Dynamic Consistency Boundaries

A Dynamic Consistency Boundary (DCB) is how EvidentSource lets you enforce cross-entity invariants at append time, atomically, without distributed transactions and without saga orchestration.

The idea is formalized at dcb.events. This page is a practical orientation.

The problem

Classic event sourcing works well when every command touches one entity. It stops working neatly when a command needs to enforce a rule across several entities — “transfer between accounts A and B only if both have sufficient funds at the moment the transfer lands.”

Traditional answers: distributed transactions (complicated, slow, lock-prone), sagas (eventual consistency, compensation logic, operational complexity), or single-entity aggregates that absorb everything they might need to touch (denormalization that fights the model).

The DCB answer

When decide returns events, it also returns conditions — predicates expressed against the event store that must hold at the exact moment of append. EvidentSource checks them atomically; if any fail, the append is rejected and decide can be retried.

return {
  events: [{ type: "TransferApproved", subject: transferId, data: { ... } }],
  conditions: [
    // No new debits on source account since we read its state
    unchangedSince(sourceAccountId, sourceState.revision),
    // No new debits on destination account either
    unchangedSince(destinationAccountId, destinationState.revision),
  ],
};

You get:

• Atomicity across multiple entities
• Linearizable read-your-write consistency
• No coordinator — the condition check is part of the append, done in the event-store core
• No compensation logic — either it all lands or none of it does

The predicates

EvidentSource supports predicate primitives over:

• Stream revisions (unchangedSince, atRevision)
• Subject last-event predicates
• Event-type presence/absence within a subject or stream
• Time-bounded versions of each

See Reference: Constraints for the full list.

When to reach for DCB

Any time a command’s correctness depends on state you read and that state could change between your read and your append. DCB conditions are how decide says: “This decision was made assuming X — if X isn’t still true, don’t apply it.”