As the name suggests, CQRS separates the responsibilities for commands and queries, namely writes and reads.

A command changes the state of the system by ultimately producing events. It is not allowed to return any data. Either the command succeeds, which results in zero or more events, or it fails with an error. The events are produced reliably.

A query retrieves and returns data, without side effects on the system. It is not allowed to modify state.

To give an example in Java code, a command acts like a void doSomething() method, which changes state. A query acts like a getter String getSomething(), which has no impact on the system's state. These principles sound simple, but have some implications on the system's architecture.

The responsibilities of the commands and queries are separated into several concerns, allowing CQRS applications to emerge in fully independent applications that either write or read. Now, how to design and implement this approach?