As expected from Daniel Abadi, this is good stuff. The cross over between isolation levels and consistency levels adds a lot of complexity, which is a real prob

As expected from Daniel Abadi, this is good stuff. The cross over between isolation levels and consistency levels adds a lot of complexity, which is a real problem when most developers don't deeply understand isolation levels already. Many distributed SQL databases (e.g. Cockroach, spanner) default to strong consistency and strong isolation. This is the right thing to do, because it saves developers from having to reason about the complex interactions between consistency and isolation. They also typically support a "recent snapshot" mode (often called "consistent prefix" in literature) for read-only, which can be good for latency optimizing read only workloads which can tolerate a little bit of staleness, while still not having to think about weak isolation (snapshot on read only transactions is equivalent to serializability).