I just (7th September 2017) found Alperin's paper when looking for something else: See here for a copy of the pdf on ResearchGate.
I was never able to find it, so I only cited the citation of it in Boyle-Lind-Rudolph. In hindsight, perhaps I could have guessed R. C. Alperin knows something about it. Though I never saw this note (that I can recall), my paper surely exists partially because I heard about this result several times, had to reconstruct a proof myself, and then took it a step further to make it worth my time.
The "automorphism group of a full shift" is of course just a fancy name for the group of reversible cellular automata. There are other kinds of groups with "automaton" in their name: in particular automata groups and automatic groups. I do not know any connections between the three (except that groups of CA and automata groups are both subgroups of the rational group). Nevertheless, these classes have similar closure properties:
Last update: 7.9.2017