@techspokes/typescript-wsdl-client 1.1.5 → 1.2.0

package/docs/content-model-s06-re-01-resolver-implementation.md

@@ -1,869 +0,0 @@
-# S06 RE01 Research Resolver Implementation
-
-Research implementation scope, verification and the future production seam. See the root [README](../README.md).
-
-## Implementation contract
-
-### Architecture recovery source staging
-
-The #232 recovery assignment permits reusable internal source owners before the complete existential theorem.
-`src/compiler/semanticContext.ts` adapts the real canonical/resolved graph to one immutable request and isolated candidate deltas under `RE01-domain-boundary-v1`.
-Its `structural-only` tag is explicit: this boundary implements the accepted B helper universe without certifying component/schema validity.
-
-`prepareSemanticRequest` preserves original records, source contexts, AU/group/anonymous identities and legal element/type recursion.
-The request's `actual` view and `candidate(delta)` share one budget; candidates have explicit membership, additions and an optional coherent endpoint replacement.
-Original records remain available through `sourceNode`, while ordinary lookup returns `found`, `missing` or `outside-context` without silent actual fallback.
-
-Opaque runtime context authentication rejects forged handles.
-Reference, owner and incoming-affiliation access to an unavailable subject raises `SemanticContextAccessError`, rather than reporting a false empty collection.
-Candidate collisions are checked against the entire original identity domain, including omitted original records.
-
-`src/compiler/semanticBudget.ts` is the single implementation of the previously accepted request accounting; `resolver-budget.ts` is a compatibility re-export.
-Default inclusive limits are 100,000 nodes and 125,000,000 work units, with sticky exhaustion and pre-work/pre-allocation charging; the [scalar calibration](#scalar-completeness-capacity-calibration) records the measured scope and headroom.
-The new request covers actual context preparation, candidate indexes and consuming kernels; inherited loading/catalog/resolution stage counters remain separate and are not relabeled one whole-pipeline budget.
-
-Semantic consumers reuse the request's original indexes and access generated content through `particleContentNode`.
-Occurrence analysis receives a compact graph containing only the transformed particle and its referenced declarations.
-`ParticleContentView.graph` is a lazy, charged diagnostic projection; normal restriction, extension and endpoint correspondence do not materialize or reindex the whole context.
-Independent scaling controls keep a fixed query unchanged while adding unrelated original components, and measure preparation separately.
-
-Source scalar/type/declaration and original-AU kernels consume authenticated contexts; no production source imports research implementations.
-Accepted DT01 A arithmetic and PW R-240 control flow have shared source owners with research compatibility adapters.
-Their exact local premises, implementation coverage and remaining aggregate obligations are recorded in the companion AU/PW/DT records and GitHub leaf handoffs.
-
-The private `SchemaValueContext<Owner>` boundary shares datatype, list and ordered-union evaluation between component operands and source syntax.
-Its diagnostic owner is generic; source assessment does not fabricate graph nodes or borrow application-schema lookup authority.
-The fixed schema-for-schemas pattern consumer shares XML name ranges with the loader, while application patterns retain their existing approved subset.
-
-Attribute-use facts are retained only within the exact authenticated context, source/raw role and original use identity.
-A completed fact includes its declaration dependencies and scalar obligations; later calls replay that complete evidence with positive request charges.
-Container uniqueness, requiredness, restriction and wildcard checks still execute for each assessment, and failed or partial facts cannot become successful cache entries.
-
-The [selected raw-AU contract](content-model-s06-re-01.md#selected-raw-matching-implicit-order-and-generated-scalar-properties) requires one compatible base use when matching same-name uses, retaining the prescribed wildcard allowance for unmatched raw uses.
-It preserves required-base coverage and wildcard checks; source replacement retains universal C1.
-Alternative raw matches can discharge a pending pair only with a fully checked compatible pair and its complete dependencies.
-
-The canonical adapter retains source contributions.
-Those contributions do not by themselves represent unrestricted abstract hypothetical AU sets or raw particle construction; source mapping must not be applied to every hypothetical candidate by assumption.
-The concrete finite checker described below consumes qualified effective-property correspondence and all generated local obligations.
-Missing-input and separately owned aggregate qualifications remain explicit under the [domain checkpoint](content-model-s06-re-01.md#architecture-recovery-domain-and-proof-checkpoint); universal witness/evidence adequacy remains #234.
-
-### Concrete source verification APIs
-
-`prepareSourceSchemaRequest` consumes a loader-authenticated `SchemaInput`, builds/resolves the actual graph once and binds the complete retained original source syntax to the new request.
-Graph-only `prepareSemanticRequest` remains useful structurally, but cannot supply missing source-document qualification to complete verification.
-`forkSemanticCandidate` constructs an authenticated direct child of the actual or hypothetical parent and preserves original source evidence separately from its active raw properties.
-
-`SemanticCandidateDelta.rawParticleWildcards` supplies complete namespace/process properties for genuinely new ANY particles, including the abstract finite empty namespace set.
-The [representation contract](content-model-s06-re-01.md#abstract-particle-wildcard-properties) retains first-birth identity protection, canonical set semantics, source/native-origin authority and every inherited fresh root.
-`rawParticleWildcardProperties` reads the context-owned tuple; semantic consumers share `readParticleWildcard` rather than interpreting the lexical diagnostic placeholder.
-The semantic projection supplies selected empty-alphabet viability to the existing occurrence owner without changing ordinary canonical analysis defaults.
-
-The internal `loadSemanticSchemaInput` entry shares the existing loader implementation and applies XML Schema whitespace mapping to XSD namespace and schema-location values before interpretation, comparison and cache keys.
-It preserves original syntax, bytes, digests and namespace bindings; `loadSchemaInput`, WSDL transport behavior and the public defaults are unchanged.
-Import referents must match both namespace presence and value, while an explicitly empty import namespace without a supplied referent is not itself forbidden.
-
-Notation declarations use a private component table without adding a persisted canonical node kind.
-`sourceSchemaBinding.ts` binds complete source interpretations to the exact resolved handle produced from one authenticated immutable input; copied handles cannot supply that authority.
-The source-owned preparation path now freezes the authenticated loader handle under its shared budget and keeps its exact private binding. The resolved graph was already deeply frozen by canonical construction, so the same authenticated source path reuses it when constructing the semantic request. Generic catalog callers still receive an owned graph snapshot. The source handle becomes immutable before canonical construction, and no source or graph data escape as writable request state.
-`schemaNotationProperties.ts` maps original XML values and checks complete abstract property tuples through one shared implementation.
-Source representations with the same notation QName coalesce only when their complete public/system identifier properties agree, including absence versus an empty value.
-Every original representation still receives its own source grammar, duplicate-root-key and annotation checks.
-
-Schema source wildcards use the fixed schema-for-schemas assessment universe, independently of the application schema being constructed.
-`schemaSourceGrammar.ts` owns global and parent-selected local declaration grammars; `schemaSourceValues.ts` evaluates their values through the shared scalar implementation and privately issued fixed type handles.
-Known named types, contextual child declarations, schema-instance attributes, seven fixed schema keys and physical-root ID/IDREF obligations have concrete owners.
-Unknown foreign declarations follow lax assessment against the complete fixed inventory, including the contemporaneous October 2004 XML import.
-Typed payload IDs participate in the same physical validation-root ID set; an inert pattern payload's string value does not construct a schema regular expression.
-The authenticated source loader forbids DTDs, so source ENTITY/ENTITIES declared-value assessment fails against its known empty unparsed-entity set; application scalar obligations remain separate.
-XML `id` contributes typed ID facts to the same physical validation root; a missing IDREF target is rejected after complete assessment.
-Schema-location hints are typechecked without fetching or extending the fixed assessment universe.
-The dated assessment episode uses the contemporaneous October 2004 XML import's `xs:language` declaration, so literal empty and whitespace-only `xml:lang` values fail shared scalar assessment, including in payload subtrees.
-The [immutable dated XML schema](https://www.w3.org/2004/10/xml.xsd), [October 2004 installation record](https://lists.w3.org/Archives/Public/public-xml-core-wg/2004Oct/0030.html) and [independent inventory review](https://github.com/TechSpokes/typescript-wsdl-client/issues/271#issuecomment-6114584589) establish the complete imported inventory.
-It declares `lang: xs:language`, `space` as an anonymous `xs:NCName` enumeration of `default` and `preserve`, `base: xs:anyURI` and `id: xs:ID`; `specialAttrs` references only base, lang and space.
-There are no named types, global elements or further imports/includes, so other XML names follow known-absence rules in their respective symbol spaces.
-The primary evidence is a complete rendered XML tree and contemporaneous approval/install linkage, not a raw-source byte attestation; runtime assessment fetches no schema.
-
-The fixed grammar records ID-typed attribute uses independently of whether those attributes occur.
-Under `cvc-complex-type` 5.2, a wildcard `xml:id` is invalid on a type with such a use, including schema and annotation types even when their ordinary `id` is absent.
-Types without that use, including documentation and lax ur-type payloads, assess `xml:id` through the shared ID owner; duplicate values remain invalid across the physical root.
-
-`checkCandidateCorrespondence(actual, parent, candidate, target, intermediate, ancestor)` computes the designated ancestor and verifies E extends A, D′ restricts E and every original effective property is preserved.
-Original particle/AU identities are exact; only authenticated source-generated composition wrappers admit a bijection preserving kind, occurrence, child order and sharing.
-Equal normalized languages do not establish raw property correspondence.
-Original scalar operand type, owner and context remain explicit, and scalar prerequisites propagate into the final checker.
-
-`verifyFiniteRe01Evidence(actual, target, evidence)` accepts a root ID and finite records containing exact parent/candidate contexts, target, intermediate, ancestor and child-evidence IDs.
-The checker constructs the obligation inventory itself, checks every actual XSD component and required original source obligation, and closes hypothetical dependencies from E, D′ and all fresh components.
-Every nonprimitive nested RE01 requirement must match supplied evidence in that exact parent context.
-Missing children remain unresolved, active proof cycles cannot pass, unrelated extra proof nodes are invalid evidence and component scheduling does not prove a judgment.
-
-Notation availability and active membership have separate completeness conditions.
-The finite checker inventories the authenticated notation table, checks every fresh declaration including unused helpers, and records `actualNotations` and each proof step's `notations` in their exact context.
-Pure property checks may transfer only through the same privately shared immutable property tuple; source syntax and active membership remain independently checked.
-
-Successful privately authenticated `RE01-finite-evidence-v1` receipts report checked actual obligations and the supplied construction.
-Their `universalCoverage`, `completeNegativeAnswer` and `searchTermination` fields are always false.
-Candidate rejection, malformed evidence, unresolved local rules and sticky resource limits remain separate from a universal negative answer.
-No public API, persisted format or activation guard consumes these internal receipts automatically.
-
-`proveRe01FreshAffiliationObstruction(actual, parent, target)` in the existing correspondence owner supplies a separate sufficient contextual contradiction.
-Its `proved-no-construction` result identifies a mandatory fresh affiliate whose fixed complex head is unreachable in every permitted two-step rebase; `not-established` supplies no verdict.
-The [proof and exceptions](content-model-s06-re-01.md#a-terminating-contextual-ancestry-contradiction) retain primitive no-change, original-intermediate identities and complete context-local obligations; this API does not assess actual legality or implement general witness search.
-
-### Borrowing an unchanged SOURCE ur-type wildcard
-
-`deriveOriginalUrWildcard(actual, wholeHandle, particle)` selects one private native ur-type wildcard from an authenticated actual SOURCE whole-content view.
-Its opaque handle identifies the unchanged `##any`, `lax`, `0..unbounded` leaf, including its actual backing and ur-type origin.
-An ordinary wildcard with the same lexical properties does not acquire that origin.
-
-A candidate imports these handles through `originalUrWildcards`; descendants inherit them monotonically.
-The borrowed leaf remains outside active component membership, while every new sequence, GROUP and wrapper is a genuine registered component and receives its complete FULL check.
-Containment, model-group and RAW content views, declaration consistency and FULL traversal resolve the same authenticated terminal leaf.
-No caller origin flag, changed backing, raw wildcard override, active addition or colliding stored identity is accepted.
-
-This supplies the missing construction for an actual SOURCE prefix embedded inside a new sequence.
-Reify its unit sequence wrappers as genuine new particles and retain only the authenticated ur-type leaf.
-The whole SOURCE handle still authenticates only its original root; a new wrapper receives no invented original root identity.
-Borrowing provenance does not introduce a FULL duty for the leaf's original storage owner, and actual SOURCE qualification remains independently required.
-
-The authority stock is finite: enumerate the native ur-type leaves exposed by actual input WHOLE views, cached by exact leaf identity.
