Published

Localization / Language

vr.wm-xct-031 · wm-xct-031-localization-language

Provide an embeddable, format-neutral declaration of locale, translation and terminology binding, so that any adopting entity can state which language, script and locale conventions apply to it, express and resolve linguistic preferences, and reference translated resources and convention profiles without duplicating them.

World Models Cross-cutting context XCT.LNG

Bundle → Layer → Finding → Questions Classifiers What it is Why it exists Distinguishing features What robots and AI may and may not do Moral aspects Owners Links to other meta-models Identity and identifiers Direct properties Recognition Capabilities and actions Hazards and failure modes Standards and interfaces Context of use Sources

Bundle → Layer → Finding → Questions Filled

7 bundles · 16 layers · 31 findings · 136 questions

Linguistic and locale identity Everything needed to state, normalize and validate which language, script and locale a host record is bound to, and to keep that binding stable as external registries change.

Language tag identity and registry binding

The governed tag string, its subtag structure and canonical form, and its dereference into the authoritative subtag registry including deprecation and replacement.

Language tag identity, structure and canonical form

The BCP 47 language tag that identifies the linguistic context, the role of each subtag it contains, the canonical and case-regularized form in which it is stored, and whether it is merely well-formed or also valid.

Which language tag identifies this linguistic context, and in which canonical form is it stored? identity
Which subtags does the tag contain and what role does each one play? composition
Is the tag well-formed only, or also valid against the subtag registry? validation
Which casing and subtag ordering conventions were applied, and are they treated as significant for comparison? constraint
Does the tag carry private-use or otherwise unregistered subtags, and what do they mean inside the adopting Dimension? interoperability

Subtag registry binding, deprecation and replacement

The pinned registry snapshot against which subtags are validated, the deprecation status and prescribed Preferred-Value replacement of each subtag, the handling of grandfathered and redundant registrations, and the macrolanguage, collection and extlang caveats that make a syntactically fine tag semantically wrong.

Which registry snapshot, identified by its File-Date, was used to validate these subtags? provenance
Is any subtag or the whole tag deprecated, and what replacement does the registry record prescribe? state
Is the identifier a grandfathered or redundant registration, and how is it mapped forward? exception
Does the primary language subtag denote a macrolanguage, a collection or an extended language, and which encompassed language is actually meant? classification
How is drift between a stored identifier and a newer registry snapshot detected and reconciled? temporal

Script and directionality declaration

The writing system a linguistic context uses and the direction and orientation metadata a consumer needs in order to display it correctly.

Script identity and script-subtag suppression

The ISO 15924 script in which the linguistic context is written, whether it was stated explicitly, deliberately omitted because the registry suppresses it, or inferred by likely-subtag expansion.

Which ISO 15924 script code applies to this linguistic context? identity
Was the script subtag omitted because the registry records a Suppress-Script value for that language? constraint
If the script was not declared, how was it inferred and with what confidence? provenance
Does the content mix writing systems, and how is that recorded without altering the declared tag? composition

Base direction and writing-mode hints

The declared base paragraph direction and layout orientation that a consumer needs in order to display a string correctly, recorded as explicit metadata rather than guessed from the characters in the value.

What base direction is declared for this string or resource, and at which scope? definition
Was the direction stated explicitly or produced by a first-strong heuristic, and what residual risk remains? quality
Which character order and line order apply to the layout of this content? classification
How are language and direction metadata preserved when a value is embedded in another string or transferred between systems? interoperability

Locale identity and convention bindings

The locale identifier that extends language identity with Unicode extension fields, and the versioned references to the externally governed convention profiles that those fields select.

Unicode locale identifier and extension fields

The locale identifier as distinct from the plain language identifier: the -u extension attributes and keywords that refine behaviour, the -t extension that records transformed content and its source language, and the canonicalization that makes two such identifiers comparable.

Which Unicode locale identifier applies here, and how does it differ from the plain language identifier recorded for the same context? identity
Which Unicode extension keys and types are asserted, and which values are merely environment defaults? composition
Does a transformed-content extension declare that this value was derived from another language, and from which source? provenance
Which canonicalization was applied to the extension sequence, and were duplicate attributes or keys dropped? validation
How does the consuming boundary handle extension keys it does not recognise or support? exception

Referenced formatting and collation convention profiles

The bindings from a locale declaration to externally maintained number, date, time, unit and collation convention profiles, held strictly as selections and versioned references rather than as duplicated convention data.

