Resolution proofs viewed as automata
This note presents a slightly different view of resolution proofs
which the author has not seen presented alsewhere. A proof of
unsatisfieability of a (CNF) formula S is viewed as a kind
of automaton which, given a truth-assignment f, produces
a clause C in S such that f(C) is false.
This view leads to a mild generalisation of resolution proof; it is
shown, however, that any generalised proof can be converted
to a resolution proof which is at least as short.