-A candidate selects from that input stock; new compositor and GROUP code remains in the ordinary registered construction.
-
-The [native controls](../test/research/re01/source-ur-wildcard-borrowing.test.ts) construct a complete RE01 forest for a proper prefix with an additional tail, including a registered GROUP variant and inherited-root replay.
-Exact whole-content and prefix-only alternatives fail their required extension or restriction comparisons; a plain copy loses the origin needed by restriction.
-Missing, forged, foreign and changed handles, active admission, override, collision and containment-cycle controls preserve the owner boundary.
-This closes the stated representation gap; arbitrary-witness coverage, scalar reconstruction, complete negatives and mathematical termination remain separate obligations.
-
-### Local rule and evidence map
-
-This map identifies the shared implementation and independent controls.
-The leaf issues retain exact reviewed revisions and unresolved findings; a mapped function is not a claim that all applicable premises pass for every input.
-
-| Rule or property | Shared owner and entry | Independent controls and scope |
-| --- | --- | --- |
-| Complete B availability, collision domains, source identity, original/fork isolation | `semanticContext.ts`: `prepareSemanticRequest`, `forkSemanticCandidate` | `semantic-context-independent.test.ts`, `derivation-domain.behavior.test.ts`; structural authority only |
-| Fresh abstract particle namespace/process properties | `schemaParticleWildcardProperties.ts`, authenticated context tuple table | `particle-wildcard-properties-independent.test.ts`, complete abstract-wildcard RE01 and source-reference controls; source origin and formal/realized occurrence remain separate |
-| Unchanged SOURCE ur-type leaf inside new registered particles | `complexTypeContent.ts`, `semanticContext.ts`, local FULL and declaration-consistency owners | `source-ur-wildcard-borrowing.test.ts`; authenticated origin, complete forests, inherited roots and rejected forged or colliding imports |
-| Inclusive request limits and immutable input/output | `semanticBudget.ts` | Context and finite-checker controls; sticky terminal result, charged snapshots and no getter/iterator invocation |
-| Actual document grammar and retained source universe | `schemaDocumentSyntax.ts`: `prepareSourceSchemaRequest`, `checkSchemaDocumentSyntax` | `schema-document-independent.test.ts`; complete notation property reconciliation, separate source syntax and namespace-mapping prerequisites |
-| Fixed assessment-schema declarations, values and XSI | `schemaSourceGrammar.ts`, `schemaSourceValues.ts`, `schemaSourceWildcards.ts` | Source and XML-inventory independent controls; contextual grammar, independent assessment universe, opaque handles, typed keys/IDs and complete dated import |
-| Original and fresh notation component properties | `schemaNotationProperties.ts`, authenticated context notation table | `semantic-notation-independent.test.ts`; source authority, active membership, collision reservations, original operands and unused fresh obligations |
-| Actual component XML grammar, attribute placement and ordered children | `schemaSyntaxAssessment.ts`: `checkActualSchemaSyntax` | Actual-source syntax controls; abstract helpers have no invented XML representation |
-| Complete builtin/simple/complex derivation, exclusions and final | `schemaTypeRelations.ts`: `checkTypeDerivation`, `checkImmediateTypeFinal`, `immediateTypeBase` | `schema-type-relations.test.ts`; component validity remains a separate premise |
-| Declaration type/value legality, affiliation versus effective substitution | `schemaDeclarationPrimitives.ts`: `checkDeclaration`; shared type-relation owner | Declaration controls and `particle-edc-independent.test.ts`; complete declared affiliation dependencies survive effective filtering |
-| Lexical/value/facet/list/union semantics and exact DT01 A | `schemaScalarKernel.ts`, `scalarSchemaAssessment.ts`, `schemaDatatypeValues.ts`, `schemaExactCalendar.ts` | `schema-scalar-independent.test.ts`, shared DT01 suite; original datatype, lexical namespace, owner, selected union member and prerequisite context retained |
-| Original generated scalar references and independent slot closure | `semanticTypeReference.ts`, `schemaTypeIdentity.ts`, `schemaGeneratedScalarShape.ts`, `checkOriginalGeneratedScalar` | `schema-original-generated-binding-independent.test.ts`; identity, all reference positions, suppressed source, exact context and private receipts |
-| Original simple property and facet correspondence | `checkCandidateCorrespondence`, shared scalar cross-context comparison | `schema-simple-correspondence-independent.test.ts`, `schema-cross-context-facets-independent.test.ts`; original typed values, fixed and annotation properties |
-| Exact supplied facet components and evaluation ownership | `scalarSchemaAssessment.ts`: `layer`, `readFacetProperties` | `schema-exact-facet-components-independent.test.ts`; fixed tuples, annotations, shared versus copied operands, evaluation order and inherited enumeration |
-| Selected P1 expression correspondence and F2 scalar restriction | `scalarSchemaAssessment.ts`: `samePatternProperty`, `exactScalarRestriction`, `rawFacetCorrespondence` | `schema-scalar-policy-independent.test.ts`, `schema-scalar-inheritance-review.test.ts`; exact inventories, compatible matching, complete value containment, universal fixed duties, source inheritance and partial calendar order |
-| Source AU C1 identity union, replacement, requiredness, fixed/default and augmentation | `schemaAttributeUses.ts`: `assessSourceAttributeUses`, `assessAttributeUse` | `schema-attribute-independent.test.ts`; omitted original identity versus explicit replacement, typed equality and absent-value conflicts |
-| Context-local AU reuse with complete evidence replay | `schemaAttributeUses.ts` | `schema-attribute-reuse-independent.test.ts`; complete declaration/scalar closure, source/raw role separation, aggregate checks and inclusive warm-hit exhaustion |
-| Scalar failure ownership through AU and aggregate consumers | `schemaAttributeUses.ts`, `schemaConstructionAssessment.ts` | `schema-attribute-origin-independent.test.ts`; original datatype invalidity versus a candidate bad value |
-| Literal raw AU construction and namespace/process algebra | `assessRawAttributeUses`; `wildcardAlgebra.ts` | Raw and selected-AU independent controls; existential compatible matching, complete requiredness/wildcard checks and unchanged source C1 |
-| Actual effective content versus explicit raw content | `complexTypeContent.ts`: `deriveActualComplexContent`, `deriveRawComplexContent` | `complex-content-independent.test.ts`; generated sequence identity, scalar identity, ordered source contribution and immutable source operands |
-| Restriction normalization and exact formal ranges | `particleSemanticView.ts`; the single `occurrenceAnalysis.ts` | `particle-restriction-independent.test.ts`, existing occurrence suite; declared and normalized views remain separate |
-| Particle restriction dispatch, ordered/unmatched mappings and R-240 | `particleRestriction.ts`, `particleWildcardRelation.ts` | Restriction, selected-order and wildcard controls; one coherent implicit-group enumeration per root query with explicit order preserved |
-| Raw particle extension and full component-property equality | `particleExtension.ts`: `checkRawParticleExtension` | `particle-extension-independent.test.ts`; raw order and original/equal-property cases, without restriction normalization or assumed transitivity |
-| All placement, UPA and EDC across particle positions | `particleLocalAssessment.ts`, `particleAttribution.ts`, `particleDeclarationConsistency.ts` | `particle-upa-independent.test.ts`, `particle-edc-independent.test.ts`; repeated group uses, original named types, effective substitutions and nested-repeat counterexamples |
-| Aggregate simple/mixed/empty/particle construction and content-dependent defaults | `schemaConstructionAssessment.ts`: `checkComponentConstruction` | `component-construction-independent.test.ts`; concrete local scalar, final, emptiability, particle, declaration and AU premises with exact dependencies |
-| Original syntax mapping to no semantic component | `isAbsentOriginalSourceComponent` | Prohibited/zero source descendants versus raw max-zero particles; referenced global declarations retain semantic obligations |
-| Original effective endpoint correspondence | `re01CandidateCorrespondence.ts`: `checkCandidateCorrespondence` | `raw-type-properties-independent.test.ts`, `re01-candidate-independent.test.ts`; caller provenance labels and copied receipts have no authority |
-| Complete finite nested obligations and primitive identity leaf | `finiteRe01Verification.ts`: `verifyFiniteRe01Evidence` | `re01-candidate-independent.test.ts`; actual obligations, complete supplied/nested evidence, missing/cyclic evidence and all universal claims false |
-
-Every component receipt separates `obligations` from concretely `discharged` predicates by exact rule, component and context ID.
-Original-context scalar prerequisites are discharged only against that actual closure; a candidate receipt for the same component ID cannot replace them.
-Privately typed facet operands also require membership in the restriction's base value space; their original primitive family alone is insufficient.
-Finite inherited enumeration components provide concrete positive or negative membership evidence.
-Additional finite lexical aliases are reassessed against the exact base and compared through shared typed equality; a failed lexical witness alone does not prove a value-space negative.
-The shared pattern owner can also enumerate its privately compiled finite literal fragment: literal code points and escapes, choices, sequences and bounded repetitions.
-This historical positive producer proposes spellings without changing the application pattern subset or assigning pattern-component equality; every spelling still requires exact base assessment and typed equality.
-Enumeration is iterative and charged before copying or repetition; resource exhaustion remains sticky, and fragments outside this producer retain their membership prerequisite.
-For admitted explicit character classes, the positive producer tries operand spelling characters and printable ASCII that match the class, then scans remaining code points from U+007F for one match when needed.
-Exact finite pattern words across all layers are checked before incomplete heuristic proposals, so a costly earlier pattern cannot starve a later finite witness.
-Within a heuristic pattern, cheap class witnesses remain ahead of large literal branches, while paths needing no class search are tried before the fallback Unicode scan; the [independent starvation control](../test/unit/schema-pattern-starvation-independent.test.ts) exercises both orderings with an explicit 1,000,000-work allowance.
-For admitted unbounded repetitions, it explores a finite prefix from the declared minimum through that minimum plus the expression and operand spelling lengths, using the same request budget.
-These proposals are checked in the exact base context against the original typed operand; their prefix and class choices do not exhaust either language. The complete local membership owner now continues with the typed inverse/image procedures below when those positive proposals miss.
-The same value-facet evaluator can establish exact length, precision or bound exclusion without promoting one union member's failure to a union-wide negative.
-Coexisting length and minimum/maximum-length facets require the specified proper restriction ancestor, including intrinsic builtin list bounds, in addition to numeric consistency.
-Raw conjunct reduction retains a correspondence prerequisite when it cannot preserve fixed metadata or an inherited distinct facet property; it does not silently discard that property.
-`schemaBuiltinFacets.ts` supplies the dated derived-builtin component tuples, including exact singleton pattern strings, fixed markers, token-list minimum length and integer bounds and precision.
-Primitive facet sets remain empty; mandatory primitive whitespace behavior is distinct from a facet component, while list construction has its separately required whitespace component.
-Raw helpers must supply their complete effective properties; local raw atomic values use the primitive plus those supplied properties, and derivation/correspondence legality is checked separately.
-A complete exact raw facet inventory must retain every kind in a known, unqualified inherited inventory. The reviewed nonfixed same-side inclusive/exclusive normalization belongs only to the legacy `facetLayers` representation.
-Supplied properties lost during reduction retain their prerequisite rather than being mistaken for absent input.
-Verifier-matched exact inherited raw facet tuples remain inherited properties, with their fixed metadata and typed values checked through shared equality.
-For exact scalar sets, changing any tuple of a kind replaces that whole kind in the reconstructed local set S. Every newly supplied bound must satisfy its local base-membership rule. Explicit source redeclarations undergo those same local rules, and source grammar still permits only one scalar facet of each kind per step.
-Unused original model groups receive their own all/UPA/EDC checks, and every fresh component is a hypothetical root.
-The approved pattern subset and schema-only NOTATION/identity/ENTITY obligations retain the distinctions required by ADR-003; runtime payload enforcement remains #184.
-
-Implicit substitution choices use a finite search over coherent enumerations within one restriction query, with immutable blocks preserving explicit list positions through flattening.
-Separate branches referencing the same implicit group cannot choose contradictory orders; distinct queries may choose independently.
-Search and alternative mapping preserve unresolved premises, try fully checked alternatives, and charge before work or copying; incomplete search returns sticky resource exhaustion.
-
-GSC1 assigns each source simple-content restriction slot its own anonymous scalar identity, including a restriction without facets; extension inherits that exact identity.
-The shared scalar property owner supplies the containing type's mapped namespace, empty own final, absent own type annotation and complete inherited/base/facet properties.
-Property equality traverses those properties with the same scalar equality owner and retains legality prerequisites; it does not equate anonymous identities or use synthetic graph nodes as scalar authority.