Which convention profiles does this locale declaration select, and where are their definitions maintained? relationship
Which data release supplied each referenced profile, and is that release pinned or floating? provenance
Where does responsibility for the actual formatting, parsing or sorting behaviour sit? ownership
What is recorded when a selected convention profile is unavailable in the consuming environment? exception

Preference, availability and locale resolution How a linguistic preference is declared, what is actually available to satisfy it, and how the two are reconciled into a single resolved locale with a recorded, reproducible reason.

Declared preferences and available inventory

The requested side and the supply side of negotiation: an ordered list of language ranges with its provenance, and the set of locales for which resources actually exist together with the configured default.

Ordered language preferences and language ranges

The requested, ordered or weighted list of language ranges that expresses a preference, the range syntax used, where the preference came from, and the limits on reusing it beyond the interaction that produced it.

Which ordered language ranges express the requested preference, and in what priority order? composition
Are the supplied ranges basic or extended, and do any of them contain wildcards? classification
Are relative weights attached to the ranges, and how are ties and zero weights treated? measurement
Where did this preference come from, and is it a stated choice or an inferred signal? provenance
What limits apply to reusing this preference outside the interaction that produced it? privacy

Available locale inventory and default locale

The set of locales for which resources actually exist within a defined scope, how complete each of them is, the configured default when nothing matches, and how the inventory is versioned and refreshed.

Which locales are available within this scope, and at what granularity is availability asserted? definition
How complete is each available locale, and what threshold counts as available at all? quality
Which locale is the configured default when no requested range matches anything available? decision
How often is the inventory regenerated, and how are stale entries detected? temporal

Matching, resolution and declaration role

The reconciliation of preference against availability into one resolved locale with an explicit fallback reason, and the role a given declaration plays relative to content, audience and interface.

Locale matching, resolution outcome and fallback reason

The record of how a requested preference was matched against the available inventory: which scheme was applied, which locale was resolved, which fallback steps were traversed, why, and against which data versions the result can be reproduced.

Which matching scheme was applied, with which parameters, and is it deterministic across hosts? process
Which locale was resolved, and is it an exact match, a truncation or the configured default? state
What fallback chain was traversed, and what reason is recorded for each step? evidence
When was the resolution performed, and against which inventory version and registry snapshot? temporal
Who may override or re-run a resolution, and where is the override recorded? authority

Content language versus interface locale and declaration provenance

Which role a locale declaration plays - the language of the content itself, the audience it is intended for, the user interface locale, or the source language of transformed content - who asserted it, and how conflicts between declared and detected values are settled.

What role does this declaration play: the language of the content, the intended audience, or the interface locale? classification
Which party or system asserted this declaration, and on what authority? ownership
Was the language declared by an authority or automatically detected, and how is a disagreement between the two settled? decision
Which declaration prevails when several apply at different scopes to the same content? constraint

Translatable Unit and Target Variant Binding What is identified as translatable, what constrains it on the source side, and how a target language variant is bound to it with a revision, a state and a verifiable reference to the resource that actually holds the text.

Translatable unit identity and source-side constraints

The source side of the binding: what the unit is, which source revision is current, what disambiguates its meaning, and what must not be changed or must be limited in any target.

Stable translatable-unit identity, source revision and disambiguating context

A translatable unit is identified by a key that is independent of the source text it currently carries, scoped to a declared namespace, and paired with the identifier of the source revision a translation was made from, plus the notes that fix its intended meaning, purpose and audience.

Which identifier designates this translatable unit independently of its source text, and within which scope is that identifier unique? identity
Which source-text revision is currently in force for this unit, and how is that revision denoted and verified? provenance
What context, meaning, purpose or audience notes disambiguate this unit for a translator or an agent? definition
What must happen to existing bindings when the source text changes while the unit identifier stays the same? exception
Which content node, message or media element in the owning system does this unit correspond to? relationship

Translatability scope, protected tokens and target-side limits

Which spans are translatable at all, which literals must survive verbatim, which locales the unit applies to, and what character-set and storage-size limits any target text must respect.

