Preface

Semantic integration is professional Work that makes separately governed meanings and representations usable together for a particular receiver. Its governed object is the maintained semantic-integration arrangement: the use contract, qualified sources and needed model content, correspondences, identity and claim dispositions, executable mappings, realization, interface, validation, and their dependencies and maintenance. An ontology file, mapping table, graph database, registry, API, or pipeline can contribute to that arrangement; none is the whole by itself.

Begin from one receiver and one query, decision, operation, or engineering workflow. Work backward to the answer claims, distinctions, sources, authorities, losses, currentness, and tests that the use needs. Preserve explicit incompatibility when it is more truthful than a common value.

An integration practitioner works with source experts and the receiving engineer, analyst, application team, or other owner of the use. A successful join may still answer the wrong question: the same code can name a product family in one source and a particular item in another, or two availability fields can carry different commitments. At the other extreme, requiring a common ontology or store for every source can delay a small useful answer and suppress a difference the receiver needs. The working problem is to connect the required meanings while keeping their authority, limits, and consequences recoverable.

The main trade-offs follow from that problem. A narrower source cut costs less but may omit a claim that changes the answer. Preserving more distinctions and provenance improves what can be inspected but increases source work, interface detail, and tests. Live access can preserve a source’s control while making availability and repeatability harder to obtain; copying can support repeatable queries while adding permission, refresh, correction, and custody obligations. The contract makes these choices answerable to the actual use.

SIE.Preface:1 - Meanings, identifiers, claims, carriers, and world-side referents remain distinct

SIE.Preface:2 - Source authority and receiving authority remain separate

A source owner can define a scheme or publish a value without authorizing the receiver’s decision. A master-data steward can decide an enterprise identity without deciding product release or recall. Systems Engineering can decide configuration and effectivity without owning a cross-source mapping. Data Engineering can operate a pipeline without deciding semantic equivalence. The receiving application or professional practice owns its operational or decision outcome.

SIE makes the semantic premises and losses inspectable, tests them for the named use, and returns unresolved decisions to their direct owners. It does not borrow their authority.

SIE.Preface:3 - Pattern relations do not prescribe a lifecycle

The patterns have information dependencies, not one mandatory calendar sequence. SIE.1 and SIE.2 often expose the first stop. A practitioner may enter SIE.5 when the use and source inventory already exist, SIE.8 when a realization choice is current, or SIE.10 when an existing interface needs validation. Discovery, alignment, mapping, implementation, and testing can iterate.

SIE.3 supplies a qualified reused or developed model when adequacy is unsettled. SIE.11 follows a changed semantic premise to the results that actually relied on it; a compatible reference repair can finish directly. SIE.12 supplies the arrangements needed by actual shared-module users. A known adequate model, an unchanged use, or one local interface can continue without performing those further Methods.

SIE.Preface:4 - The first whole result

The first useful whole is SemanticIntegrationPackage@Use. It is a connected, inspectable set of eight results, not necessarily one file and not necessarily RDF or a graph:

1. use contract;
2. source manifest;
3. correspondence set;
4. cross-source identity disposition when it is load-bearing;
5. source-qualified claim composition;
6. mapping specification;
7. realization and interface;
8. validation account.

The package references any SIE.3 model qualification on which it relies. Preserve identity premises where the answer depends on them; explain their non-use only when it affects interpretation or later reliance. A package may return explicit conflict or non-comparability. Positive validation requires matching evidence for every load-bearing premise of the claimed whole use or contract-permitted subset.

SIE.Preface:5 - Use the repertoire at the scale of the missing result

The package anatomy states what must be inspectable when the promised result is a whole semantic interface package. Use the Methods whose results are missing or whose qualifications need reopening. With a supplied use contract and qualified source meanings, SIE.4 can return one warranted correspondence or an incompatibility and stop. With an existing interface and its premises, SIE.10 can identify a failed receiving-use obligation without rebuilding that interface. Reuse an available result when its subject, source editions, use, and conditions still match; reopen the contribution whose premise changed.