-The content projection carries the authentic original scalar selector directly; it no longer fabricates a simple-type graph node.
-
-`SemanticTypeReference` adds this private selector without changing canonical or persisted `Reference`.
-The same selector is available in raw scalar bases, list items, ordered union members, complex simple bases and content, declaration types and typed facet operands.
-One declared-type accessor honors the explicit raw declaration type; a canonical placeholder cannot override it.
-One original shape mapper supplies immutable original identities and properties, while the scalar checker reassesses their legality in the selected active context.
-
-Original generated slots have their own context membership, separate from the containing complex type's membership.
-Syntax suppressed by a zero particle or prohibited attribute use creates no scalar slot; an otherwise real original slot cannot be invented or silently restored after omission.
-Structural subsets may retain a real slot independently of its containing type, but full finite verification requires complete original slot availability.
-Actual-schema checking covers every original slot; hypothetical checking covers slots consumed by E, D′, fresh components and their concrete dependencies.
-Availability alone does not turn an unrelated original slot into a hypothetical legality root.
-
-Finite receipts keep generated scalar construction evidence separate from graph-component evidence.
-Type dependencies identify both their context and their component, builtin or generated-slot identity; each must have matching semantic construction evidence in that context.
-Source-syntax-only evidence cannot discharge type validity, and a containing complex-type receipt cannot discharge its generated scalar's own rules.
-Ordinary inline types suppressed by source mapping also remain diagnostic syntax rather than selectable semantic types.
-Raw properties cannot turn such syntax into an original component or make its selection an actual-schema invalidity.
-
-Complete raw properties for an original simple type are compared with its immutable source properties through the shared scalar owner.
-The comparison preserves base, item and ordered member identities, typed facet values, fixed metadata and annotation properties or an exact original facet-component binding; an equal component ID alone does not prove correspondence.
-An authenticated original facet binding preserves its original recipe/component authority and source operands, while active type and value legality remain separate obligations.
-Its native nonexact clearance alone does not establish equality to an unreduced complete tuple inventory; the existing normal-form qualification and independent property comparison still apply.
-The stronger exact-inventory claim remains an open contract/normal-form reconciliation, not a discharged constructor theorem or an authorized weakening of P1/F2.
-Missing annotation metadata is unresolved unless authentic component identity establishes correspondence; explicit absence is data, and independently constructed multi-operand groups retain any unsettled mapping prerequisite.
-
-Exact supplied facet components retain their typed values in the owning scalar description before effective constraint reduction.
-Enumeration operands belong to their installed constraint layer, including when a descendant inherits that layer.
-Reusing the same supplied facet object in another datatype cannot change earlier equality or membership results; independent controls cover both evaluation orders and copied-object contrasts at the unchanged request limits.
-For an inherited exact raw row, the scalar owner preserves each component occurrence through a private token for that authenticated immutable row and ordinal.
-Distinct rows, repeated incidences and independent births do not share this identity merely through caller data or type IDs.
-Only the same incidence's annotation and immutable pattern spelling can use it; actual typed values, full formation and context-qualified dependencies are still computed independently.
-The [complete inherited-facet controls](../test/research/re01/inherited-raw-facet-identity.test.ts) keep every fresh root and required child; this is not a whole-description or new-alias binding.
-
-Enumeration value-set coverage and component-introduction validity are separate checks.
-The scalar owner associates an exact raw enumeration component with its own installed layer; same occurrence or determinate equal annotation can establish retention only after typed values match.
-Otherwise every member must have a complete immediate-base membership witness.
-Unknown retention with unproved new formation retains the existing scalar prerequisite, while successful new formation suffices without resolving annotation equality.
-Source and original-bound fallback metadata comes only from that layer's own facets and existing issued identity.
-The [complete enumeration controls](../test/research/re01/enumeration-retention-annotation.test.ts) preserve both contexts, all inherited fresh roots, required children and missing-evidence failures.
-
-Raw group import copies documented fields before any internal identity processing.
-An extra caller `identity` field cannot populate the private layer-identity map or bypass pattern/value comparison.
-Exact component annotations remain data; genuine original-recipe-issued identities retain their separate authenticated path.
-The [raw-group authority controls](../test/research/re01/raw-facet-group-authority.test.ts) distinguish matching patterns from differing patterns under shared, distinct, primitive and absent caller hints in complete contexts.
-The separate [scalar witness record](content-model-s06-re-01-scalar-witnesses.md) states the new bounded endpoint problem and the remaining full F2 materialization gates.
-
-Failure ownership is also context-qualified.
-An intrinsically invalid original datatype remains an actual-schema failure through scalar, AU and aggregate consumers.
-A bad candidate lexical value assessed against a valid original datatype remains a candidate failure; the lookup context alone does not establish original invalidity.
-
-The context privately owns sanitized immutable references and dependency records; ordinary caller objects are inspected and copied before reuse.
-Private completed scalar property plans may share their frozen publication, while each use repeats active selection, legality prerequisites and dependency collection.
-These copies and positively charged reuse paths preserve the shared request accounting and never transfer validity between contexts.
-
-### Complete #268 rule inventory
-
-This inventory covers every local type, declaration and scalar obligation of [#268](https://github.com/TechSpokes/typescript-wsdl-client/issues/268). It uses the October 28, 2004 second editions of [Structures](https://www.w3.org/TR/2004/REC-xmlschema-1-20041028/) and [Datatypes](https://www.w3.org/TR/2004/REC-xmlschema-2-20041028/datatypes.html). Clause identifiers below refer to those dated texts, not a later XSD edition or an engine's behavior.
-
-The accepted interpretations are AU C1, PW R-240, DT A, GSC1, P1 expression construction and F2 scalar restriction. [ADR-003](decisions/003-content-model-contracts.md#xsd-10-profile-and-enforcement-s02-d04) defines the existing application profile; schema-only operands retain their applicable local rules even where payload capability is excluded. The source owners in this section are all under `src/compiler/`; controls are under `test/unit/` unless another location is given.
-
-#### Preconditions and finite execution arguments
-
-`C` means an authenticated immutable semantic context, original identities and caller-owned shared budget. Availability or a successful local relation never establishes aggregate component validity. `S` additionally means retained source syntax, namespace mappings and source-local declarations; abstract helpers instead supply complete explicit properties. `V` means a formed scalar descriptor and operands assessed in their original type, lexical namespace, whitespace and owner context. `F` means complete effective facet properties, including fixed flags and annotations, rather than a partial inventory.
-
-`T1` traverses a finite component/pair graph with an explicit stack and active-cycle detection; a repeated active pair is unresolved or invalid, never successful validity. `T2` scans finite declaration, affiliation or namespace tables; effective substitution uses a visited finite universe and retains completeness obligations for structural subsets. `T3` parses finite operand text and evaluates finite pattern relations over its code points; decimal/calendar arithmetic charges digit-dependent work before construction. `T4` traverses the finite scalar dependency DAG and exact inventories, preserving inherited context-qualified obligations.
-
-`T5` decides membership by intersecting complete typed inverse predicates with admitted spellings, and containment by checking complete selected value images. Its datatype-specific execution arguments are T6 through T9 below. Finite producer misses contribute no negative answer; a positive probe is assessed through the original datatype before it can establish membership or widening.
-
-Every argument uses inclusive 100,000-node/125,000,000-work engineering defaults, configurable independently. The [scalar capacity evidence](#scalar-completeness-capacity-calibration) records their measurement basis. Traversal, parsing, copying, arithmetic and publication consume the same sticky budget; exhaustion returns `resource-limit`. `semanticBudget.ts`, `semantic-context-independent.test.ts`, `schema-pattern-starvation-independent.test.ts` and explicit former 1,000,000/25,000,000 controls retain the accounting contract.
-
-`T6` constructs finite Thompson machines and explores their finite subset product over XML code-point intervals. It detects productive cycles before enumerating a finite language and proves absence only after complete reachability. Infinite productive languages are never treated as exhausted finite witnesses.
-
-`T7` compares exact IEEE rounding-cell endpoints and reduces mantissa length and exponent constraints to finite weighted-path and ultimately periodic count products. A selected-image comparison first proves lexical embedding where possible; otherwise it removes one complete rounding cell after each exact failed value test. Each format has finitely many typed values, including zero and nonfinite values, so that loop terminates independently of the engineering budget; budget exhaustion remains a resource result.
-
-`T8` constructs exact decimal and temporal languages and images using finite field choices, the Gregorian 400-year residue partition, finitely many timezone offsets and bounded carry transducers. Duration order uses the four dated anchors and exact month/second arithmetic. Fixed-value temporal inverse images enumerate all legal field decompositions and zones within bounds derived from that value, including unbounded zero-padding as regular languages.
-
-`T9` constructs finite symbolic token automata whose states retain whole-word DFA position, earlier-list failure masks and capped enumeration position. Symbolic subset products decide sequence differences using complete atomic label predicates; fixed-value list membership uses a finite position/automaton/mask product. Every label decision terminates by T6 through T8, and no prefix or finite sample substitutes for complete list or union selection.
-
-#### Type relations and substitution
-
-| Obligation and primary clause | Accepted interpretation and source function | Preconditions / execution | Independent controls |
-| --- | --- | --- | --- |
-| Type identity; `cos-ct-derived-ok`, `cos-st-derived-ok` 1 | Immutable named/anonymous/builtin identity; `schemaTypeRelations.ts:relationStep` | C / T1 | `schema-type-relations.test.ts`: identity versus copies |
-| Complex immediate and recursive bases; `cos-ct-derived-ok` 2 | Explicit excluded extension/restriction sets; `relationStep`, `typeDerivation` | C; component validity retained / T1 | Complex ancestry and excluded-method controls |
-| Simple immediate/recursive bases; `cos-st-derived-ok` 2.1–2.3 | Restriction exclusion and immediate base final; `relationStep` | C; simple role checked / T1 | Atomic/list/union ancestry controls |
-| Simple ur-type and union alternatives; `cos-st-derived-ok` 2.4–2.5 | List/union ur ancestry and ordered member alternatives; `relationStep` | C; member availability / T1 | Union alternatives and restricted list controls |
-| All builtin ancestry; Datatypes 3, Structures ur-type definitions | Complete `builtinParents`, builtin ur and list entries; `info`, `immediateTypeBase` | C / T1 | Builtin hierarchy controls |
-| Immediate final; `cos-st-restricts`, `cos-st-derived-ok`, `cos-ct-extends` | Constructor item/member final differs from base restriction final; `immediateFinal` | C; local properties present / T1 | List/union constructor and complex final controls |
-| Final/block token grammar; element/type XML representations | `#all`, category-specific defaults and methods; `declarationMethods` | C,S or explicit raw flags / T2 | Invalid-token and schema-default controls |
-| Immediate affiliation; `e-props-correct` 4 | Head final affects affiliation; head block does not erase it; `checkSubstitution` affiliation mode | C; global declaration/type roles / T1,T2 | Final versus block controls |
-| Transitive affiliation and substitution; `cos-equiv-derived-ok-rec` | Preserve incoming properties; collect type-path methods/blocks; `affiliation`, `substitutionTypes` | C; local type relations / T1,T2 | Blocked runtime and retained affiliation controls |
-| Effective substitution group; `cos-equiv-class` | Abstract filtering and complete incoming census; `members`, `effectiveSubstitutionMembers` | C; complete universe or explicit obligation / T2 | Effective members and recursive affiliations |
-| Relation qualification and reuse; #266/#267 contract | Context/options-qualified immutable direct facts and dependency replay; `directQuery`, `run` | C / T1,T2 | Context isolation, warm-cache and exhaustion controls |
-
-#### Declaration properties and source-local rules
-
-| Obligation and primary clause | Accepted interpretation and source function | Preconditions / execution | Independent controls |
-| --- | --- | --- | --- |
-| Element/attribute name and namespace; `e-props-correct`, `a-props-correct` | XML NCName, forbidden attribute namespaces/names; `schemaDeclarationPrimitives.ts:checkDeclaration` | C / T2,T3 | Declaration name/namespace controls in `schema-type-relations.test.ts` |
-| Element abstract/nillable and value-property kinds; `e-props-correct` | Boolean properties and default/fixed discriminant; `checkDeclaration` | C / T2 | Declaration-property controls |
-| Source name/ref, type/anonymous, default/fixed and local/global rules; `src-element`, `src-attribute` | Check local mutually exclusive properties without inventing raw XML; `checkDeclaration` | C,S / T2 | Source declaration contrasts; `schema-syntax-assessment.test.ts` |
-| Declared-type resolution and roles; `e-props-correct` 2, `a-props-correct` 2 | Attribute type simple; element simple or complex; `declaredType`, `resolveSemanticType`, `checkDeclaration` | C / T2 | Wrong-role, anonymous and generated-slot controls |
-| Declared scalar legality; `st-props-correct` | Describe authentic scalar; carry local scalar obligations; `checkDeclaration`, `declaredPlan` | C,V / T4 | `schema-scalar-independent.test.ts`, generated-binding controls |
-| Default/fixed membership; `e-props-correct` 2, `a-props-correct` 2 | Assess original typed value; mixed content emptiability stays #271; `checkDeclaration`, `assessConstraint` | C,V; mixed-content premise retained / T3,T4 | Declaration values and `component-construction-independent.test.ts` |
-| ID-derived value constraints; `e-props-correct`, `a-props-correct` | Schema-only prohibition still applies; `assessConstraint` | C,V / T4 | Schema-only scalar/declaration controls |
-| QName/source lexical legality; XML representation rules | Source grammar/values owns QName resolution, original lexical data retained; `schemaSourceValues.ts`, `checkActualSchemaSyntax` | C,S / T2,T3 | `schema-source-assessment-independent.test.ts`, XML inventory controls |
-| Immediate substitution affiliation; `e-props-correct` 4 | Local declaration primitive consumes qualified type relation; `checkDeclaration` | C; global head available / T1,T2 | Affiliation validity versus substitutability |
-
-Particle-position EDC, UPA, complete type formation and mixed-content emptiability remain #270/#271. Declaration receipts name those obligations; type relations have no reverse dependency on particle or aggregate checking or RE01 search.