Is this unit or span marked as translatable, and which rule or inheritance established that value? classification
Which literal tokens, codes or brand strings must appear unchanged in every target variant? constraint
To which locales does this unit apply, and is that list inclusive or exclusive? requirement
How is a breach of a protected token, allowed-character set or storage-size limit detected before publication? validation

Target variant binding, revision, state and external resource reference

The target side of the binding: which language variant exists, which revision it is, what state it holds, who may advance that state, and where the text or media actually lives.

Target variant binding, revision, staleness and translation state

A binding pairs one translatable unit with one BCP 47 target language tag and one variant revision, records the source revision it was made from and whether it is now stale, carries the workflow-independent translation state and approval, and separates the time the translation was produced from the time this model observed it.

Which canonicalised language tag and variant revision together identify this target variant of the unit? identity
What translation state does this variant hold, and who is authorised to advance it to the next state? authority
Is this variant stale relative to the current source revision, and by which comparison was that decided? state
When was this variant produced, reviewed or approved, and when was each of those facts recorded here? temporal
Was the target text produced by a machine, by a human, or by a mixed post-editing process? provenance

External resource, catalogue and non-text alternative references

Verifiable references from the binding to the systems that actually hold the target text, media asset or alternative, pinned by digest and resolution time, including locale-specific non-text and time-based media alternatives.

Which external resource holds the actual target content, and how is it addressed and dereferenced? interoperability
What digest or version pin makes the referenced resource verifiable at a stated point in time? evidence
Which non-text or time-based media alternatives must accompany this variant in this locale? composition
What is asserted when the referenced resource is unreachable or its digest no longer matches? exception

Message Form Contract and Fallback Resolution The formal contract a target variant must satisfy to be substitutable for its source, and the declared, ordered behaviour when no acceptable variant exists.

Selection and placeholder contracts

The structural obligations a target variant inherits from its source: which categories it must supply and which variables, placeholders and codes it must reproduce.

Plural, ordinal and select categories with selector declarations

Which selectors a message declares, in what order they are evaluated, which language-specific plural or ordinal categories the target language requires, and whether a catch-all variant guarantees that every selector value resolves.

Which selectors does this message declare, and in what order are they evaluated against variant keys? composition
Which plural or select categories must the target variant supply for its own language? requirement
Is a catch-all variant present so that every possible selector value resolves to a pattern? validation
Are the categories cardinal plural, ordinal, or an enumerated select set defined by the message author? classification

Variables, placeholders, type constraints and code isolation

The set of named variables and placeholders the target must reproduce, the value-type and option constraints on each, how original formatting codes and markup are isolated from translatable text, how bidirectional isolation is applied, and what is emitted when a placeholder cannot be resolved.

Which named variables and placeholders must appear in the target, and may they be copied, deleted or reordered? constraint
What value type, function or option constraints apply to each placeholder? requirement
How are original formatting codes and markup kept isolated from the translatable text of this unit? composition
How is bidirectional isolation applied around placeholders in mixed-direction text? quality
What is emitted when a placeholder or variable cannot be resolved at format time? exception

Fallback chain, missing-translation policy and override provenance

The declared ordered degradation path when a requested variant does not exist, what happens at the end of it, which layer actually supplied a served value, and how that fact is disclosed.

Ordered fallback chain, terminal behaviour and override provenance

An explicit ordered list of language tags or locales tried when the requested variant is absent, the mandatory defined behaviour when nothing matches, the record of which chain position supplied a served value and whether that entry was inherited or an override, and the prohibition on presenting a fallback as a translation.

What is the ordered list of language tags or locales tried when the requested variant is absent? process
What is the terminal behaviour when no entry in the chain matches? decision
Which chain position actually supplied the value served, and was that entry inherited or an override? provenance
Is the consumer told that the served text is a fallback rather than a translation into the requested language? interoperability
Under what conditions is fallback prohibited and an explicit absence or error preferred instead? exception

Terminology Binding, Quality Assertions and Verification Evidence What a term occurrence is required to mean, how good a variant is asserted to be, who vouched for it, and what testing shows about whether it actually renders and is usable.

Terminology and concept binding

Binding a term occurrence to a language-independent concept in an externally owned termbase, and stating the usage constraint that applies to it in this locale.

Terminology concept binding and term usage constraints

Which concept a marked term occurrence denotes, in which termbase that concept lives, which target term is preferred, admitted or deprecated for it in the target locale, how a usage breach is recorded, and what evidence supports an automatic term identification.

