[!todo] Seed note. A starting point, not a finished note yet.
PACELC, proposed by Daniel Abadi in 2010, refines the CAP theorem: if there is a Partition (P), a system chooses between Availability (A) and Consistency (C); Else (E), it chooses between Latency (L) and Consistency (C). Its value over CAP is that the Else branch is always live. Even on a healthy network, keeping replicas in agreement costs a coordination round trip, so a replicated database is trading latency against consistency at all times, not just during the rare partition. The four corners classify real systems cleanly: DynamoDB and Cassandra are PA/EL, Spanner and CockroachDB are PC/EC. It is a vocabulary for the question to ask before evaluating any database: what does this workload need in the normal case, and in the failure case? Seeded from A Mental Model for Databases.