-
-#### Scalar descriptors, lexical spaces and values
-
-| Obligation and primary clause | Accepted interpretation and source function | Preconditions / execution | Independent controls |
-| --- | --- | --- | --- |
-| Atomic/list/union construction; `st-props-correct`, `cos-st-restricts` | Finite DAG, roles, final and corresponding item/member ancestry; `scalarSchemaAssessment.ts:ensure`, `dependencies` | C,S or complete raw properties / T1,T4 | `schema-scalar-independent.test.ts`, policy and properties controls |
-| Source simple-content and generated identities; `src-ct`, simple-content mapping | GSC1 own identity/final/namespace, mixed-base obligation; `normalizeReference`, `originalGeneratedShape`, `ensure` | C,S; exact original slot availability / T4 | `schema-selected-gsc-independent.test.ts`, generated-binding controls |
-| Primitive/derived builtin facets; Datatypes 3, 4 | Empty primitive component sets versus mandatory normalization; `schemaBuiltinFacets.ts`, `builtin` | C / T4 | `schema-scalar-properties-independent.test.ts` |
-| XML characters and string/name/language families; Datatypes 3.2.1, 3.3.1–3.3.7 | XML code points, type grammar and exact Unicode length; `schemaDatatypeValues.ts:atomic` | V / T3 | XML character, whitespace and name controls |
-| Whitespace; `rf-whiteSpace`, `cvc-whiteSpace-valid` | Preserve/replace/collapse; union members normalize independently; `whitespaceFor`, `operandStep` | V / T3,T4 | Original token/string and ordered-union contrasts |
-| Decimal and integer families; Datatypes 3.2.3, 3.3.13–3.3.25 | Exact coefficient/scale and subtype bounds; `atomic`, `decimal`, `compareDecimal` | V / T3 | Adjacent large integers and decimal aliases |
-| Boolean; Datatypes 3.2.2 | Four lexical forms, two values; `atomic` | V / T3 | Alias versus primitive-disjoint controls |
-| Float/double; Datatypes 3.2.4–3.2.5 | Exact rational IEEE rounding and nonfinite values; `ieee`, `atomic`, `compare` | V / T3 | Floating-point rounding and bound controls |
-| Date/time and reduced families; Datatypes 3.2.7–3.2.14 | Selected DT A signed-year leap/skip-zero/second-60 and exact zones; `schemaExactCalendar.ts:parseCalendar`, `parseReduced` | V / T3 | Shared DT suite; `schema-scalar-independent.test.ts` |
-| Duration; Datatypes 3.2.6 and Appendix E | Exact months/seconds equality, four-anchor partial ordering; `parseDuration`, `compareDuration` | V / T3 | DT duration aliases and indeterminacy controls |
-| Calendar equality/order; Datatypes 3.2 temporal clauses | Zone presence retained; indeterminacy explicit; `sameCalendar`, `compareCalendar`, value `compare` | V / T3 | BCE, exact fractions, zone/boundary controls |
-| Hex/base64 binary; Datatypes 3.2.15–3.2.16 | Canonical decoded octets, distinct primitive families, byte length; `atomic` | V / T3 | Binary aliases, padding and byte-length controls |
-| anyURI; Datatypes 3.2.17 | Escaped URI reference, no dereference; `schemaUri`, `atomic` | V / T3 | URI scalar controls |
-| QName; Datatypes 3.2.18 | Resolve original lexical namespace, expanded-name equality; `atomic`, `sameValue` | V; original namespaces / T2,T3 | QName alias/rebinding controls |
-| NOTATION; Datatypes 3.2.19 and `enumeration-required-notation` | Schema-only enumeration and active source notation table; `atomic`, `declaredPlan` | C,V; notation source universe / T2,T3 | `semantic-notation-independent.test.ts`, NOTATION controls |
-| ID/IDREF/ENTITY families; Datatypes 3.3.8–3.3.12 | Schema operands retain local grammar; ENTITY needs unavailable DTD context; `atomic`, `builtinListItem` | V; DTD prerequisite explicit / T3 | Schema-only identity and ENTITY qualification controls |
-| Lists; Datatypes 2.5.1.2 | Collapse, atomic or union-of-atomic items, item count; `operandStep`, `listTokens`, `listOperand` | V / T3,T4 | List aliases, exact values and item-count controls |
-| Ordered unions; Datatypes 2.5.1.3, AU original-context contract | First successful declared member, exact selected identities; `operandStep`, `schemaValueSequences.ts:chooseUnionMember` | V / T3,T4 | Boolean/decimal order and whitespace contrasts |
-| Cross-context equality/order and immutable publication; #267, DT/AU contract | No coercion or validity transfer; `schemaScalarKernel.ts:equivalent`, `compare`, `compareSchemaFacetProperties` | C,V; same authentic budget / T3,T4 | Cross-context, origin and forged-context controls |
-
-#### Facet components, construction and containment
-
-| Obligation and primary clause | Accepted interpretation and source function | Preconditions / execution | Independent controls |
-| --- | --- | --- | --- |
-| Facet applicability; Datatypes 4.3 | Primitive/variety-specific facet kinds; `allowFacet` | C,V,F / T4 | Scalar properties and illegal-facet controls |
-| Exact component inventory; `st-restrict-facets`, c-fr | Retain every value/fixed/annotation tuple; legacy normal form separate; `scalarTuples`, `readFacetProperties`, `rawFacetCorrespondence` | C,V,F / T4 | `schema-exact-facet-components-independent.test.ts`, inheritance review |
-| Pattern construction/correspondence; `st-restrict-facets`, `rf-pattern` | P1 singleton or supplied-order `|`; equality distinct from language inclusion; `patternExpression`, `samePatternProperty` | C,V,F / T4 | `schema-pattern-property-independent.test.ts`, policy controls |
-| Pattern membership; `cvc-pattern-valid`, Appendix F | Evidenced whole-string XSD subset, disjunction within layer and conjunction across layers; `schemaPattern.ts:compile`, `acceptsOperand` | V; admitted regex constructs / T3,T4 | Pattern witness/starvation controls |
-| Enumeration; `rf-enumeration`, `cvc-enumeration-valid` | Typed base membership and exact value equality; `facetOperand`, `operandStep`, `excludedByEnumeration` | C,V,F / T3,T4,T5 | Membership, witness and inherited-layer controls |
-| Universal fixed duties; each facet valid-restriction clause | Every fixed base value applies, even with another compatible match; `exactScalarRestriction`, `layer` | C,V,F / T4 | Fixed empty restriction and repeated-fixed controls |
-| F2 same-kind matching; each facet valid-restriction clause | Existential compatible match plus whole-value containment; `compatibleScalar`, `exactScalarRestriction` | C,V,F / T4,T5 | Retained minima 3/5 and weakening controls |
-| Length/minimum/maximum; `rf-length`, `rf-minLength`, `rf-maxLength` | Exact type units, ancestor coexistence duty; QName/NOTATION value vacuity; `checkLengthProperties`, `valueFacetFailure` | C,V,F / T4 | Length ancestry, Unicode/binary/list contrasts |
-| Precision; `rf-totalDigits`, `rf-fractionDigits` | Exact value digits/scale and cross-kind consistency; `layer`, `valueFacetFailure`, `exactScalarRestriction` | C,V,F / T3,T4 | Exact decimal precision and repeated scalar controls |
-| Numeric/calendar bounds; `rf-minInclusive`, `rf-minExclusive`, `rf-maxInclusive`, `rf-maxExclusive` | Clause-specific prohibited orders, endpoint membership, exclusive repetition exception; `layer`, `exactScalarRestriction` | C,V,F / T3,T4,T5 | Indeterminate order, excluded endpoint and inherited-kind controls |
-| Local S reconstruction and annotations; `st-restrict-facets` | Replace complete scalar kinds; retain unknown annotation prerequisite; `annotationRelation`, `rawFacetCorrespondence`, `exactScalarRestriction` | C,V,F; annotation data or authentic identity / T4 | Whole-kind reconstruction and annotation review controls |
-| Complete value containment; Datatypes restriction definition and F2 | Complete derived value space is a subset, including pattern/facet interaction; `exactScalarRestriction`, `selectedUnionContainment` | C,V,F / T5–T9 | Retained minima, interacting bounds, arithmetic images and selected-union contrasts |
-| Complete typed membership; facet-value clauses | Some legal lexical alias selects the original typed value; `membershipWitness`, `regularMembership` | C,V,F; original namespaces / T5–T9 | Infinite aliases, typed lists, original QName bindings and exact exclusions |
-| Value constraint formation; declaration clauses and AU C1 | Typed default/fixed membership without payload augmentation; `assessConstraint`, `checkValue` | C,V / T3,T4 | Declaration/AU scalar origin and equality controls |
-
-#### Gap and qualification census at the fetched baseline
-
-The starting main is `a103ca96a3609bd0b525f2331daa7173879ad187`, tree `5ddb85e204e29e4561ebdd1a937434b2c8d1a6fa`. `exactScalarRestriction` lacks general pattern/scalar implication: exact base minima `{3,5}`, retained `[a-z]{5,}`, and derived minimum `{3}` is contained but qualified. Retained `a+` supplies the widening contrast `aaa`; a finite literal-only certificate does not cover the infinite positive.
-
-`membershipWitness`, `schemaScalarWitnesses`, `literalLanguage` and `positiveWords` are finite positive producers. They do not exhaust decimal zero aliases, calendar/timezone/duration aliases, contextual QName forms, binary forms, list products or ordered-union selection. General inverse-value language and datatype-aware containment are implementation obligations on complete admitted input, not permitted missing-input qualifications.
-
-Permitted input qualifications are absent authentic contexts/properties, absent facet/annotation metadata, an incomplete active source notation universe, legacy namespace normalization and the documented DTD prerequisite for ENTITY. Source multioperand annotation aggregation remains the existing explicit construction prerequisite; known annotation data must be compared. Mixed-base emptiability, complete declared-type validity, complete substitution universe and aggregate root obligations remain with their named consumers.
-
-Existing pattern exclusions are Unicode block escapes, application XML-name escapes, `w/W`, character-class subtraction and unsupported nested class escapes. Fixed S4S and intrinsic builtin consumers have their separately evidenced name support. Application document identity, DTDs, identity constraints, redefine, XSD 1.1 and dynamic payload polymorphism keep their ADR exclusions; those exclusions do not remove schema-only type/declaration/scalar duties. General RE01 witness adequacy and search termination stay #234, while production activation stays #179/#153.
-
-### Selected scalar comparison policies
-
-The [authorized continuation](https://github.com/TechSpokes/typescript-wsdl-client/issues/232#issuecomment-6141810918) selects P1 for pattern properties and directs evidence-qualified F2 scalar restriction. The comparison asks which component property was constructed; accepted-language equality and legal derivation are separate judgments.
-
-P1 preserves a singleton expression and joins complete expressions in supplied order with `|`. It adds no grouping, sorting or deduplication. Thus `['a', 'b']` constructs `a|b`, while `['b', 'a']` constructs a different property. A later language-equivalence operation can have its own authority without changing component correspondence. Pattern membership continues to use the compiled XSD expression semantics.