Which terminology concept does this term occurrence denote, and in which termbase does that concept live? relationship
Which target term is preferred, admitted or deprecated for that concept in this target locale? classification
How is a term-usage breach recorded when the approved target term is not used? validation
What evidence supports a term identification made automatically rather than by a terminologist? evidence

Quality assertions, review provenance and rendering verification

Asserted quality and coverage, the agents who vouched for it, and the independent observations that show whether the localized product actually renders and is usable.

Confidence, quality issue, rating and coverage measures

Machine confidence on its declared scale, recorded quality issues with type and severity, any overall rating computed against a named profile and threshold, and coverage measures that state exactly what they count and what they exclude.

What confidence value does the producing system assert for this variant, and on which declared scale? measurement
Which quality issues were recorded against this variant, of what type and at what severity? quality
Against which published quality profile and score threshold was any overall rating computed? evidence
What proportion of the unit set has an approved variant in this locale, and how is that proportion defined? measurement
Which entries does the coverage measure exclude, and is fallback-served content excluded from it? validation

Translation and review agent provenance

Which person, organisation or tool produced and which revised the target text, what review occurred and under whose authority it was signed off, where the full workflow history is held, and how attribution personal data is minimised and retained.

Which person, organisation or tool produced the target text, and which agent revised it? provenance
What review was performed on this variant, and under whose authority was it signed off? authority
Where is the full workflow task history held, and what exactly is referenced from here? ownership
What personal data appears in translator and reviewer attribution, and how is it minimised? privacy
How long must a review assertion be retained after the variant it describes is superseded? retention

Pseudo-localization and rendering verification evidence

Observed evidence that the localized surface actually renders: pseudo-locale runs that expose hardcoded strings, concatenation, expansion and bidi defects, measured fit against declared size limits, the build revision verified, and an explicit statement of what such testing cannot show.

Which pseudo-locale runs were executed against this unit set, and what defects did they expose? evidence
Did the rendered target text fit the declared size, storage and layout constraints? measurement
Which build or product revision was verified, and when was the run observed? temporal
Which defect classes remain untested by pseudo-localization and rendering checks alone? exception

Localization governance, authority and conformance Who is accountable for a localization binding, who may declare its defaults and fallback, how the origin and pinned inputs of every bound value are recorded, and how the binding is validated, reconciled, disputed and mapped to external carriers.

Stewardship, role rights and declaration authority

Accountable ownership of the binding, the roles permitted to propose, vet, review, approve and release changes, and the authority to declare source, default and fallback locales under a regional or organizational profile.

Accountable stewardship and decision rights over the binding

Names the party accountable for the localization binding on its host object, the roles authorized at each decision point, how distributed locale expertise is weighted, and what happens when no qualified authority exists for a target locale. Roles are references to parties governed elsewhere; only the assignment is owned here.

Which named party is accountable for this localization binding on the host object? ownership
Which roles are authorized to propose, vet, review, approve and release a change to this binding? authority
How is contributor standing weighted when locale expertise is distributed across several organizations? decision
What is recorded when no qualified language expert or release authority is available for a target locale? exception

Authority to declare default locale, fallback chain and locale profile

Establishes who may declare the host's source locale, default locale, permitted-tag set and ordered language priority list, which matching scheme governs resolution, and the mandatory behaviour when nothing matches. Profiles are declarative constraint sets; the resolution engine that consumes them is external.

Who is authorized to declare the default locale and the ordered fallback chain for this host object? authority
Which matching scheme governs resolution, and what behaviour is defined when no tag matches? decision
Which regional or organizational locale profile constrains the tags permitted on this binding? classification
Which subtag combinations does that profile require, prohibit or normalize away? constraint

Provenance and revision pinning

The origin, agent, method and confidence of each bound value, and the exact registry, profile and standard revisions under which the binding was created and must be re-evaluated.

Provenance of tags, translations, terminology and inferred values

Records how each bound value came to exist: human declaration, derivation from a registry record, machine detection or machine translation, together with the responsible agent, the activity, the prior version it revises and any confidence measure. Uses PROV starting-point terms rather than a local vocabulary.

