SIE.11:11 - SoTA-Echoing
The practice question is which receiving uses must change when one semantic premise has changed and the affected scope is still unsettled. In §5.2, both the present-availability endpoint and other consumers exist; knowing the changed field alone does not settle their reliance. The selected line combines A.10/A.10.1’s premise comparison and scoped discovery with the integration-specific returns in §4.2.
The serious alternative is a bounded full regression: identify the same receiving scope, update the tests to the new source meaning, and revalidate every registered use in it. This can be preferable when the suite is small and inexpensive or a reliable dependency analysis would cost more. It is not merely rerunning old tests against an unchanged schema.
| Answer to the same multi-use horizon change | Work and evidence at the same scope | Selection and accepted trade-off |
|---|---|---|
| Discover changed reliance and revalidate affected results | Compare the two meanings, inspect source-side and receiver-side dependencies, qualify coverage, and retest results that rely on the changed horizon. Keep applicable evidence for the independent catalogue. | Adopt A.10/A.10.1 in §4.1.1–3/5–7; adapt §4.2 and §5.2 to semantic model, mapping, interface, and validation results. This avoids repeating the catalogue’s unrelated checks and lets supported branches finish. It spends effort establishing dependencies and leaves an inaccessible consumer unresolved. |
| Revalidate every identified use in the same receiving scope | Make the same meaning and coverage comparison, then execute the appropriate receiving-use checks for every member, including those ultimately found independent. | Retain this alternative when those checks are cheaper or more dependable than selective impact analysis. It may reduce reliance on a detailed dependency model, but can repeat unaffected checks and delay a common release. It still cannot clear an unknown consumer or infer meaning from schema compatibility. |
For the worked horizon case, the established catalogue independence and separable endpoint results justify selective revalidation. The accepted trade-off is the work needed to establish that independence and the explicit limit on the unknown consumer; no universal cost or completeness advantage is claimed. §4.1.7 prevents local completions from being mistaken for a wider no-impact result. For one already known use, the direct domain result remains sufficient under §1.
LOT4KG supplies a current candidate line for changes that propagate between an ontology, mappings, graph content, constraints, and validation. Adapt those relationships in §4.1.4–5 and §4.2 when a KG realization is selected. They help locate which integration result needs work; they do not choose source truth, application action, or a universal regression policy. For other realizations, the same return question is answered from their actual semantic dependencies. Reject a changed version alone as evidence that every use failed, and an unchanged carrier alone as evidence that every use passed.
Reopen this comparison when one actual affected use was missed, when a dependency or coverage claim is defeated, or when the cost of qualifying selective impact exceeds the available full receiving-use regression. Reconsider only the affected discovery and checking choice in §4.1.3–5. Changed results still require their domain evidence whichever strategy is selected.