ARC 4Integrate Application EngineeringChapters 15–18

Arc 4 Recap — From Collection APIs to Application Engineering (Chapters 15–18)

The application layer now depends on a fulfillment coordination port. DSM locators, typed handles, Spring assembly, and test evidence each have a clear owner.

Engineering boundary

FulfillmentService
  → FulfillmentCoordination (business port)
    → DSM adapter (failure translation)
      → composition root / Spring auto-configuration
        → Runtime + named typed handles

A fake supports fast business-decision tests while the adapter still runs against a real Runtime. Spring Boot is another composition root and does not alter Register, Lease, or CRDT semantics.

Four distinctions

Distinction	Left responsibility	Right responsibility
domain port / DSM handle	business action and result	collection operation and technical failure
bean name / locator	select a bean in one container	distributed collection identity
clusterId/serviceId / locator	communication and sync domain	collection and state domain
test pass / production readiness	evidence inside a fixed surface	deployment, capacity, security, and operations acceptance

How this arc realizes DSM's value

The domain port contains collection APIs, the composition root centralizes locator/codec/lifecycle, Spring Boot provides testable assembly, and the evidence ladder limits every conclusion. This structure reduces integration and evolution cost while keeping DSM types, identity, and failure semantics out of business services.

Transfer checklist

Does a business service import com.leanowtech.dsm.* outside a composition root or adapter?
Does every locator have one source of truth and a schema ID?
Does a Lease result retain uncertain, reason, and fencing token?
Does Spring fail fast on duplicate locator, missing codec/merger, and invalid Lease settings?
Is a cluster/service identity change handled as protocol migration?
Does each test layer record proves and doesNotProve?

Evidence in this arc

The book example provides an E1 domain-port test and E2 real-Runtime test.
Spring auto-configuration fixes Runtime, three collection beans, configuration failures, and lifecycle behavior.
Runtime integration fixes locator isolation, CRDT behavior, and Lease fencing.
TwoNode/Chaos fixes E3 replication, repair, and partition behavior.