Which agent generated or revised each bound language tag, translation or term, and in what role? provenance
What evidence supports an inferred or machine-generated value, and what confidence accompanies it? evidence
Which prior binding version or upstream source is this value derived from or a revision of? relationship
At what confidence threshold must an inferred locale or translation be escalated to human review? quality

Registry, profile and standard revision pins

Captures the exact external revisions under which a binding was interpreted and judged valid, so that a later registry or profile advance produces a detectable re-validation obligation rather than a silent change in meaning.

Which language subtag registry revision was in force when this binding was validated? temporal
Which locale-data release and profile revision were pinned for fallback and likely-subtag resolution? provenance
Which exchange-format and terminology standard versions does this binding claim alignment with? interoperability
What triggers re-validation when a pinned registry, profile or standard revision advances? process

Validation, conflict disposition and interoperability

Checks that determine whether a binding is well-formed, valid, canonical and profile-conformant; how contradictory localization claims are ranked and disposed; and how the binding is mapped onto external carriers as evidence-backed alignments.

Validation, canonicalization and reconciliation of declared against observed language

Defines the conformance checks applied to a binding and how a declared tag is reconciled against detected or transport-asserted content language, while preserving the originally recorded tag so that canonicalization and deprecation mapping never destroy the input.

Is the bound tag well-formed, valid against the pinned registry, and in canonical form? validation
Does the binding satisfy the declared profile's permitted-tag and required-subtag constraints? constraint
How is a deprecated subtag with a preferred value mapped without silently rewriting the recorded original? process
How is the declared tag reconciled against detected or transport-asserted content language? quality

Conflict disposition and declared review limitations

Records how competing localization claims are ranked and disposed, how macrolanguage and specific-language tensions are represented, and where accessibility or cultural review could not be completed. Disposition decisions are localization decisions; enforcement of the outcome belongs to the consuming system.

Which precedence rule applies when the declared tag, the transport-level claim and the profile default disagree? decision
How is a conflict between a macrolanguage tag and a more specific individual-language tag recorded? relationship
What is recorded when a cultural or accessibility review cannot be completed for a target locale? exception
Which accessibility obligations does this binding discharge, and which remain with the rendering layer? requirement

Interoperability crosswalks and export bindings

Declares how the binding maps onto external localization carriers, with each mapping recorded as an evidence-backed alignment and an explicit lossiness statement rather than an unqualified conformance claim.

Which external carrier fields does this binding map onto for interchange? interoperability
Is each mapping lossless, lossy or one-way, and what is dropped on round-trip? constraint
Which identifier correlates the binding with an external localization unit across systems? identity
What evidence supports each claimed alignment, and is conformance asserted or only alignment? evidence

Binding lifecycle, disclosure and disposition The governed lifecycle of the binding on its host object, and the declarative rules that control who may see localized values, how long binding records are kept, and what survives their withdrawal.

Binding state, version and supersession

Workflow state and permitted transitions for the binding, its monotonic version and effective interval, and how superseded or deprecated versions are represented so that correction never rewrites history in place.

Binding state, version, effective interval, supersession and deprecation

Governs the lifecycle of the binding itself on its host object: current workflow state and permitted transitions, monotonic version, interval over which a version is authoritative, supersession by a successor, deprecation with replacement guidance, and correction performed by issuing a superseding version rather than mutating a published one.

What is the current workflow state of this binding and which transitions are permitted from it? state
Over which effective interval is this binding version authoritative for the host object? temporal
Which binding version supersedes or is superseded by this one, and on what ground? lifecycle
How is an erroneous published binding corrected without destroying the superseded record? event
When is a binding deprecated with replacement guidance rather than withdrawn outright? decision

Disclosure, redaction, retention and disposition

Declarative sensitivity, embargo and locale-filter rules for localized values and their export redaction, and the retention class, legal hold status, integrity evidence and tombstone content that govern binding records at end of life.

Sensitivity, embargo and finding/artifact disclosure of localized content

Classifies which localized findings or referenced resources are restricted, embargoed until a release moment, or filtered to specific locales, and states the redaction or substitution applied on export. Authorization scope is expressed only at supported bundle, layer, finding and artifact grains; evaluation and enforcement remain external.

