Great post!
A few more positive shiboleths.
One would be eventual consistency.
Another would be discussing write paths vs read paths (or patterns) and recognizing that those can be decoupled (or a mention of CQRS).
A few more positive shiboleths.
One would be eventual consistency.
Another would be discussing write paths vs read paths (or patterns) and recognizing that those can be decoupled (or a mention of CQRS).
In the vein of the TFA, this would be a negative shibboleth (which just goes to show how being pedantic in certain contexts is just silly). The reason being that eventual consistency has no guarantee on what "eventual" means. If your replicas converge ten years from when a change is made, you can (correctly) claim to have eventual consistency.
Eventual consistency can be a rational choice or a quasi-necessity, but in practice it's a shortcut for careless optimists; I wouldn't expect them to analyze how inconsistent states can go wrong, whether they are an acceptable risk, and how to deal with abnormal situations.