> How can it do this if the system is partitioned, and the node receiving the last write was unable to propagate this to other nodes? I think this is pretty sp

> How can it do this if the system is partitioned, and the node receiving the last write was unable to propagate this to other nodes? I think this is pretty spot on. You could suggest a workaround that a write must have reached a quorum of nodes before it can be considered 'written'. But then you can just make another hypothetical and find a new problem with that solution. > Is the CAP theorem really making a non-trivial claim? People still claim to beat it every now and then: https://news.ycombinator.com/item?id=39676493 https://news.ycombinator.com/item?id=39676493