Common Misconceptions About CAP
You'll learn: The most widespread misunderstandings about CAP Theorem and what it actually means for your system design.
The "Pick Two Forever" Myth
The biggest misconception is that CAP forces you to permanently choose two properties and abandon the third for your entire system. People say things like "MongoDB is a CP database" or "Cassandra is an AP database" as if these labels are fixed.
The reality: CAP describes behavior during network partitions, not your normal operating mode. Most of the time, when your network is healthy, you can have all three: strong consistency, high availability, and the network is connected. CAP only applies when partitions actually occur.
It's About Network Partition Moments
Think of CAP like an emergency protocol. It's like asking: "When the bridge is out, do you wait for repairs (sacrifice availability for consistency) or take the detour (sacrifice consistency for availability)?" During normal times, you use the bridge and don't choose at all.
During normal operation (no partition), a well-designed distributed system delivers both consistency and availability. You only face the tradeoff when nodes can't communicate.
CAP Isn't Binary
Another misconception: systems are either "CP" or "AP." In reality, as you learned with "CAP as a Spectrum," different operations in the same system can make different choices. You might accept eventual consistency for user profiles but demand strong consistency for financial transactions.
CAP also doesn't mean "zero availability" or "zero consistency"—it means you prioritize one during partition scenarios.
Key Takeaway: CAP doesn't force you to sacrifice a property forever; it describes the tradeoff during network partitions only. Outside partitions, systems can be both consistent and available—CAP is an "emergency mode" decision, not a permanent architecture label.