Product Release Cut and Cut Transition

This live case tracks two provisional objects exposed by Kungfu’s release and updater work:

Does exact product-world identity require a Product Release Cut, and does authorized movement between two such worlds require a distinct Cut Transition?

The implementation separates exact identity from movement. KFD-5 must still determine whether either responsibility is independently necessary or whether existing Project Cut, Warrant, Decision, Admission, release-passport, manifest, and SemVer semantics already close the problem.

Current result

candidate: Product Release Cut
status: provisional
minimum closure: inconclusive
dogfood: implementation witness exists; qualification not run

candidate: Cut Transition
status: provisional
minimum closure: inconclusive
dogfood: implementation witness exists; qualification not run

The source evidence is Kungfu commit a2967f22d0d23766eb887168e8fffc41251bad54, especially framework/upgrade/kungfu-product-release-cut.contract.json and docs/evolution/stages/10-product-release-cut-updater.md.

Case surfaces

Claim boundary

This case allocates no KFD number and accepts neither candidate. It does not activate KFD-13, make Kungfu product terminology universal, prove historical novelty, certify a release, or establish independent adoption. A no-new-primitive outcome remains first-class for both tracks.

Genesis: Exact Product Worlds and Authorized Movement

Initial observer

The initial observer was implementing release selection and upgrade behavior. The natural objects were versions, manifests, platform artifacts, signatures, channels, installers, and rollback records.

That ontology made SemVer appear to be product identity and made updater selection appear to authorize movement.

Pressure and perspective change

Equal-SemVer successors, local dogfood artifacts, signed public supersession, rollback, conflict, and offline installation changed the consequence-bearing question:

Which exact qualified product world is this?
What exact evidence authorizes movement to or from it?

Kungfu’s implementation answered with two roots: releaseCutRoot for one product world and cutTransitionRoot for movement between worlds.

Why implementation is not qualification

The split may reflect two durable responsibilities. It may also be a local contract decomposition of existing KFD objects:

Product Release Cut may be a KFD-13 Project Cut specialization;
Cut Transition may be Warrant-bound Decision plus Admission;
release passports and manifests may already carry the evidence;
SemVer plus artifact digests may be sufficient in less demanding systems;
neither candidate may justify a new Primitive.

This case freezes those alternatives before successful release dogfood can turn one implementation into an inevitability narrative.

KFD Method Trace

KFD-1: freeze the exact source and alternatives

The two candidates are bound to an exact Kungfu commit and immutable KFD-5 cuts. Later release success cannot rewrite their genesis or erase a no-new-primitive result.

KFD-2: separate evidence from confidence

An executable contract is an implementation witness. It is not independent adoption, release certification, Primitive qualification, or proof that the split is minimal.

KFD-3: preserve participant value

Operators, release owners, installed products, and independent verifiers need the same exact identity and transition evidence without one participant’s cache, version label, or network view becoming universal authority.

KFD-4: transform the observer

The release producer sees manifests and channels. The installed product and operator bear consequences of exact identity, conflict, upgrade, recovery, and rollback. That perspective change exposes the two candidate responsibilities.

KFD-5: keep alternatives and negative outcomes live

Each candidate retains Project Cut, generic Cut, Warrant/Decision/Admission, release passport, manifest, SemVer, and no-new-Primitive alternatives. Deletion and fuse tests must hold all other roots fixed.

KFD-6 boundary

No autonomous discovery claim is made. Humans and agents reconstructed the implementation pressure and opened provisional tracks. There was no plural method experiment, fixed-ontology baseline, held-out promotion set, or transfer evaluation.

Ontology Alternatives

Proposed split

Responsibility	Candidate identity	Candidate authority
Exact qualified product world	Product Release Cut root	Declared product-profile composition only; source, artifact, signing, and qualification authorities remain external
Authorized movement between two exact worlds	Cut Transition root	Declared relation, compatibility, migration, rollback, trust, and evidence binding only

Retained alternatives

Alternative	Compression claim	Main test
Generic Cut	One fact-coordinate abstraction is sufficient	Can it preserve product qualification and platform slices without product-specific identity?
KFD-13 Project Cut specialization	Release identity is one software-project settlement profile	Does it cover installed product worlds without confusing source settlement and runtime identity?
Release passport or manifest	Existing evidence containers already identify the product	Can identity remain exact across equal-SemVer successors, local trust domains, and rollback?
SemVer plus artifact digest	Version and bytes are sufficient	Can movement authority and compatibility remain inspectable without another object?
Warrant + Decision + Admission	Existing authority objects already own movement	Does a transition add durable semantics beyond a bound decision and admitted action?
No new Primitive	Both names remain Kungfu Profile vocabulary	Does deletion change any consequential decision outside the local contract?

Forbidden fusion

Neither candidate may absorb source facts, artifact bytes, signature authority, qualification evidence, Warrant, Decision, Admission, installer execution, participant policy, or rollback occurrence. Qualification requires useful coordination while those owners remain independently inspectable.

Conditional Distinguishability

Product Release Cut deletion test

Hold source, artifacts, platform manifests, signatures, qualification, SemVer, channel, and trust domain fixed. Remove only Product Release Cut identity.

Deletion matters only if a participant can no longer answer which exact qualified product world is current, distinguish equal-SemVer successors, or bind rollback to the exact prior world without reconstructing the same object ad hoc. A convenient aggregate hash is not enough.

Cut Transition deletion test

Hold the two product worlds, Warrant, Decision, Admission, compatibility, migration, rollback, and evidence fixed. Remove only Cut Transition.

Deletion matters only if authorized movement, conflict, recovery, or rollback becomes ambiguous or requires repeated reconstruction not already owned by the fixed authority objects.

Fuse tests

Product Release Cut is not separate if a generic Cut or Project Cut profile preserves identical decisions at lower ontology cost. Cut Transition is not separate if Warrant-bound Decision plus Admission preserves identical movement and rollback consequences without hidden reconstruction.

Conversely, separation gains evidence if one exact product world participates in several valid transitions, one transition can be assessed independently of either endpoint’s identity, and endpoint identity remains unchanged when movement authorization is denied, replaced, or rolled back.

Required negative cases

same SemVer, identical Cut;
same SemVer, verified successor;
same SemVer, unsigned conflicting public Cut;
local dogfood Cut that is publication-ineligible;
diverged or unknown relation;
rollback to an exact prior Cut;
valid endpoint Cuts with an invalid or revoked transition;
valid transition evidence with a missing or mismatched endpoint.

Current verdict

Both candidates are analytically distinguishable enough to incubate but remain provisional. Minimum closure, deletion, fuse, independent adoption, and cross-domain transfer are inconclusive.

Propagation Hypothesis

If the candidates qualify, a product could preserve one common identity and movement boundary across build, publication, installation, activation, recovery, and rollback:

qualified product inputs
  -> Product Release Cut
  -> separately authorized Cut Transition
  -> verified installation or rollback
  -> retained successor identity and movement receipt

This might transfer beyond Kungfu wherever a version label is insufficient to identify exact deployed reality and movement has consequences distinct from identity.

The hypothesis is unverified. A Project Cut Domain Profile plus existing decision and admission semantics may provide the same value at lower ontology cost. A product with immutable version-to-artifact mapping may need neither candidate. Propagation that hides local trust, compatibility, rollback, or participant authority would contradict KFD-3 even if deployment succeeded.