There are a significant variety of “bottoming out” of messages. There is a partial ordering among these but not a simple ordering nor even a most fundamental, I think. I think that the adopted foundation for messages that we take subtly influences our designs. This is not bad but may causes difficulties in understanding other peoples designs. I collect here some notes on these issues with hopes that they might become organized.
Butler proposes that recipients can be trusted to ignore messages that should not have been sent. This precludes solving the confinement problem which Butler first defined in print. Lets walk thru the Granovetter process and see how close it comes. Initially Bob accepts messages from Alice by knowing and accepting messages with Alice’s number, Caller ID, as it were. Alice knows Bob’s number in order to send the message. Alice and Carol are connected just as Alice and Bob are. Alice sends a message to Carol advising Carol to accept messages from Bob and then sends Carol’s number to Bob.
One might presume a non user replaceable comm layer in each object that enforced message behavior as described. Indeed that describes several distributed capability schemes. Here is the partial ordering of functions in this case;