Yeah, that wasn't the best choice of wording. Much prefer how you phrased it. It's really just about carefully defining the behavior of the system in the face

Yeah, that wasn't the best choice of wording. Much prefer how you phrased it. It's really just about carefully defining the behavior of the system in the face of failures. Typically that means in a CP you tolerate some failures, but eventually you lose the majority and (perhaps only on one side of a network partition) and you have to sacrifice write availability at the minimum to maintain consistency - and reads might now be stale.