The system uses an algorithm that keeps track of causal relationships between different requests. When there is a conflict between two writes (A, B) and one is determined to be the cause of the other one (suppose A is the cause of B), then the resulting write (B) is retained.

However, there can still be writes that are not causally related, i.e., requests are actually concurrent. In such cases, the system cannot make an easy decision.

We’ll elaborate more on some of these approaches later in the chapters about time and order.