SIE.5 - Resolve Cross-Source Identity without Erasing Identifier Authority
Type: Method pattern Status: Eternal alpha Normativity: Normative method guidance within SIE; examples are constructed and non-normative.
Primary working result: a
CrossSourceIdentityDisposition@Usethat states, for each load-bearing endpoint pair or set, the required entity grain and interval, identity-related disposition, source identifier schemes and issuers, evidence, authority boundary, unresolved cases, and the exact mappings, claim compositions, interfaces, and tests that may consume it.
SIE.5:1 - Problem Frame
Use this when an integration needs to decide whether records or identifiers from different sources concern the same entity, a part, a version, a variant, an alternate description, a family, or different or unresolved entities. The recognizable failure is a join, merged key, or sameAs assertion that makes identifiers appear unified while erasing their issuers, grains, intervals, and authority.
The primary EntityOfConcern is one bounded cross-source identity question for a named use. The first move is to state the entity grain and time or effectivity interval at which sameness matters. The first result is a disposition that downstream work can inspect and challenge without replacing source identifiers or master-data authority.
The payoff is that the integration can join only warranted cases, return exact unresolved branches, and preserve source trace. Do not use SIE.5 to create identifiers, choose enterprise master identity or survivorship, decide authoritative values, or prove a source claim true. Those decisions remain with identifier owners, MDM, or the direct domain authority.
SIE.5:2 - Problem
The same real-world item can carry manufacturer, supplier, enterprise, lot, serial, package, event, and temporary process identifiers. Conversely, the same character string can be reused by different issuers or at different times. Version, configuration, part-whole, and alternate-description relations can be mistaken for identity because they support some joins.
If an integration collapses those distinctions, provenance becomes decorative: the merged record no longer shows which identifier asserted what, when, or under whose rules. Contradictory events appear to concern one thing without evidence, and a master key silently becomes a domain decision.
SIE.5:3 - Forces
| Force | Tension |
|---|---|
| Joinability | A common key makes computation easy, while premature merging can create false identity and irreversible provenance loss. |
| Grain | Receivers want one “item”, while wafer, die, device, package, lot, feature, version, and event answer different questions. |
| Time and effectivity | Identity may be stable across one interval, while reuse, replacement, split, merge, and revision change the warranted claim. |
| Authority | Identifier issuers and master-data stewards govern different schemes and decisions; crosswalk custody transfers neither authority. |
| Incomplete evidence | A bounded decision is needed, while some pairs must remain unresolved without blocking independent rows. |
| Reuse | A disposition can help several mappings, while its grain, interval, and use boundary prevent blanket equivalence. |
SIE.5:4 - Solution
Frame identity as a source-qualified, grain- and interval-specific disposition for one receiving use. Preserve every source identifier and issuer. Compare the entity criteria, relation evidence, version/effectivity, and authoritative decisions that matter. Return same, different, part, version, variant, alternate-description, family, or unresolved only at the precision supported; do not force every useful relation into identity.
SIE.5:4.1 - Pattern-Use Unfolding
- Name the receiving identity question. State which mapping, claim composition, query, or decision needs an identity premise and what a wrong merge or split would change.
- Fix entity grain and interval. Name the relevant individual or occurrence kind, part-whole level, version/configuration, lifecycle state, jurisdiction or organization, and time/effectivity interval.
- Bind source endpoints. Reference each
SIE.2identifier scheme, issuer/owner, syntax, resolution behavior, restrictions, identified-entity claim, edition, and source provenance. - Separate candidate relation kinds. Consider same individual at the stated grain, different individuals, part/whole, earlier/later version or specialization, variant, alternate description, family membership, or unresolved. Do not use one boolean until the use truly needs only that distinction.
- Collect discriminating evidence. Use issuer records, source assertions, configuration/effectivity, provenance, physical or domain characteristics, event histories, and counterexamples. A shared string or co-occurrence is only candidate evidence.
- Recover authority. Name who may decide scheme-local identity, master identity, authoritative values, or the domain relation. SIE may record and use a bounded disposition; it does not acquire those authorities.
