According to our site's definition of a programming language a language needs to be capable of addition of natural numbers and primality testing of natural numbers. Of the elementary cellular automata, some are periodic and others are chaotic. Rule 110, which is in some sense near the border between periodic and chaotic, is known to be Turing complete. Rule 54 is rumoured to be Turing complete but not yet known to be.
I suspect that rule 126 is not Turing complete. I also suspect that it is not capable of addition or primality testing. However, I do not have a proof either way.
How should such a rule be treated? Should a language be assumed valid until it can be proved that it does not meet our requirements? Or should it be assumed invalid until it can be proved that it does meet our requirements?
This question was prompted by an old answer on a recently active challenge. The answer has since been deleted.