Which localized findings or artifacts are restricted, and to which audience classes? access
Until what moment is a localized variant embargoed, and what may be disclosed before it? access
Does the localized value contain personal or otherwise regulated data that constrains disclosure? privacy
What redaction or substitution is applied when an unauthorized audience requests the binding? access

Integrity, retention, legal hold and tombstone declarations

Declares the retention class and minimum period for binding records, whether a hold blocks disposition, what integrity evidence proves a superseded binding is unaltered, and what a tombstone must preserve. Appraisal authority and the physical execution of destruction or transfer sit with the records system.

What retention class and minimum retention period apply to this binding record? retention
Is the binding subject to a hold that blocks disposition, and who asserted it? retention
What integrity evidence shows that a superseded binding version is unaltered? evidence
What must a tombstone preserve after a binding is withdrawn, and who executes the destruction? retention

Classifiers Filled

Family: World Models
Category: Cross-cutting context
Entry kind: mixin
Navigation path: NAV.XCT.LNG
Domain: XCT.LNG
Industry: Cross-industry
Tags: localizationlanguagexct.lng

What it is Filled

WM-XCT-031 is a mixin that attaches linguistic and locale context to a host record. It covers BCP 47 language-tag identity and canonical form; the binding of subtags to the IANA Language Subtag Registry including deprecation and replacement; Unicode locale identity with -u and -t extension fields; script identity and base-direction/writing-mode hints; ordered preference declarations, available-locale inventory and the explicit resolved-locale outcome with fallback reasons; the role distinction between content language, intended audience and user-interface locale; versioned references to translated resources, terminology assets and formatting/collation convention profiles; and the governance needed to keep every declaration reproducible against a pinned data version. It declares context; it does not hold localized content, run translation production, identify people or places, or execute access, retention or audit decisions.

In scope

BCP 47 language-tag identity, well-formedness, validity and canonical form, covering language, extlang, script, region, variant, extension and private-use subtags
Binding to the IANA Language Subtag Registry, including Deprecated and Preferred-Value handling, grandfathered and redundant registrations, Scope classification and macrolanguage caveats
Unicode locale identity as distinct from language identity, including -u and -t extension fields, their canonicalization and likely-subtag expansion
Script identity by ISO 15924, Suppress-Script handling, declared base direction and character/line order hints carried together with the value they qualify
Ordered or weighted preference declarations, basic and extended language ranges, available-locale inventory and the configured default locale
Matching, filtering and lookup outcomes recorded as requested, available and resolved locale with an explicit fallback chain and per-step reason
Role distinction between content language, intended audience and user-interface locale, and between declared, detected and negotiated values
Versioned references to translated resources, terminology assets and number, date, time, unit and collation convention profiles
Governance of the mixin itself: pinned registry and CLDR data versions, canonicalization policy, provenance fields and reproducibility records

Out of scope

The localized content itself: translated text, media, their versions and editorial state
Translation production workflow, vendor assignment, translation memory operation and terminology approval processes
Person, party or account identity; a language preference is bound to a party by the party model, not defined here
Geographic territories, boundaries, political status and legal jurisdiction; a region subtag is neither a location nor a legal fact
Calendar, numbering, collation, measurement and time-zone rule data; only selection keys and pinned references are held
Character encoding, Unicode normalization, text segmentation, font selection, shaping and the bidirectional algorithm itself
Machine-translation and language-detection engines; only their identity, output and confidence are recorded
Execution of access decisions, erasure requests, retention schedules and audit-trail storage
Currency amounts, pricing and monetary conversion, even where a currency key appears in a locale extension

Why it exists Filled

Distinguishing features Filled

A mixin that attaches language and locale context to a host record using BCP 47 tags.
Separates language identity from region, so a language is not inferred from a country.
Covers preference resolution, fallback chains and translation bindings, not the translation work itself.
Leaves formatting rules to CLDR data rather than restating them.

What robots and AI may and may not do Filled

Must not

Infer a person's language or nationality from location or name.
Publish machine translation as reviewed without review.
Drop placeholders or markup that change meaning.
Override a person's declared language preference.
Use deprecated subtags without recording their replacement.

Only with a human decision

Approving translations of legal, medical or safety text.
Releasing a new locale to users.

May

Parse, canonicalize and validate language tags against the subtag registry.
Resolve a user's declared preferences to a supported locale with fallback.
Check placeholder and markup parity between source and translation.
Record review assertions on translations.