- Assign the narrowest warranted disposition. State conditions, confidence or uncertainty where meaningful, rejected alternatives, and unresolved evidence. Preserve source IDs even for accepted same-entity rows.
- Propagate the premise explicitly. Reference the disposition from every mapping, claim composition, interface row, and identity test that relies on it. Never copy only the merged key.
- Test difficult cases and reopen. Include identifier reuse, split/merge, version/configuration change, part-whole confusion, contradictory source events, and absent-link cases. State the source or observation change that reopens the row.
SIE.5:4.2 - Record the Result
| Identity position | Required content |
|---|---|
| identity question and use | receiver, dependent action, wrong-merge/wrong-split consequence |
| entity boundary | kind and grain, part-whole level, version/configuration, interval/effectivity, relevant context |
| source identifiers | scheme, identifier, issuer/owner, syntax/resolution/restrictions, source edition and provenance |
| candidate relations | same, different, part, version/specialization, variant, alternate description, family, unresolved alternatives |
| warrant | evidence, counterevidence, counterexamples, uncertainty/confidence, rejected alternatives |
| authority | scheme-local, master-data, domain, and receiving-decision scopes |
| disposition | narrowest warranted relation or unresolved result, conditions, source-ID preservation rule |
| downstream and reopen | mapping/composition/interface/test references and exact change or observation that reopens |
SIE.5:4.3 - What Changes in Practice
The integration stops minting a common key as proof of identity. Every load-bearing join points to a disposition with grain, interval, source identifiers, evidence, and authority. Unresolved identity becomes an explicit interface and validation branch instead of a row silently dropped or merged.
SIE.5:5 - Archetypal Grounding - Semiconductor Recall Trace
A recall-triage query crosses a wafer identifier from a fabrication source, die coordinates, device serials, package identifiers, supplier lot identifiers, GS1 event identifiers, and an enterprise material master. The receiving question is whether a shipped device can be traced to an affected fabrication lot.
| Endpoint question | Constructed disposition |
|---|---|
wafer W-204 and supplier lot L-88 | not identity; the supplier record asserts that the wafer is a member/output associated with the lot under a bounded process interval. Preserve both identifiers and relation provenance. |
die coordinate (x17,y9) on wafer W-204 and device serial D-771 | same physical die/device at the selected post-singulation grain only when the packaging event and issuer records connect them; accepted with interval and event evidence. |
device serial D-771 and package code P-55 | unresolved because the package identifier is reused after rework and the event history lacks effectivity. Stop the dependent trace branch. |
enterprise material master M-12 and device serial D-771 | family/type membership, not individual identity. The master describes a material/product class and authoritative enterprise attributes within its stewardship scope. |
GS1 event E-901 and device D-771 | event and object are different entities; the event may concern the device. Never merge their identifiers. |
Tests cover reused package identifiers, a die split/merge error, contradictory supplier events, a missing cross-enterprise link, and a later correction. The interface returns supported trace segments and the unresolved P-55 branch. MDM or the domain steward may later decide a master identity; SIE.5 has not done so.
SIE.5:6 - Bias-Annotation
| Lens | Likely drift | Repair |
|---|---|---|
| Governance | The integration registry becomes the de facto master-data authority. | State the bounded disposition and retain MDM/domain decision ownership. |
| Architecture | One canonical key is required because the platform prefers it. | Carry a surrogate technical key only with source IDs and the identity disposition it indexes. |
| Ontology/Epistemology | Same individual, part, version, variant, alternate, and family relations collapse. | Fix grain and interval, then choose the narrowest warranted relation. |
| Pragmatics | Every identifier pair receives a costly adjudication. | Decide only pairs that are load-bearing for the receiving use; leave independent rows untouched. |
| Didactics | “Same” is read as timeless universal identity. | Display conditions, interval/effectivity, scheme, and forbidden reuse beside the disposition. |
SIE.5:7 - Conformance Checklist
- The receiving action and wrong-merge/wrong-split consequence are named.
- Entity kind, grain, part-whole level, version/configuration, and interval/effectivity are explicit.