-
-F2 requires each effective scalar facet to have a compatible same-kind base facet when that kind exists. It additionally checks every applicable fixed base value, inherited cross-kind restrictions, complete local S construction, typed bound membership and whole-value containment. Retaining exact `minLength` components 3 and 5 succeeds; retaining only 5 succeeds; retaining only 3 fails when the result admits a value of length 3. A retained enumeration containing only `abcdef` can prove containment even with minimum 3.
-
-The dated primary rules are [Structures st-restrict-facets](https://www.w3.org/TR/2004/REC-xmlschema-1-20041028/#st-restrict-facets) and [Datatypes constraining facets](https://www.w3.org/TR/2004/REC-xmlschema-2-20041028/datatypes.html#rf-facets). The bound restriction clauses prohibit specified mathematical relations; an indeterminate calendar comparison alone is not such a relation. Value assessment separately requires the value to satisfy each bound, and a sufficient implication proof needs a determinate order.
-
-The checker proves scalar containment from retained scalar implications, complete finite enumeration and the exact language/image authorities below. An assessed result value outside the base proves widening. The retained `[a-z]{5,}` pattern with weaker scalar minimum has a complete regular certificate; IEEE, temporal and selected list/union images have their own complete datatype-aware procedures.
-
-Exact component inventories preserve repeated tuples and fixed metadata.
-Source descendants apply their known local-S construction: scalar kinds are replaced, while enum/P1 occurrences with distinct whole values remain and equal-valued base occurrences are replaced by the local component.
-The current component inventory and complete historical enforcement layers are different views.
-The legacy `facetLayers` view denotes its proved normal form; comparing it with an exact inventory uses that normal form, preserving equality transitivity.
-An unproved normal form or missing annotation evidence retains its qualification.
-
-The 43 independently specified policy controls and 14 additional reviewer controls cover these distinctions. Reviewer regressions include dropped inherited calendar constraints, fixed duties on an empty source restriction, hidden old constraints after replacement, whole-kind local reconstruction, erased inherited bound kinds, repeated exclusive endpoints and mixed-representation equality. Retained bound kinds compare annotation properties as well as value and fixed metadata; unknown metadata keeps a qualification when no valid local construction is proved. The lower inclusive/exclusive prohibition applies to the complete datatype, while the upper-pair prohibition applies to the local derivation step, following their distinct dated clauses.
-
-### Immediate-base raw component selectors
-
-`RawFacetComponentEntry` adds `{baseComponent: number}` beside the existing literal `RawScalarFacetComponent`.
-`validateRawSimpleTypeShape` requires an own data ordinal that is a nonnegative safe integer and rejects annotation overrides.
-Only that ordinal is interpreted; extra caller fields supply no component authority or metadata.
-`semanticContext` admits selectors only when a fresh canonical SIMPLE definition and its raw row are first installed together.
-The guard includes inactive stored parent changes and original rows; inherited selector rows stay immutable.
-The usual dependency traversal permits a co-born base to appear later in the supplied array.
-
-`scalarSchemaAssessment` owns a lazy, per-kernel replay table for completed raw-exact rows.
-Each entry pairs an actual component with its immutable public facet recipe; neither a `Description` nor caller identity hints serve as imported evidence.
-For a selected entry the owner forms the recipe unbranded in actual H, checks its typed value/fixed/P1 properties, then attaches the issued component identity.
-The row's own TYPE identity, BASE edge and dependency obligations remain separate.
-Tables are published after ordinary dependency/qualification assembly and are charged before allocation and replay.
-The RAW path stays lazy; eligible SOURCE property readers also request their immutable metadata projection so authority never depends on a prior selector lookup.
-Existing preparation/locality and explicit budget controls remain required.
-
-The supported positive sector requires whole actual enum/P1 KEEP, distinct ordinary new P1 properties, and whole-kind bound KEEP or one ordinary NEW bound.
-Additional NEW enumeration components are permitted after ordinary current-base membership and complete c-fr/F2 checks; issued KEEP identities never authenticate a new component's annotation.
-Coverage uses the completed base's actual main-component associations, not merely its raw array; uncovered intrinsic or source properties retain qualification.
-Exact LIST ITEM and ordered UNION MEMBER identities are sector conditions, not new schema-validity rules.
-Different genuinely derived constituents qualify only after ordinary c-fr/F2 and enum formation have run; genuine nonderivation and weakening still fail those rules.
-Other available but unsupported combinations likewise preserve known formation failures before returning qualification.
-
-An unavailable raw table is an unresolved construction prerequisite, not permission to invent an input inventory or prove an omitted component.
-Known literal/parser/type checks still run before that qualification.
-An available ordinal outside its table fails `raw-base-facet-component`; a changed replayed property fails `raw-base-facet-component-value`.
-Malformed or late-installed selectors fail context representation admission.
-Resource exhaustion keeps its separate sticky outcome.
-
-See the [scoped contract](content-model-s06-re-01.md#authenticated-immediate-base-component-retention), [complete forest controls](../test/research/re01/immediate-base-component-retention.test.ts) and [independent scalar controls](../test/unit/schema-exact-facet-components-independent.test.ts).
-Qualified source-only atomic DECIMAL/STRING issuance is described below; other source/generated inventories, general scalar recipes and general RE01 coverage/negative/termination remain open.
-
-### Current source repeatable-component ownership
-
-The scalar owner keeps source-constructor provenance private and computes the current enum/P1 view from genuinely formed current-H data.
-It uses whole typed enum-set equality and selected P1 property equality; annotation comparison does not decide known-source replacement.
-Distinct overlapping enum sets and P1 a/(a) keep their separate components.
-Known source grouping is separate from an unknown annotation payload; an unavailable grouping keeps its qualification and acquires no authority through a caller field.
-
-Publication and `rawFacetCorrespondence` coverage/enum introduction use the same current base/result components.
-They cannot find a discarded old component in enforcement history and thereby bypass an actual new member's current-base admission.
-Parser/membership/containment, source formation and original-recipe replay keep every required historical layer and dependency.
-This reader distinction preserves authentic original binding even when a historical enum member is outside the final datatype's image.
-
-Source steps start from current base plus actual S; complete RAW rows start from their own supplied inventory.
-Authentic recipe provenance describes its internal source-constructor sequence, rather than adding the current alias base's missing components.
-Replay forms actual typed operands in H before recomputing dominance; no cached original visibility result substitutes for current-H equality.
-The private view is charged and kernel-local, and does not replace `Description.exactComponentSet`, ordinary F2 or full validity.
-
-See the [contract and discriminator](content-model-s06-re-01.md#current-source-components-and-historical-enforcement) and [complete source controls](../test/research/re01/source-facet-current-components.test.ts).
-Qualified SOURCE ordinal issuance now reuses this current view; full RAW same-value metadata reconstruction and general RE01 coverage/adequacy/bounds/negatives/termination remain separate gates.
-
-### SOURCE replay tables and metadata authority
-
-The scalar owner qualifies the complete SOURCE restriction closure after ordinary formation, without re-entering ensure from its table builder.
-It pairs actual scalar tuples and current repeatable components with their authenticated defining source facets.
-Scalar-kind lexical order followed by current survivor order is deterministic in the admitted source-only DECIMAL/STRING sector.
-Intrinsic/derived-builtin prefixes, contextual constructors, RAW ancestors, explicit typed operands and unavailable grouping/replay backing prevent the whole table.
-
-A private original-context/source-record/kind incidence cache supplies annotationIdentity on completed immutable projections.
-It never changes the existing value/P1 identity fields or generic current-component cache.
-Exact value/fixed/P1 comparison runs before selected clones receive this metadata authority.
-Annotation correspondence and authenticated whole-KEEP association may use it; parser selection, replacement dominance and scalar equality may not.
-The base projection is available independently of a selector lookup, preserving read-order independence.
-
-SOURCE entries bind actual formed H values and lexical/source context through the existing private per-facet binding maps.
-They retain genuine current dependencies and outstanding obligations, with no Gamma validity transfer.
-Installed selected components remain RAW inventory entries without source-step provenance.
-Unavailable tables preserve known independent literal/type failures; available empty or out-of-range tables retain actionable ordinal failures.
-All table, cache, token, copy, comparison and publication work is charged before allocation or processing.
-
-See the [qualified SOURCE contract](content-model-s06-re-01.md#authenticated-current-source-component-retention),
-[new source controls](../test/research/re01/source-component-retention.test.ts) and
-[existing RAW controls](../test/unit/schema-exact-facet-components-independent.test.ts).
-No general source issuer, builtin/generated materializer or negative-search bound follows from this extension.
-
-### Complete scalar membership and containment authorities
-
-`schemaRegularLanguage.ts` owns authenticated private language programs, Boolean operations, complete reachability, quotients, concatenation, reversal, whitespace preimages/images and finite-language classification. `schemaPattern.ts:language` reads a private copy of compiled atoms and root, so mutating a published plan cannot manufacture a certificate. Character partitions use the existing admitted Unicode authority, include only XML code points, and preserve intrinsic XML-name ranges and their colon exclusion.
-
-Private Boolean reductions, productive-state pruning, deterministic Moore minimization and completed witness caches reduce repeated products without changing P1 component properties. Cached programs and published automata do not share mutable arrays; every reuse charges work before copying or reading, and incomplete exploration is never cached as an empty language.
-
-`regularLengthContainment` intersects every retained pattern layer with normalized spellings and supplied value-length constraints. String lengths use code points, hex/base64 lengths use decoded octets, list lengths count tokens, and QName/NOTATION length predicates are vacuous under the dated value clauses. Sufficient overapproximations may prove a positive; a failed sufficient proof proceeds to the complete admitted-domain authority rather than creating an algorithm qualification.
-
-After scalar formation prechecks and ordinary scalar coverage, an empty intersection of the result's main pattern-component languages proves whole-value containment directly.
-Each component contributes its alternative union, and every main component tests the same selected normalized result, including an ordered union.
-The owner returns its existing qualification flag; it neither combines separately normalized conjuncts nor removes the whitespace guard from potentially nonempty containment proofs.
-Outer c-fr, enum introduction, constituent, source, NOTATION and complete current-context validity still run.
-The [complete-context controls](../test/research/re01/empty-main-pattern-certificate.test.ts) distinguish the previously unresolved whitespace case, an existing positive, fixed violations and missing required-child evidence.
-
-`schemaDecimalLanguage.ts:relation` and `precision`, consumed by `decimalContainment`, supply exact regular preimages for signed decimal bounds and total/fraction digits. Decimal zero, signs, leading zeros, omitted integer digits and trailing fractional zeros retain their value meaning. Every selected counterexample is reassessed through the shared exact decimal owner.
-
-`inverseExpression`, `inversePredicate`, `fixedAtomicPredicate` and `regularMembership` cover complete atomic inverse images. Whitespace normalization belongs to the selected primitive; QName/NOTATION aliases retain the original namespace context, and URI grammar is shared with the actual URI parser. Hex case, base64 padding and whitespace, boolean spellings and decimal leading/trailing zero aliases map to their actual typed values.
-
-`schemaIeeeLanguage.ts` derives exact ties-to-even rounding cells, subnormal and overflow endpoints, signed zero, NaN and infinity. `intervalWitness` combines regular mantissa paths with exact exponent/count relations; an empty complete weighted product proves exclusion. Decimal-only and nonnegative integer-exponent images have complete regular specializations; other exponent images retain the exact numeric guard. Cached completed rounding cells retain private endpoint copies.
-
-`schemaCalendarLanguage.ts`, `schemaDurationLanguage.ts` and `schemaTemporalAliases.ts` connect admitted spellings to DT A exact order/equality. They cover BCE and skipped year zero, leap residues, reference dates, 24-hour and leap-second aliases, all legal zones, unknown-zone indeterminacy, arbitrary fractional precision and duration field decomposition. `schemaCalendarImages.ts`, `schemaIntegerImages.ts` and `schemaValueImages.ts` produce canonical temporal/decimal/binary images through finite field and carry products, preserving zoned versus unzoned values.
-
-`schemaLexicalPredicates.ts` and `schemaNumericImageConditions.ts` decide Boolean combinations of regular admission and exact IEEE input guards. `atomicImageLabels` retains first-success item selection, primitive value identity and global constraints; `schemaSymbolicValueSets.ts` decides label intersections/differences using canonical images or complete IEEE cell elimination. A failed exact value test removes its full inverse cell, including both zero signs and nonfinite literals, never just the attempted spelling.
-
-`schemaTokenImages.ts` and `schemaSymbolicSequences.ts` retain whole-word pattern state while factoring typed tokens and earlier-list failures. Lists keep ordered typed items; preserving string and URI images keep maximal non-whitespace fragments and each intervening whitespace code point. This encoding is injective for the normalized string, and numeric conditions are applied before erasing a selected predecessor's raw whitespace or token separators.