Moral aspects Filled

Language access affects whether people understand their rights, risks and instructions.
Minority and indigenous languages are often under-served and need deliberate support.
Language preference can reveal ethnicity or origin and should be handled as sensitive.

Who is affected

Speakers of the languages served
Translators and reviewers
People who depend on accurate safety or legal text

Owners Filled

Steward

The adopting Dimension names a single accountable owner for localization bindings within its scope and a distinct release authority, and records both as resolvable party references rather than free text.

Roles

Localization binding owner: Holds accountability for the binding on its host object and keeps the stewardship charter current; Nominates language experts, translators, reviewers and approvers within the governed scope; Records the accountable delegation chain when accountability moves; Decides whether an unstaffed target locale is left provisional or removed from the profile
Language and locale expert: Declares and justifies the tag, script, region and variant values for a locale within their competence; Advises on macrolanguage versus individual-language selection and on region subtag choice; Records declared limitations where locale conventions are disputed or insufficiently evidenced; Assesses whether a detected or machine-generated value meets the confidence threshold
Translator: Produces target-locale values bound to a source unit and records the generation method used; Attaches terminology references for the host subject and flags terms lacking a concept entry; Advances a binding to a translated state without approving or releasing it; Records machine-translation engine identity, version and confidence where a value is post-edited
Localization reviewer: Verifies a binding against the governing profile, the validation report and terminology bindings; Records review outcome, findings and any cultural or accessibility limitation that could not be closed; Returns a binding for correction rather than editing it when the proposer is better placed to fix it; Never approves a binding version they reviewed, preserving separation of duties
Release authority: Approves a binding version and sets its effective-from and optional effective-to moments; Closes the predecessor version's effective interval at the successor's effective-from moment; Approves early lifting of an embargo and records the basis for doing so; Authorizes deprecation with a preferred replacement and a sunset date
Localization records custodian: Binds the model's records to a retention class and monitors the disposition trigger; Records hold assertion and release with the asserting authority, and blocks disposition while a hold is in force; Prepares the tombstone payload and hands execution to the records system that owns destruction and transfer; Maintains integrity digests and escalates verification failures as integrity-suspect

Links to other meta-models Filled

aligned

IETF BCP 47 language tag and language tag matching standard - Adopt the normative tag syntax, subtag roles, canonical form, well-formed versus valid distinction and the basic filtering, extended filtering and lookup schemes. Alignment is claimed for the structures actually cited; no conformance claim is made for behaviour this model does not execute.
Unicode CLDR and UTS #35 locale identity and locale data - Adopt the Unicode locale identifier, its canonical syntax and alias-based canonicalization, likely-subtag expansion, the -u and -t extension field structures, and the layout orientation elements. Divergence from BCP 47 canonical form is recorded as a conflict rather than silently reconciled.
Language range matching and lookup behaviour defined by RFC 4647 - Bind the declared fallback chain to a named matching scheme and to the requirement that defaulting behaviour be defined when no tag matches; the algorithm itself is not reimplemented here.
Unicode CLDR/LDML locale inheritance, plural rules and language matching reference data - Cite the released plural category sets, parent-locale chains and root fallback as reference data that constrains what a target variant must supply, without forking or restating that data locally.
Unicode MessageFormat 2 message, selector and placeholder model - Ground the selection and placeholder contract in a normative message model, including catch-all variants, option and operand type constraints and format-time fallback representations; the formatting runtime remains external.
OASIS XLIFF bitext interchange documents and their optional modules - Map unit and segment identity, state and sub-state, inline codes, notes and validation rules to an interchange vocabulary so bindings survive round-tripping through localization tooling.
W3C ITS 2.0 data categories - Adopt named data categories for translatability, notes, terminology, locale filtering, provenance, external resources, quality issues, ratings and MT confidence rather than inventing parallel vocabulary.
Accessibility conformance model covering non-text and time-based media alternatives - Record that locale-specific text alternatives, captions and audio descriptions are bound and evidenced, and that language of page and of parts must be programmatically determinable; conformance evaluation belongs to that model.
IETF BCP 47 (RFC 5646 language tag syntax and RFC 4647 matching) - Adopt the normative tag grammar, subtag types, well-formedness versus validity distinction, canonical formatting, and the filtering, lookup and mandatory no-match behaviour that govern resolution.
Unicode CLDR locale identifiers and locale data (UTS #35) - Bind canonicalization, likely-subtags handling and parent-locale fallback chains to a pinned locale-data release, while leaving locale data maintenance and its release approval with the CLDR Technical Committee.
OASIS XLIFF 2.1 translation interchange - Map source and target locale binding and workflow state onto the interchange format's attributes and modules for export, without owning the interchange document, its segmentation or its inline markup.
W3C Internationalization Tag Set (ITS) 2.0 data categories - Reuse Language Information, Translate, Terminology, Domain, Locale Filter, Provenance, Quality Issue and Rating, and MT Confidence rather than inventing local equivalents.
W3C PROV-O provenance ontology - Express agent attribution, generation, derivation and revision of bound values in a standard vocabulary; graph-scale provenance storage and inference stay with the provenance model.

