• The concept of message passing originated in the late 1960's, when general-purpose ILLIAC-IV multiprocessor architecture (and supercomputer precursor) was built and the first computer network - ARPANET was cretated (1969).
• The multiprogramming nucleus developed by Per Brinch Hansen in 1970 for the Danish RC4000 computer was the early proposal of the message-passing system supporting client/server communication using a shared pool of fixed-length buffers.
• In the earliest models of concurrency, processes communicate through global shared variables - program variables that can be read and written by all processes.  Initially, the shared variables were accessed by the ordinary program operations of expression evaluation and assignment; later variations included synchronization primitives such as semaphores and monitors to control access to shared variables.  Global shared-variable models provide a natural representation of multiprocessing on a single computer with one or more processors connected to a central shared memory.
• Most operating systems for single processors and shared-memory multiprocessors use shared variables for process interaction within the OS. However, these OS provide also message passing primitives so that processes can communicate with other machines on a network (e.g., Windows and Unix sockets, remote procedure calls)
• Of necessity, OS for distributed memory multiprocessors (multicomputers) provide and use message passing primitives.

It is more efficient to communicate using shared variables than by message passing, for the price of simplicity and ease of implementation.

• In the term distributed computing, the word distributed means spread out across space. Thus, distributed computing is an activity performed on a spatially distributed system.
• The models of computation generally considered to be distributed are process models. Other models, such as Petri nets, are usually not studied under the title of distributed computing, even though they may be used to model spatially distributed systems.
• The process models that are most obviously distributed are ones in which processes communicate by message passing - a process sends a message by adding it to a message queue, and another process receives the message by removing it from the queue.
• The boundaries, when we should use shared-memory or message passing primitives are not sharp. If processors and memories are spatially distributed, then it is more natural use message passing.



However, message passing and shared memory primitives are equivalent - whatever can be done using shared memory primitives can be done using message passing and vice versa.