Some relations need separate answers even when one practitioner handles them. A correspondence supplies a relation between meanings; SIE.5 supplies a cross-source identity disposition when the answer depends on the same entity at a particular grain and interval. SIE.6 then qualifies the composition of source claims. Identity can hold while claims conflict, and claims can be compared without merging their subjects. SIE.7 specifies executable behavior from those premises; SIE.8 compares ways of supplying it; SIE.9 carries the qualified result into the receiver’s work. Their results constrain one another, while a defect can return to any supplying pattern.

A bounded interface package needs adequate models, its other load-bearing premises, and the required implementation and validation evidence. The five applications demonstrate interface and shared-module uses. Their differences change the work: AP242/QIF needs configuration and effectivity premises; semiconductor traceability makes identity grain and issuer rules central; an analytic can fail on a unit conversion or hidden default; live provider comparison may need qualified difference and non-comparability instead of a shared value. The shared-equipment commons adds module decisions, dependencies, and semantic change. Each application uses the same repertoire at the scope of its receiving question.

If a required distinction cannot be expressed, SIE.3 reuses, extends, or constructs the needed semantic content. An unresolved source definition still limits the dependent mapping or interface claim. SIE.11 compares changed reliance and obtains the affected domain results; its wider account is conditional on a receiving need. SIE.12 supplies shared-module maintenance and decision arrangements where actual users require them. The source, implementation, and receiving owners retain their respective returns.

SIE.Preface:6 - Qualify the combined arrangement

Three questions remain distinct. Is each source meaning, correspondence, identity disposition, or composition rule supported? Does that result permit the particular use made of it? Can all relied-on results and their realization satisfy the receiving contract together? A true relation may be too lossy for one transformation. Two individually current sources may concern different effectivity intervals. Several feasible components may exceed a shared access limit. The relevant bodies answer their local questions; the whole-use conclusion also needs these joins and common conditions.

For the combination, bind the same receiving question and the actual subjects, grain, intervals, source and rule versions, accepted losses, authority, and failure branches. Where contributions use shared requests, credentials, storage, time, or operating support, compare their total demand with the applicable conditions, including required trace and recovery behavior. A permission to query does not imply permission to replicate or disclose. Keep a common source or derivation visible when several results rely on it; repeated citations or passing tests of the same premise are not independent support for a different claim. The comparison belongs in SIE.8, Solution and the bounded validation in SIE.10, Solution.

SIE.Preface:6.1 - A shared request limit changes the provider result

Consider a constructed variation of APP-SIE-04. The receiving contract allows an explicitly incomplete purchasing answer. Assume that adequate models and the required source, correspondence, identity, and composition premises have been supplied. Provider A still means “on hand”; provider B means “available to promise”; their claims must keep those meanings. Permissions allow live retrieval and the agreed immediate presentation, but prohibit replication.

For this illustration, four calls retrieve A’s required rows, four retrieve B’s, and three retrieve the common source and rule trace required even for a partial answer. All calls consume one gateway allowance of ten requests in the agreed request window; none is shared or counted twice. Retries also consume the allowance, and the contract accepts neither a copied cache nor deferral to another window for this answer. Each pair of contributions fits: 8, 7, or 7 calls. The complete answer requires at least 11 before retries, so pairwise feasibility does not establish the whole arrangement.

SIE.8 therefore cannot select that arrangement as supplying the complete result. Under the stated incomplete-answer permission, an A-only result with the required trace uses seven calls before retries and can be proposed with B explicitly marked as not retrieved under the request limit. SIE.9 must preserve A’s on-hand meaning and the incomplete branch; it must not render B as zero stock or call the response a complete provider comparison. For example, an A source value on_hand = 12 observed at T becomes a receiving row 12 on hand at T, with its source and rule trace and the branch B not retrieved. It supplies no available-to-promise value.

SIE.10 tests the remaining required conditions before any bounded positive validation. If both providers are mandatory, return a stop and the exact missing result: for example, an access-owner decision changing the allowance or an implementation that demonstrably reduces calls while preserving meanings and trace. Omitting provenance is not an equivalent repair.