-
-`selectedUnionContainment` checks the union of first-selected typed images, not the unordered union of member languages. Narrowing an earlier member can introduce a later member's different typed value even with identical spelling; mixed float/double, list/string/URI and binary/list contrasts exercise that distinction. Lexical-embedding and universal-padding certificates prove entire images; optional finite probes provide positive evidence only, and complete image products decide the remaining cases.
-
-The [nonauthor controls](../test/unit/schema-scalar-completeness-review.test.ts) specify positive/widening contrasts, arbitrary-length aliases, original contexts, fixed duties, temporal/numeric exactness, ordered selection, private mutation, allocation accounting and explicit old limits. Historical finite-producer controls now distinguish a complete inverse-language exclusion from failure of an old producer. Profile exclusions, fixed duties, annotation construction, original contexts and payload activation remain as accepted.
-
-### Remaining legitimate qualifications and consumer obligations
-
-The local scalar procedures decide complete admitted inputs, subject to configurable resource exhaustion. The following qualifications require missing input, existing external prerequisites or separately owned aggregate obligations; they do not represent renamed scalar algorithm gaps.
-
-| Case | Current result and owning continuation |
-| --- | --- |
-| Unsettled independent source facet aggregation | Independently constructed multioperand annotation/component groups retain the existing external construction prerequisite |
-| Incomplete original facet or annotation correspondence | Exact cross-context properties and authentic original bindings are compared; missing metadata or unsettled independently constructed groups retain their precise prerequisite |
-| Missing authentic context or complete effective properties | Retain the precise context, facet inventory or original operand prerequisite |
-| ENTITY source environment | Retain the documented DTD prerequisite; ID/IDREF application identity constraints retain ADR exclusions |
-| Incomplete notation source availability or active membership | Context-qualified lookup retains its named prerequisite; complete source-bound original/fresh tables support concrete property checking |
-| Legacy namespace normalization | The internal semantic loader supplies normalized XSD values; legacy-loaded spellings that disagree retain the source receipt's qualification |
-
-Application identity constraints remain outside the approved profile; checking the fixed assessment schema's own keys does not add application key/keyref/unique support.
-Complete declared-type validity, substitution-universe completeness, mixed-base emptiability and aggregate roots remain explicit duties of their named consumers. Local type-relation success does not certify the schema or remove these duties.
-
-These local duties are separate from #234's universal RE01 witness/evidence adequacy, complete negative answers and search termination obligations. Production activation remains #179/#153; the complete local #268 procedures do not enable payload augmentation or publish a release.
-The [inherited-root continuation](content-model-s06-re-01.md#inherited-fresh-roots-and-the-october-13-proof-continuation) repairs a nested finite-certificate false positive: every candidate rechecks all fresh additions relative to immutable actual Gamma, including additions inherited from its parent.
-The permanent affiliation controls distinguish complete availability from complete legality roots and retain exact context-local assessment.
-The same entry point accepts a discriminated primitive root and issues a separately authenticated `checked-primitive` receipt after checking every actual obligation, including supplied evidence for unrelated nonprimitive types.
-This represents the existing no-change rule without inventing a restriction edge, which can itself invalidate a consumed affiliation; existing construction-root callers retain their input and receipt contracts.
-
-### Historical research seam
-
-The [derivability continuation](content-model-s06-re-01.md#derivability-and-affiliation-continuation) adds a model-tagged research comparison boundary and an explicit reverse-affiliation dependency census.
-The maintainer addendum supersedes universal hypothetical whole-schema preservation while retaining complete actual-schema assessment.
-The historical resolver remains a supplied-candidate seam; its conditional owner ledger is not a complete semantic authority for the revised judgment.
-
-The [reviewed resolver specification](content-model-s06-re-01-resolver.md) defines R1–R7 and behavioral cases B01–B23.
-Its selected domain is one supplied abstract two-step construction over immutable actual component records.
-The research modules do not construct or search the complete RE01 witness family.
-
-The [literal continuation](content-model-s06-re-01.md#literal-preparation-and-fresh-attribute-helper-continuation) adds `prepareLiteralContexts` with whole-Γ and non-derivation preservation, plus a scoped fresh-AU synthesis theorem/prototype.
-Its structural success does not supply literal semantic authorities; the D1 validation ledger reports that missing qualification explicitly.
-
-The historical resolver material under `test/research/re01/` consumes independently prepared records.
-It imports no unaccepted #231 checkout; shared source owners introduced by recovery are identified above.
-The [RE01 proof record](content-model-s06-re-01.md) and [joint research handoff](content-model-s06-research-handoff.md) retain their acceptance gates.
-
-| Leaf | Module | Independently prepared controls |
-|---|---|---|
-| R1 | `resolver-context.ts`, `resolver-types.ts` | Identity, ownership, outgoing closure, legal recursion, forbidden cycles, source records and deterministic diagnostics |
-| R2 | `resolver-resolution.ts`, `resolver-comparison.ts` | Caller-context slots, authenticated handles and anchored effective-property correspondence |
-| R3 | `resolver-relations.ts` | Dated complex/simple derivation, final exclusions, affiliation and retained substitution groups |
-| R4 | `resolver-validation.ts` | All-member inventory, source AU/group roles, direct owner requests and honest missing-authority results |
-| R5 | `resolver-budget.ts` | Request isolation, cache options, cumulative inclusive limits and the separate numeric work ledger |
-| R6 | `resolver-integration.test.ts` | Combined lookup, correspondence, conditional owner checks and retained/omitted substitution contrasts |
-
-The files are under `test/research/re01/`; [`resolver-types.ts`](../test/research/re01/resolver-types.ts) defines their shared record contract.
-Their adjacent literal tests cover specification cases B01–B23, including B17a's anchored particle policy.
-The existing endpoint, group, particle and substitution reference controls remain independent evidence.
-The S06 research manifest pins executable records, tests and documentation without rewriting its historical 147-artifact checkpoint.
-
-## Historical resolver guarantees and qualification
-
-Context preparation certifies structural consistency and explicit outgoing closure.
-Reference resolution certifies one declared slot's target in the caller's context.
-Endpoint comparison certifies only a supplied identity correspondence over the specified finite effective-property projection.
-
-Named predicate receipts identify their context, operands, rule and authority.
-A checked-candidate receipt requires all applicable inventory entries and endpoint correspondence to pass under qualified authorities.
-It applies only to the explicitly listed component schema and reports excluded incoming relationships.
-
-An unimplemented or unqualified semantic owner returns a named `unresolved` result.
-A rejected candidate or correspondence proves nothing about other candidates.
-Resource exhaustion returns no partial successful receipt and proves no schema-invalidity or witness-nonexistence conclusion.
-
-The relation authority is the dated rule control flow over supplied facts and a fixed seven-entry builtin ancestry table:
-`anyType`, `anySimpleType`, `string`, `decimal`, `integer`, `long` and `int`.
-Other builtin identity queries may pass by identity; a query requiring an absent builtin relation premise is unresolved.
-The table supplies ancestry only, with no scalar assessment or facet evaluator.
-Immediate derivation final checks remain separate from full type construction and simple list/union formation legality.
-
-The validation module has six fixed predicate owners: type construction, attributes, particle restriction, attribution, declarations and scalar operands.
-Each receives the caller context, shared budget and original operands, and must return a matching context/rule/ordered-operand/authority receipt.
-The implemented AU role checks preserve original identity sets, universal replacement matches, requiredness and fixed-kind obligations.
-Complete fixed-value equality, wildcard admission, UPA, EDC, particle normalization and unrepresented constraints retain their named owning checks.
-Accepted AU01 C1, PW01 R-240 and DT01 A remain the selected contracts for any such invoked predicate; a missing qualified implementation stays unresolved.
-
-Conditional owners in tests only attest captured requests.
-Their positive result tests receipt combination and never supplies a complete XSD predicate authority.
-
-## Future production adapter
-
-The production entry remains [`prepareResolvedCompilationInput`](../src/compiler/semanticCatalog.ts).
-The future #179 adapter consumes `resolved.graph`, `resolved.links` and original source context from that entry.
-It must qualify preparation authority for every actual record, including any constraint omitted from the narrow facts view.
-
-Canonical IDs remain anchors with their global role/QName or original containing declaration and local path.
-Group uses, particles, local declarations, anonymous scalar operands, original AU IDs and declaration IDs retain separate identities.
-Candidate-local additions use disjoint identities; only the designated endpoint can have a proposed replacement definition.
-
-Final/block values and defaults come from original declared and schema attributes with lexical namespace context retained.
-Original prohibited attributes and empty/repeated attribute-group references stay source records and outgoing operands.
-Unsupported syntax and unrepresented constraints become `unassessed` entries with a named rule and owner.
-
-`composed.types` may supply effective content contribution roots.
-Its QName-collapsed `attributes` cannot reconstruct original AU constraints or sharing.
-The adapter obtains original AU/declaration/group contribution IDs and their fixed/default/type operands directly from `resolved.graph`.
-
-The accepted AU01 C1 source-replacement predicate checks every original matching intermediate AU.
-Retention preserves the entire original same-QName AU set, including separate fixed obligations.
-The accepted PW01 R-240 and DT01 A predicates are invoked only for the obligations within their reviewed domains.
-
-The single accepted [`OccurrenceAnalysis`](../src/compiler/occurrenceAnalysis.ts) supplies exact total range and `schemaEmptiable` to its requesting particle owner.
-Changed particle inputs require a context-qualified view of that same contract.
-Reusing an unchanged particle summary requires an identical content/occurrence closure; changed type composition is recomputed.
-
-Type relations, normalized particle views, scalar descriptions and semantic caches must take explicit context and complete predicate options.
-An actual assessment receipt cannot silently become a proposed-context receipt through an unchanged component.
-Group attribution must retain use positions, while EDC includes the explicitly retained implicit substitution members.
-
-## Remaining acceptance owners
-
-#234 owns full-type witness soundness, complete candidate-family reduction and the negative-answer/termination theorem.
-Its proof must discharge ambient-schema preservation, source-valid construction, identity/incidence and delegated semantic assumptions.
-#232 owns joint independently reviewed research acceptance with the accepted AU01/PW01/DT01 contracts.
-
-#271 owns local component construction and every prerequisite needed by its admitted finite evidence.
-#179 owns guarded whole-assessment orchestration, operation capabilities and integration of the complete RE01 decision.
-#153 owns combined S06 acceptance, and #184 retains payload scalar enforcement.
-This research delivery does not change S07/S08 readiness, public support, catalog/profile contracts or the legacy default.
-
-## Reproduction and delivery evidence
-
-The linked implementation PR and issue handoff identify exact reviewed/head/merged commits and trees.
-They record accountable implementing agents, independent findings and resolutions, final local commands, hosted results and measured budget evidence.
-Historical observations remain separate from current reference observations and conditional supplied-predicate checks.
-
-```bash
-npm ci --cache tmp/cache/npm
-npm run reference:setup
-npm run check:toolchain
-npm run typecheck:research
-npm run typecheck:reference
-npm run docs:validate
-npm run test:research:ports
-npm run ci
-npm run test:reference:full
-npm run test:conformance
-git diff --check
-```
-
-The independently authored `resolver-work-ledger.test.ts` derives one complete finite preparation and lookup count from charged operations.
-Its exact and one-below controls are separate from measured counts and historical witness-probe boundaries.
-The two-actual-component fixture requires 44,683 preparation units and 271 lookup units, totaling 44,954.
-At 44,953 units its final six-unit output reservation fails at 44,948, returning only the shared terminal result.
-The node controls separately admit 100,000 distinct actual identities with sufficient work and reject 100,001 before index allocation.
-All request operations share cumulative counters; an exhausted request returns its reserved terminal result on subsequent calls.
-The final revision must pass substantive independent review after all semantic or resource changes.
-Normal protected merges and merged-main verification complete delivery; an unmerged parent branch alone does not satisfy it.
-
-## Preparation measurement evidence
-
-The [strict TypeScript harness](../test/research/re01/preparation-measure.ts) measures complete local source inputs through the loader, canonical graph, portable provenance, fingerprints and resolved preparation entry.
-The [corpus manifest](../test/research/re01/preparation-corpus.ts) pins original bytes and complete import/include closures; the [recorded comparison](../test/research/re01/preparation-evidence.json) contains 16 corpus/limit rows, each with one cold and two warmed samples.
-These are structural preparation measurements, not schema-validity, witness-coverage or public activation receipts.
-
-The four committed tiers are the complete 21,204-byte Weather WSDL, a relative-import WSDL/XSD closure, the chameleon include closure and the shared-recursive XSD control.
