Explicit lifecycle
Admission, running, terminal success, failure, refusal, reconnect, duplicate, resume, and capacity outcomes are messages, not timing guesses.
Development infrastructure · Chapter 78
Scenario-Driven Development makes the cycle inspectable.
Reframe treats an end-to-end scenario as a versioned contract, not a loose test script or screenshot checklist.
The cycle
Contract first, evidence last.
A scenario names prerequisites, actors, typed actions, lifecycle events, terminal predicates, failure states, cleanup, and the boundary of its claim before implementation begins.
- 01 · ContractWrite the complete journey and expected authorities.
- 02 · Source sliceImplement only the bounded behavior the scenario exercises.
- 03 · Typed executionEnter through the same operation boundary as a production software peer.
- 04 · Independent proofRead AX, visual, Store, telemetry, and provenance evidence separately.
- 05 · ProjectionPublish only the sanitized intersection of what the product proved.
Why MIDI2 matters here
Development communication becomes a protocol surface.
Older end-to-end automation often relied on sleeps, polling, private harness commands, and log interpretation. In Reframe, MIDI 2.0 lets a scenario actor participate as a software peer: it invokes typed operations and receives correlated lifecycle events over Universal MIDI Packets. This complements, rather than replaces, BDD/ATDD, contract testing, and CI.
Independent witnesses
AX proves operability, CoreGraphics/VRT proves appearance, FountainStore proves durable behavior, and telemetry explains conditions.
Bounded claim
MIDI2-native execution proves owned software-peer interoperability. Physical hardware interoperability still requires a hardware witness.
Org infrastructure
A reusable seam, product-local truth.
FountainScenarioKit is the public Swift/MIDI2 kit for generic lifecycle, UMP encoding, idempotent replay, transport boundaries, and deterministic fixtures. Reframe retains its operation namespace, Store adapter, UI witnesses, scenario corpus, and acceptance evidence.
Research boundary
Established neighbors, bounded claims.
BDD and ATDD are neighboring practices; Cucumber is an implementation tool, not scientific evidence. The Book cites the BDD systematic mapping study and a controlled TDD study, and reports no universal productivity claim.