advanced 1 min answer Multiple choice

PACELC extends CAP with an "else" branch. Why does that branch matter more in day-to-day operation?

cappacelclatencyconsistencycoordination
Pick one
Show the full answer Hide the answer

What is being tested

Whether you recognise which half of the trade-off you actually live with.

The reasoning

CAP describes behaviour during a partition: consistency or availability, pick one. Partitions happen, and for most systems they happen rarely.

PACELC adds: else, latency or consistency. That is the normal case — every request, every day.

A system that chooses consistency in the "else" branch pays a coordination round trip on every write. Cross-region, that is 60–150 ms added to each one, forever. Over a year that cumulative cost usually dwarfs the partition behaviour everyone argues about in design reviews.

So the more useful question is not "what happens during a partition" but "what does this cost me when everything is working?"

The decision that actually resolves it

Asymmetry of failure cost, per operation.

A false decline is recoverable — the customer retries, the terminal falls back. A double-spend is unrecoverable and scales with the duration of the problem.

Operation Choice Why
Authorisation against a shared balance CP Double-spend cannot be undone
Low-value contactless under a floor limit AP Bounded, insurable loss
Recording a completed settlement AP Append-only; no shared invariant
Fraud scoring AP, degraded A stale model beats no decision

Card networks are explicitly hybrid: offline stand-in processing exists because the network chose AP with a value cap for a defined class of transaction. The cap is the risk control that makes availability affordable.

The practical trap

Two regions cannot form a quorum. A two-region active-active deployment with no witness in a third location has designed a system that cannot make a consistent decision during a partition — regardless of which letter it claims.