The quantities are construction assumptions, not measurements of providers. Actual use needs evidence for the gateway condition, call demand, permissions, freshness, behavior under failure, and the receiver’s interpretation. Passing source lookups or several pairwise tests cannot supply that evidence for the whole. A changed allowance reopens the affected realization and use validation; an unchanged correspondence can remain usable.

Constituent actions in ongoing work. While processing an exchange, resolving a source identifier can constitute part of interpreting a correspondence, within an ongoing integration of the receiving information. If the receiving use changes from present eligibility to eligibility at an earlier date, a lookup that returns only the present subject state is no longer sufficient. The team may have parser and database skills but lack the intermediate meaning or temporal-identity account. Supply that contribution through the relevant SIE Method rather than treating a successful query as a successful integration. FPF B.1.5.EW helps recover the connection; SIE.8 and SIE.10 keep the whole arrangement and its warranted validation in view.

SIE.Preface:7 - Architectural Rationale

The language is organized around the results that make a receiving use possible. Source recovery, relation truth, bounded identity, claim composition, executable semantics, realization, interface, and validation can fail independently and return to different owners. Keeping their Methods directly accessible permits a useful early stop and replacement of one contribution without inventing a new lifecycle for the entire arrangement.

The complete repertoire joins model adequacy, semantic interfaces, changed reliance, and commons maintenance through their actual results. Completing a small model or interface question remains useful on its own. A continuing service requires its operational results, and a commons requires the rights and dependencies of its actual users.

SIE.Preface:7.1 - Source contributions behind the arrangement

The source-use account gives the qualified source cut and its dates. The shared architecture combines contributions that answer different questions; it does not treat a standard or tool family as a complete integration Method.

Terminology and registry practice, including ISO 704:2022 and ISO/IEC 11179-3:2023, makes source objects, concepts, definitions, designations, items, and versions distinguishable. SIE.2 adapts that contribution into the smallest inventory needed by the receiver. This improves on schema inspection plus informal recollection: a mapping can point to the operative definition and edition. It still cannot infer a cross-source relation from a registry entry. FPF’s source-local meaning and direct Bridge distinctions supply the separate generic questions used by SIE.2 and SIE.4.

SKOS supplies different correspondence forms; SSSOM 1.0 makes endpoints, predicates, justification, provenance, and source versions inspectable. OAEI 2025 contributes task-dependent alignment evidence. In SIE.4, SoTA-Echoing, SIE.4 uses those contributions to separate candidate generation, direct relation judgment, and a receiving-use qualification. A two-column crosswalk or score can help find a candidate, but the chosen relation, permitted loss, and counterexamples determine whether it can support this transformation. A new endpoint sense or defeating counterexample reopens that row.

ISO 8000-115:2024 contributes identifier ownership, semantics, restrictions, and resolution inputs within its declared scope. PROV-O supplies distinctions for derivation, attribution, revision, specialization, and alternate descriptions. SIE.5 and SIE.6 adapt these into two different results: an identity disposition and a source-qualified claim composition. A merged key cannot replace the first, and provenance cannot settle the second’s truth or authority. The SEMI and GS1 traceability sources in SIE.5, SoTA-Echoing make grain, issuer, and event-time failures concrete; they do not authorize a recall or establish an enterprise master identity.

R2RML and QVT 1.3 contribute bounded declarative mapping and transformation forms. The Ontop line demonstrates a virtual realization over mappings. SIE.7 and SIE.8 adapt those contributions into an independently inspectable semantic rule and a choice among whole arrangements. This costs explicit specification and comparison, but it permits an implementation to change while preserving the required behavior. Neither RDF nor MOF nor a graph store is required. Failure of a mapping premise returns to the semantic rule; an implementation discrepancy returns to its implementer.