-Synthetic sizes 4, 16, 64 and 128 add independently recursive types that share one two-element group; their scale results remain labeled synthetic.
-No larger redistributable real-service snapshot with complete closure was available in the repository, so the real-service scale claim stops at Weather.
-
-The baseline is archived main `7d652fea79fddce8fd407f5c18dc2645ad203cc5`.
-Reports hash all measured source files, the harness and lockfile, record Node/platform/CPU details, and reject source changes during a run.
-The current source snapshot starts at main `9eb027cd144cdb2ac7e48e1d87990fea7b92342a`; its 112-entry source/harness/lock map is `9c4a9db5175739d9656ed4a2dbc58d63e2550a2ef9bdaa77d45929d99362b8c0`.
-The recorded comparison's SHA-256 is `437b0074f22f78e4fc62d82752ff97e51cf933b77a458a41029aae42f88e8016`; [PR276 measurements](https://github.com/TechSpokes/typescript-wsdl-client/issues/272#issuecomment-6114655396) remain historical evidence for their exact source and URI.
-The [PR274 measurement handoff](https://github.com/TechSpokes/typescript-wsdl-client/issues/272#issuecomment-6113569258) remains a separate historical delivery record.
-
-| Complete input | Portable objects before | After | Persisted boundary |
-|---|---:|---:|---|
-| Weather WSDL | 8,406 | 5,077 | Identical |
-| Relative import closure | 1,025 | 634 | Identical |
-| Chameleon closure | 372 | 205 | Identical |
-| Shared-recursive XSD | 2,876 | 1,254 | Identical |
-
-The [provenance transformer](../src/compiler/catalogProvenance.ts) preserves original object sharing, keyed by input identity and the complete parent expanded name.
-It does not intern content-equal records from different semantic contexts.
-Every before/after row preserves the same input closure/options, format-2 canonical JSON bytes, catalog digest and semantic fingerprint; the normal writer's trailing newline is outside the reported canonical JSON byte count.
-
-[Independent transformation controls](../test/research/re01/preparation-provenance.test.ts) retain chameleon namespaces, rebound QNames, distinct inherited bases, original diagnostic/source locations and context-sensitive credential checks in both encounter orders.
-The finite sharing series counts unique declarations and distinct group-use positions without unfolding legal element/type recursion.
-Object/field counts are deterministic retained-allocation proxies, not measurements of V8 allocated bytes.
-
-Stage wall time, lifetime process peak RSS, logical work, admitted nodes and serialized/provenance-table sizes remain separate quantities.
-`process.resourceUsage().maxRSS` includes the TypeScript loader and all prior stages in that child; subtracting successive high-water values would not measure stage memory.
-No forced garbage collection is used, OS file caching is uncontrolled, and cold means the first call after module loading rather than cold disk or process-start time.
-
-Existing uninstrumented stage counters are recorded as `null`.
-Weather's baseline gathering, merge, resolution and composition work is respectively 739,119, 183,415, 472 and 27 under their inherited independent stage budgets.
-The separately delivered wildcard owner increases shared-recursive composition work from 87 to 141; that charge increase is not a provenance optimization.
-
-The [candidate workload](../test/research/re01/preparation-candidates.ts) compares one reusable actual preparation with deliberately rebuilding the same new adapter under one cumulative budget per workload.
-Baseline main has no equivalent real-graph request API, so this is not a historical semantic-API comparison.
-Its three distinct deltas retain full membership, select the endpoint's closure with changed `mixed`, and restore full membership in a sibling with changed `abstract`.
-
-Weather diagnostic actual-plus-delta work is 3,419,023 with reuse versus 9,599,863 with rebuilding, at active sizes 140, 28 and 140.
-The three delta costs are 140,673, 36,346 and 151,505 work units, with original source evidence still available separately from active membership.
-[Request controls](../test/research/re01/preparation-candidates.test.ts) verify actual/sibling isolation, exact inclusive allowance and sticky exhaustion after earlier successful deltas.
-
-The separately named `source-owned-actual-preparation` workload measures the authenticated [source request factory](../src/compiler/schemaDocumentSyntax.ts) used by the finite checker.
-Its loader stage has separate limits; its owned source snapshot and context share one request budget, while canonical construction and reference resolution retain their inherited stage limits.
-The exact immutable resolved handle now binds the complete original source universe, whose notation properties are prepared through the shared property owner.
-This workload reports source-input and request-graph allocation censuses and runs no schema-validity assessment.
-Baseline main has no matching source-owned API, so these current preparation costs are reported separately from both the historical provenance comparison and the structural reuse controls.
-
-| Complete input | Source-owned diagnostic work | Source-owned default outcome |
-|---|---:|---|
-| Weather WSDL | 5,473,160 | Resource limit at 1,000,000 work, before context nodes |
-| Relative import closure | 690,163 | Prepared |
-| Chameleon closure | 211,276 | Prepared |
-| Shared-recursive XSD | 1,106,574 | Resource limit at 999,974 work, with 52 admitted nodes |
-
-Synthetic source-owned diagnostic work is 417,561, 1,311,759, 4,890,194 and 9,667,888 at sizes 4, 16, 64 and 128 respectively.
-Charged source work includes original source URI string lengths; the task paths are recorded in the issue handoff, and moving a checkout can change these counts.
-The current source-owned samples use `/workspace/recovery-232-correspondence`; earlier PR276 samples used `/workspace/recovery-232-scalar-finite`, PR275 used `/workspace/recovery-232-finite` and PR274 used `/workspace/recovery-232-followup`.
-Differences therefore include source URI lengths and source changes, without attributing them to one isolated optimization.
-
-Default and diagnostic rows remain separate; diagnostic work uses an explicitly named 50,000,000 allowance while defaults remain unchanged.
-At the post-277 baseline, Weather source/catalog preparation succeeds at defaults, but its new actual context exhausts the 1,000,000-work allowance before any candidate.
-In the separate #278 matched-path continuation from post-277 main, Weather source-owned work falls from 5,461,538 to 988,098 and shared-recursive work from 1,103,022 to 214,199. Both now prepare under the unchanged 1,000,000-work default, with identical catalog fingerprints and request-graph census. The two checkouts have equal-length paths (`/workspace/recovery-278-baseline` and `/workspace/recovery-278-proposed`) and use identical fixture bytes. This measures structural source entry only: Weather's separate actual-plus-candidate context workload still exhausts the default allowance. Warm wall-time samples improved, but OS cache and GC are uncontrolled; logical work and identical graph facts are the primary comparison.
-Default import and shared-recursive contexts finish with reuse while the rebuilding controls exhaust; synthetic16 exhausts its third delta after two successes, and synthetic128 exhausts source gathering at the legacy reported step 1,000,001.
-
-The separate cold finite candidate accounting runs three unchanged constructions with fresh source requests and no prior cache warming; [PR276 accounting](https://github.com/TechSpokes/typescript-wsdl-client/issues/271#issuecomment-6114656155) remains historical evidence.
-Its before/after baseline imports exact frozen `b443034e2aa93f92dc6ad70da7b75c5b98220249` source while using the same fixture bytes and `/workspace/recovery-232-correspondence` input URIs as the current source, so source-path length is matched for this comparison.
-The current report's SHA-256 is `9a67d7d9f3c4de465e0bd77f7b0c3fe6a90c8e3a67f7e8b3d88ab40594a1d0ae`.
-
-| Fixed construction | Baseline diagnostic work | Current diagnostic work | Current default outcome |
-|---|---:|---:|---|
-| Particle, AU and wildcard | 1,455,248 | 1,426,117 | Resource limit at 999,962 work |
-| Nested helper | 1,437,731 | 1,556,788 | Resource limit at 999,959 work |
-| Typed-attribute scalar | 953,192 | 974,151 | Checked candidate at 974,151 work |
-
-Diagnostic allowance is 100,000,000 for these finite checks; the baseline scalar also passed the unchanged 1,000,000 default allowance while its two larger constructions exhausted it.
-The current typed-attribute success preserves the full actual and candidate receipts and is an executed result, not subtraction of estimated cache savings.
-The changes include new semantic dependency checks and safe owned-data reuse; this is not a general speedup claim, and the earlier PR275 AU reuse improvement remains scoped to its own matched baseline.
-Candidate accounting is separate from structural preparation costs and does not establish universal witness coverage.
-The [independent permanent regression](../test/unit/re01-candidate-independent.test.ts) uses the same fixture bytes at a fixed offline URI so its default-budget assertion does not depend on checkout-path length.
-At that fixed URI, the matched scalar baseline uses 942,932 work and this source uses 966,375; both complete under defaults.
-
-### Reproduce and retain measurements
-
-Run these commands from the repository root with installed dependencies.
-The [baseline helper](../test/research/re01/preparation-baseline.ts) archives the pinned revision without changing any checkout and removes only its own temporary archive.
-Add `--corpus weather` or `--corpus chameleon` to both measurement commands for a smaller matching comparison.
-
-```bash
-node --import tsx test/research/re01/preparation-baseline.ts --output tmp/conformance/preparation/baseline.json
-node --import tsx test/research/re01/preparation-measure.ts --contexts --output tmp/conformance/preparation/candidate.json
-node --import tsx test/research/re01/preparation-summary.ts tmp/conformance/preparation/baseline.json tmp/conformance/preparation/candidate.json tmp/conformance/preparation/evidence.json
-```
-
-The [summary generator](../test/research/re01/preparation-summary.ts) rejects changed input/output boundaries and nondeterministic stage or request work across repetitions.
-Committed corpus identities and the reviewed evidence record are durable; raw reports and generated synthetic files under `tmp/conformance/preparation/` are disposable task outputs.
-Full CI, independent review, protected merge and merged-main verification remain delivery requirements; these bounded observations do not close #232/#234 or activate #179/#153.
-
-### Accepted #278 finite capacity qualification
-
-The original [fixed-URI finite harness](../test/research/re01/finite-capacity-measure.ts) compared actual main `2a057d22a48bfe78a7af2deb368f86a4c762533f` (tree `845417705b22259f6ebf6ce93abb481d31dfedbc`) with the #278 change. Both checkouts used the same source bytes and `https://capacity.example/` offline URIs. Each workload used a fresh request for each of three samples, with separate then-default 1,000,000-work and diagnostic 100,000,000-work profiles; both allowed 100,000 nodes. The table preserves that accepted historical experiment; logical work and nodes were identical across its three repetitions.
-
-| Complete finite workload | Main diagnostic work | Main default | #278 default work | #278 nodes | Remaining default work |
-|---|---:|---|---:|---:|---:|
-| Particle, AU and wildcard | 1,204,704 | Resource limit | 917,562 | 61 | 82,438 |
-| Nested helper | 1,438,063 | Resource limit | 976,270 | 11 | 23,730 |
-| Typed attribute | 857,115 | Checked | 593,138 | 7 | 406,862 |
-
-The #278 diagnostic and default profiles produce the same complete checked receipts and totals. Their deterministic JSON receipt-projection SHA-256 values match current main's diagnostic receipts for all three workloads. Each retains checked actual obligations and the supplied candidate; universal coverage, complete negative answer and search termination remain false.
-
-Main's diagnostic receipts and the #278 receipts retain the same particle object and string census; the nested and typed #278 receipts retain fewer objects and strings. This census traverses reachable receipt values and is a retained allocation proxy, not a V8 allocation counter.
-
-| Checked receipt | Main objects | #278 objects | Main string values | #278 string values |
-|---|---:|---:|---:|---:|
-| Particle, AU and wildcard | 1,234 | 1,234 | 1,649 | 1,649 |
-| Nested helper | 1,157 | 1,127 | 1,513 | 1,478 |
-| Typed attribute | 690 | 684 | 908 | 901 |
-
-Source preparation, actual lookups and candidate preparation retain the same logical work on main and #278. The particle stages use 46,257, 171 and 129,222 work; the nested stages use 24,846, 108, 103,325 and another 75,936 for its nested candidate; the typed stages use 24,846, 108 and 86,100. Complete finite verification changes from 1,028,747 to 741,605 for particle, 1,233,553 to 771,760 for nested and 745,982 to 482,005 for typed. The main default runs stop during verification, so their truncated verification counts are not comparable completed costs.
-
-The retained source-owned structural controls remain 988,098 work and 140 nodes for Weather, and 214,043 work and 53 nodes for shared-recursive, under both main and #278. Weather's separate three-candidate structural reuse workload consumes 905,654 work and 55 nodes. These are source and context preparation controls, not complete finite proofs. The separate generic Weather actual-plus-candidate context path still exceeds the default; this qualification does not claim that path passes. Weather is the largest available redistributable complete real-service closure.