- Every identifier retains its scheme, issuer/owner, restrictions, source edition, and provenance.
- Useful non-identity relations are not compressed into a boolean same/different field.
- Shared strings, co-occurrence, and crosswalk membership are treated only as candidate evidence.
- Evidence, counterevidence, counterexamples, and uncertainty are inspectable.
- Scheme-local, master-data, domain, and receiving-decision authority scopes remain separate.
- Accepted same-entity rows do not delete source identifiers.
- Every dependent mapping, composition, interface, and test references the disposition.
- Reuse, split/merge, version/configuration, part-whole, contradiction, and absent-link cases are tested where applicable.
SIE.5:8 - Common Anti-Patterns and How to Avoid Them
| Anti-pattern | Repair |
|---|---|
| Same string, same entity | Add scheme, issuer, grain, interval, and discriminating evidence. |
Universal sameAs | Use the narrowest relation and record unresolved or different cases. |
| Master key by integration fiat | Return master-identity and survivorship decisions to their authority. |
| Version equals object | State whether the use concerns one continuing entity, a version-specific configuration, or a relation among them. |
| Part joined as whole | Fix the entity grain and preserve part-whole relations explicitly. |
| Drop unresolved rows | Expose them as interface and validation branches with dependent actions. |
SIE.5:9 - Consequences
The package can join warranted identities without losing source trace and can stop only the branches that depend on unresolved identity. Corrective source or steward decisions can be replayed against explicit consumers. The method also reveals when the useful relation is part, version, membership, or event participation rather than identity.
The cost is domain evidence and authority work. Some convenient global keys remain technical indexes rather than semantic truths, and some desired traces return incomplete.
SIE.5:10 - Rationale
Identifiers are governed signs, not entities. Cross-source identity is therefore a relation claim that depends on scheme, grain, interval, evidence, and use. Preserving source identifiers and authority makes the claim reversible and prevents integration convenience from becoming a master-data or domain decision.
SIE.5:11 - SoTA-Echoing
The best-known line combines explicit identifier ownership and restrictions, provenance-qualified alternate/specialization relations, and domain traceability practice. The serious default is entity resolution that optimizes a match score or golden record. Its defect is not probabilistic matching itself; it is allowing the score or merged record to erase grain, issuer, interval, and decision authority. SIE.5 mutates the line into a bounded disposition consumed explicitly by downstream semantic results.
| Source line | Adopt, adapt, or reject | Role and limit |
|---|---|---|
| Current FPF identity and relation law | adopt | Keeps identity and direct relations as explicit claims; generic law does not decide a domain pair. |
| ISO 8000-115:2024 | adapt | Contributes identifier owner, use restriction, semantics, and resolution inputs; excludes identifier creation, query/response syntax, and resolution method and does not decide equivalence. |
| PROV-O | adapt | Supplies specialization, alternates, derivation, attribution, revision, and invalidation relations useful to evidence; provenance relations are not blanket identity. |
| SEMI traceability standards and GS1 EPCIS 2.0.1 | adapt | Supply semiconductor and event-based identity/traceability probes; scheme use does not create enterprise master identity or authorize recall. |
| golden-record or score-only merging | reject as sufficient | It may support candidate generation or a steward decision but cannot erase source identifiers, relation kinds, and authority. |
Reopen when an identifier scheme or issuer rule changes, new event/configuration evidence changes the disposition, the required grain or interval changes, or repeated cases show that the disposition vocabulary cannot preserve an action-changing relation.
SIE.5:12 - Relations
SIE.1supplies the receiving action, identity-sensitive answer claims, loss boundary, tests, and stop.SIE.2supplies identifier schemes, issuers, restrictions, source claims, editions, provenance, and authority gaps.SIE.4may expose an identity question but does not answer it merely by accepting a semantic correspondence.SIE.6consumes identity dispositions when composing claims about the same bounded entity; it preserves contradictory claims separately.SIE.7andSIE.9reference dispositions rather than copying only merged keys.SIE.10tests every load-bearing identity premise. MDM and domain authorities retain master identity, authoritative values, and action.