SHACL supplies declared graph-constraint tests. DQV and the data-quality sources in SIE.10, SoTA-Echoing contribute dimensions, measurements, and process questions. SIE.10 retains their distinct evidence roles and adds the representative receiving-use replay. A passing shape can coexist with the wrong identity or a concealed stale branch. The arrangement therefore uses categorical pass, narrow, unresolved, and stop results under the contract, rather than a compensating overall score. Evidence for implementation and operation remains necessary when the conclusion relies on actual service behavior.

LOT contributes ontology requirements, development, publication, and maintenance to SIE.3. Sufficient model reuse remains a completed result. LOT4KG distinguishes ontology work from graph construction and relates changes to dependent mappings, constraints, and validation; SIE.11 uses those relationships for graph realizations. OBO Foundry principles contribute scoped module, term-stability, maintenance, and communication practices to SIE.12, with their community-specific rules qualified in the source account.

Reconsider the affected choice when a simpler qualified contribution supplies the same result, a represented use repeatedly needs a missing Method, or a source or case defeats a relied-on boundary. When the need is a reusable Method repertoire, ME.2 returns inspectable alternatives, source contributions, relations, and gaps for that comparison; the integration-specific question stays in SIE. A substantially independent practice remainder can reopen the field split described above.

SIE.Preface:8 - Costs, perspectives, and correction

The gain is an answer whose meanings, sources, qualifications, and unresolved branches can survive into receiving work. The cost is source recovery, explicit relation and loss judgments, trace, implementation evidence, and representative tests. Keep that burden proportional to what can change the answer. A single correspondence question does not require a service architecture; a claim of a usable whole interface cannot omit a load-bearing identity or provenance obligation merely to stay cheap.

The source cut reflects the receiver’s question and the sources practitioners can inspect. A well-documented schema or a familiar formal vocabulary can receive more attention than an inaccessible local rule. Source experts and receiving users may also differ over which distinctions matter. Make missing access, authority, meanings, and perspectives visible in the contract and inventory instead of treating absence as agreement. The five constructed applications illustrate how to work; their coverage limits supply no empirical claim about production performance or effectiveness in other settings.

Correct the result at the point that owns the defect. A hidden default returns to SIE.7; an unsafe interpretation of a partial response to SIE.9; unsupported identity to SIE.5 and the required domain authority; a pipeline discrepancy to Data Engineering. For the AP242/QIF case, SYSE.13 supplies the decision-specific configuration basis, including actual subjects and effectivity; SIE consumes those premises and can return a mismatch, but does not decide the engineering configuration. The other owner boundaries remain in force. Preserve unaffected qualified contributions and retest what the correction can change.

SIE.Preface:9 - Before relying on the whole result

Recognition asks which available pattern can supply the next useful result. Assurance asks what supports the actual reliance. Use the checklists in the selected bodies and inherit their answers only while the subject, content, use, source versions, and relevant conditions match. For a combined package, answer these questions in the account the receiver needs:

• Can the receiver recover the question, answer claims, accepted losses, authority boundary, and useful stop? An early correspondence or inventory result must not be presented as a validated interface.
• Can every load-bearing correspondence, identity disposition, and claim-composition premise be traced through the mapping and interface, with its grain, interval, limits, and unresolved branches?
• Does the whole arrangement meet shared semantic, access, resource, currentness, provenance, and failure conditions? Which implementation or owner result is still absent?
• Does each evidence item support the claim made from it, and does the representative receiving-use test cover positive, negative, and unlike cases? Several passing layers cannot compensate for a hard stop in another.
• Is a narrow result recognizable as narrow, with excluded or unexamined branches and their consequences visible? In the request-limit example, an A-only answer cannot stand for the complete provider comparison.
• Do a model gap, changed semantic premise, or shared-module decision reach SIE.3, SIE.11, or SIE.12 where needed? Are source, implementation, and receiving decisions returned to their owners, with unresolved dependence made visible?

These questions address the failures in the applications: field-name equivalence, merged identifiers without authority, source claims flattened into one value, healthy transport mistaken for semantic fitness, and missing branches hidden from the receiver. Repair the specific premise or interface, obtain the missing contribution, narrow under the contract, or stop.

SIE.Preface:End