-
-The finite harness reports loading separately with `null` shared-budget work because loading uses inherited limits. It separates source ownership, actual lookup and each candidate construction, but groups original local assessment, candidate local assessment, dependency discharge and evidence publication in the complete finite-verification stage. Earlier structural harnesses expose canonical construction and reference resolution independently. An exact per-rule allocation or verifier substage claim would require additional instrumentation. Wall time and process RSS are observed per stage; OS cache, JIT and garbage collection are uncontrolled, and process peak RSS is a lifetime high-water mark. The deterministic work, node and receipt observations support this bounded capacity result.
-
-The archived comparison used the original #278 harness in equal-length checkouts. The expanded harness below measures the current default separately from the explicit former limit; its new imports and workloads must not be substituted into that archived experiment without a new matched measurement.
-
-The optimization caches only positive direct relation and content-identity facts in their owning context. Every call re-owns references, charges cache lookup and publication, and replays the checked dependencies; finite builtin dependencies are checked once per exact context and identity. Typed attribute constraints use one scalar assessment. Candidate-specific and nested proof obligations still run for the supplied evidence. These historical controls qualify named workloads only. The later authorized continuation selects P1 and evidence-qualified F2; #234 proof and #232 full acceptance remain open.
-
-### Successor budget calibration
-
-The [authorized #232 continuation](https://github.com/TechSpokes/typescript-wsdl-client/issues/232#issuecomment-6141810918) treats request defaults as engineering choices. This historical calibration selected 25,000,000 work units and retained 100,000 nodes. The scalar-completeness calibration below supersedes that work default; callers can override either positive safe-integer limit, and both remain inclusive.
-
-The [expanded harness](../test/research/re01/finite-capacity-measure.ts) and [recorded evidence](../test/research/re01/finite-capacity-evidence.json) measure complete named workloads on Node `v24.19.0`, Linux x64, Intel Xeon Platinum 8573C. The integrated evidence was freshly measured at source revision `024460832f975b57d2f8b57a35cf56a8f175e7a1` and records exact source/dependency hashes and the production diff hash. Earlier baseline measurements remain in the budget commits and linked issue record. Each workload/profile runs in a fresh child with one cold and two warm requests; fixed offline URIs remove checkout-path differences, and request identifiers keep equal lengths.
-
-The three finite cases execute `verifyFiniteRe01Evidence` fully, including original and candidate obligations, dependency discharge and receipt publication. Weather and the shared-recursive cases execute one complete actual preparation and all three candidate deltas, including changed membership and endpoints. These structural cases do not perform complete semantic validation or search for a finite witness; their `structural-only` result is retained in every record.
-
-| Workload | Generic actual plus three candidates | Source-owned actual plus three candidates | Request nodes |
-|---|---:|---:|---:|
-| Weather WSDL | 3,370,424 | 1,316,746 | 142 |
-| Shared-recursive fixture | 885,542 | 408,819 | 55 |
-| Synthetic 4 recursive types | 355,831 | 162,966 | 29 |
-| Synthetic 16 recursive types | 1,076,816 | 451,882 | 89 |
-| Synthetic 64 recursive types | 3,964,383 | 1,610,775 | 329 |
-| Synthetic 128 recursive types | 7,820,612 | 3,158,811 | 649 |
-| Synthetic 256 recursive types | 15,540,094 | 6,262,935 | 1,289 |
-
-The synthetic inputs use the existing graded generator: every type shares one two-element group and independently recurses across an element boundary. Weather remains the largest available redistributable real-service closure; these synthetic sizes do not add evidence about larger real services. The generic path calls the structural adapter on a built and resolved graph, while the source-owned path retains #278's authenticated immutable sharing.
-
-The replacement work limit targets at least 50 percent headroom above the largest measured completed request, rounded up to the next 5,000,000 units. The resulting 25,000,000 limit leaves 9,459,906 units, or 60.9 percent of the observed 15,540,094 cost, for input and workload variation. The largest measured request uses only 1,289 nodes, so this evidence gives no reason to raise the existing node guard.
-
-| Default-profile workload | Cold wall ms | Two warm wall ms | Peak process RSS MiB | Maximum observed post-stage heap MiB |
-|---|---:|---|---:|---:|
-| Generic Weather sequence | 51.660 | 30.272, 26.300 | 76.9 | 27.2 |
-| Generic synthetic 64 sequence | 48.786 | 34.264, 25.572 | 81.6 | 32.6 |
-| Generic synthetic 128 sequence | 106.208 | 81.164, 55.500 | 98.2 | 32.5 |
-| Generic synthetic 256 sequence | 147.151 | 108.520, 98.680 | 133.3 | 64.3 |
-| Source-owned synthetic 256 sequence | 111.096 | 76.480, 52.233 | 108.3 | 39.1 |
-
-Wall totals include loading and measured preparation stages but exclude process/module startup. RSS is a process-lifetime high-water observation including the loader and preceding requests; heap is observed at stage boundaries, not its allocation total or guaranteed peak. OS cache, JIT, GC and concurrent environment activity are uncontrolled, so these samples support practical calibration without promising throughput or a memory ceiling.
-
-For generic Weather, actual preparation costs 3,041,697 units and the three deltas cost 140,673, 36,408 and 151,567; the initial terminal reservation adds 79. The 256-type generic case costs 12,971,156 for actual preparation and 1,269,596, 20,235 and 1,279,028 for its deltas, plus the same reservation.
-
-The compact committed summary stores shared inputs and claims once, each profile's result and deterministic stage costs, and three total-wall, peak-RSS and maximum post-stage-heap samples. Full raw reports under `tmp/conformance/budget-calibration/` retain each stage's timing and before/after RSS and heap; the summary records their hashes and reproduction commands. The integrated summary comes from fresh execution of all three profiles; the compact representation retains the measured numerical evidence.
-
-After the integrated scalar and proof repairs, the three complete finite cases cost 918,564 for particle/AU/wildcard, 977,067 for the nested helper and 593,453 for typed attributes. These are new measurements; the historical #278 totals remain in its preceding section. They complete under the explicit former 1,000,000 limit as well as the new default and diagnostic limits. Default and diagnostic runs have identical completed totals, graph hashes and checked receipt projections for corresponding samples; the [reducer](../test/research/re01/finite-capacity-summary.ts) checks these facts before emitting evidence.
-
-The accounting schedule remains unchanged. Its text, descriptor, traversal and allocation precharges account for variable work at finer granularity than graph nodes or transitions; the large numerical totals coexist with modest observed time and memory in this corpus. Rescaling those units would obscure existing independent ledgers without demonstrating an operational benefit, and the provisional inherited stage budgets in ADR-003 concern separate stages and units.
-
-The [capacity controls](../test/research/re01/finite-capacity.test.ts) pin the former allowance where it is the regression oracle. Generic Weather still returns `resource-limit` at 1,000,000, while an exact measured complete sequence succeeds and one unit less fails without a partial candidate. Starvation, source-ownership and symbolic-particle controls also use explicit former limits so a default increase cannot mask those defects.
-
-Oversized text is rejected by its precharge before scanning or copying, repeated cyclic references retain sharing, and reserving 100,001 nodes fails at the inclusive 100,000 boundary. Failure stays terminal with the same envelope and unchanged counters, including subsequent attempted operations. Exhaustion provides no negative semantic answer; neither this numerical calibration nor a successful supplied finite check proves universal coverage or search termination.
-
-Reproduce all profiles and reduce them only after completing a stable source tree:
-
-```bash
-mkdir -p tmp/conformance/budget-calibration
-node --import tsx test/research/re01/finite-capacity-measure.ts --profile legacy > tmp/conformance/budget-calibration/legacy.json
-node --import tsx test/research/re01/finite-capacity-measure.ts --profile default > tmp/conformance/budget-calibration/default.json
-node --import tsx test/research/re01/finite-capacity-measure.ts --profile diagnostic > tmp/conformance/budget-calibration/diagnostic.json
-node --import tsx test/research/re01/finite-capacity-summary.ts tmp/conformance/budget-calibration/legacy.json tmp/conformance/budget-calibration/default.json tmp/conformance/budget-calibration/diagnostic.json tmp/conformance/budget-calibration/summary.json
-```
-
-Loading, canonical construction and reference resolution retain their independent limits; their `null` shared-work counters are not zero-cost claims. The earlier preparation harness still measures inherited gathering/merge/resolution/composition stages separately, and its request-context profile now reads the actual semantic defaults instead of reusing inherited stage-step limits. A request that fits this work guard can still encounter a separate loader, graph or source-stage limit.
-
-### Scalar completeness capacity calibration
-
-The [scalar harness](../test/research/re01/scalar-capacity-measure.ts), [validated summary](../test/research/re01/scalar-capacity-evidence.json) and its four committed raw reports measure 25 independently motivated workload classes with three fresh authenticated requests per profile. Profiles retain 100,000 nodes and use explicit 1,000,000, 25,000,000, default 125,000,000 and diagnostic 250,000,000 work allowances. Each workload/profile starts a fresh child; cold and two warm-process samples have independent request budgets.
-
-Every report hashes the compiler and loader source, fixture, harness and lockfile. The reported HEAD is the starting main a103ca96a3609bd0b525f2331daa7173879ad187; the recorded source digests identify the measured uncommitted implementation, rather than claiming that baseline commit already contains it. The [strict reducer](../test/research/re01/scalar-capacity-summary.ts) checks shared provenance, all three deterministic work/node/result observations, profile equivalence on completed requests and sticky exhaustion before emitting the summary.
-
-| Profile | Complete checked samples | Complete widening samples | Resource-limit samples |
-| --- | ---: | ---: | ---: |
-| Explicit original 1,000,000 | 24 | 3 | 48 |
-| Explicit previous 25,000,000 | 45 | 15 | 15 |
-| Default 125,000,000 | 51 | 21 | 3 |
-| Diagnostic 250,000,000 | 51 | 21 | 3 |
-
-Default and diagnostic decisions, work and node totals are identical for all 25 workloads. All 24 semantic workloads complete; the adversarial expansion deliberately returns resource-limit under every profile. The mixed double/float selected-image contrasts cost 63,778,263 for the positive and 54,652,769 for widening; the previous 25,000,000 allowance exhausts both, while the revised procedures complete them without a membership qualification.
-
-| Default workload | Total request work | Query wall ms, cold then warm | Process peak RSS MiB, all three samples |
-| --- | ---: | --- | --- |
-| Mixed double/float positive | 63,778,263 | 115.196, 118.347, 70.999 | 94.0, 103.5, 103.5 |
-| Mixed double/float widening | 54,652,769 | 140.322, 107.368, 56.368 | 87.3, 107.7, 107.7 |
-| URI/list negative-exponent positive | 73,419,705 | 423.369, 406.575, 365.415 | 113.4, 157.5, 160.0 |
-| URI/list negative-exponent widening | 79,018,339 | 578.656, 539.975, 715.640 | 116.1, 163.5, 212.8 |
-
-The final 125,000,000 default leaves 45,981,661 work units, 58.2 percent above the largest completed observed cost, for workload variation. Measured query latency remains practical across these cases, and the largest process-lifetime RSS observation is 212.8 MiB with maximum observed post-request heap 72.3 MiB. These observations support this engineering allowance; no example throughput figure or fixed timing threshold determines acceptance, and defaults remain revisable with further workload, latency, memory and adversarial evidence.
-
-The scalar cases allocate at most 21 semantic nodes; the earlier structural calibration reached 1,289, so neither measurement supports raising the 100,000 node guard. Independent controls exercise the inclusive node boundary, one beyond it and sticky exhaustion, as well as exact measured work and one unit less. Variable state arrays, masks, namespace copying, endpoint arithmetic and cached-result copies are charged before work or allocation; conservative digit-dependent work units are not milliseconds or byte counts.
-
-The adversarial pattern requests fifty million repeated atoms and is rejected by its expansion precharge before constructing that machine. Every profile stops with 141,951 consumed work and 16 nodes, preserves the identical terminal receipt and invokes no work on retry. An exhausted allowance never supplies containment, exclusion or a partially published cache fact.
-
-Query timing excludes source/candidate preparation; total request work includes it. RSS includes module loading and earlier requests in the same child; heap snapshots are observations rather than guaranteed peaks or allocation totals. JIT, GC and OS activity remain in the samples, which measure local scalar decisions rather than universal RE01 witness adequacy or production activation.
-
-Raw reports are [default](../test/research/re01/scalar-capacity-results/default.json), [original limit](../test/research/re01/scalar-capacity-results/legacy.json), [previous limit](../test/research/re01/scalar-capacity-results/previous.json) and [diagnostic](../test/research/re01/scalar-capacity-results/diagnostic.json). Reproduction commands are stored with each report hash in the summary; run the reducer over those four reports to validate the committed evidence.