Message passing

Message passing is a model of communication between concurrent or distributed entities in which information is transmitted by explicit send and receive operations, rather than through shared memory. In message-passing systems, each process has its own local state, and the only way for processes to affect each other is by exchanging messages through a communication channel. This model is the native coordination mechanism of distributed systems, parallel computing architectures, and actor-based programming languages.

The message-passing abstraction hides the complexity of the underlying network — packet fragmentation, retransmission, ordering guarantees, and failure detection — behind a simple interface: send(message, destination) and receive(message). But this simplicity is deceptive. The semantics of message passing vary dramatically across systems: some guarantee FIFO ordering, some guarantee causal ordering, some guarantee only eventual delivery, and some guarantee nothing at all. The choice of semantics determines what distributed algorithms are possible and what failures are detectable.

Message passing is not merely a software abstraction. It is a structural feature of any system whose components communicate across boundaries with incomplete information. Biological systems use molecular message passing — hormones, neurotransmitters, cytokines — to coordinate cells that share no direct cytoplasmic connection. Social systems use linguistic message passing — speech, writing, protocols — to coordinate individuals who share no direct neural connection. The message-passing model, properly understood, is the universal syntax of coordination across informational boundaries.

The shared-memory model is often preferred by programmers because it feels natural — we imagine multiple threads reading and writing the same space. But shared memory is an illusion maintained by expensive cache-coherence protocols that are themselves message-passing systems in disguise. Message passing is the deeper truth: all coordination across boundaries is message passing, and the systems that embrace this explicitly are more robust than those that hide it.