references

IANA Language Subtag Registry - Dereference every subtag to its registry record for validity, deprecation, Preferred-Value replacement, Prefix, Suppress-Script, Macrolanguage and Scope. This model pins a snapshot by File-Date and carries record references; it does not maintain, extend or republish the registry.
ISO 15924 script code registry - Bind the script subtag to an authoritative four-letter script code and its record. Script naming and code assignment stay with the registration authority; no direction property is imported because the consulted code list publishes none.
Party, person and account profile model of the adopting Dimension - Carry the preference declaration and its provenance while the party model owns the binding of that preference to a person or account. A language preference is never treated here as an identity attribute or as evidence of capability.
Localized content, document and terminology resource models - Point from a locale declaration to the translated resource or terminology asset it applies to, by identifier and version. Content storage, versioning, editorial state and translation production remain entirely with those models.
Access control, privacy, retention and audit models of the adopting Dimension - Supply the retention class, linkability classification and reproducibility fields that those models need. Policy evaluation, enforcement, erasure execution and audit-trail storage are owned and performed there; this reference grants this model none of those semantics.
BCP 47 language tag identity governed by the IANA Language Subtag Registry - Take language, script, region and variant identity from the governed registry, including Preferred-Value canonicalisation of deprecated subtags, instead of minting local language codes.
Terminology resource governed as a TBX termbase - Carry outgoing concept references and locale-scoped usage constraints while concept definition, term admission and termbase versioning stay with the terminology system.
Translation management, translation-memory and localization workflow systems of the adopting Dimension - Reference match candidates, job records and task histories held elsewhere; this model records the resulting binding and assertions and does not execute or reconstruct workflow.
HTTP content negotiation and Content-Language exposure at the delivery runtime - Supply the binding and fallback declaration that a negotiating runtime consumes, and require that the served representation's actual language be declared; negotiation execution and its fingerprinting exposure are owned by the runtime.
IANA Language Tag Extensions Registry (u and t singletons) - Constrain which extension singletons a locale profile may permit, and attribute their definition to the registered authority.
ISO 30042:2019 TermBase eXchange terminology resources - Reference concept entries and term records for the host subject and carry the correlation identifier and dialect pin, leaving termbase structure and maintenance external.
ISO 639-3 and ISO 15924 registration authorities - Attribute language and script code assignment, change and retirement to their registration authorities so no local code lifecycle is created.
IETF RFC 9110 HTTP content negotiation - Bind the resolve and export surfaces to proactive negotiation semantics and record the documented hazard that a content language header states intended audience rather than the language of all text.
W3C WCAG 2.2 accessibility conformance - Supply the programmatically determinable language value for a page or passage and record which conformance obligations remain with the rendering and accessibility layers.
Adopting-Dimension access-control, policy-evaluation and audit model - Hand declared sensitivity classes, embargo moments, locale filters and redaction strategies to an external component for evaluation and enforcement, and reference its audit records without owning audit-trail semantics.
Adopting-Dimension records retention and disposition model - Bind retention class, hold status and tombstone content to a governing schedule whose appraisal authority, disposition execution and transfer remain external.

composes

Any adopting Dimension entity that carries linguistic context, such as content records, catalogue entries, party preferences and message payloads - Embed the locale declaration alongside the value it qualifies so that value, language and direction travel together. The host entity keeps its own identity, lifecycle and access class; this mixin adds only the linguistic declaration and its provenance.
Translatable content entity of the adopting Dimension (message, field, document node or media asset) - Attach translatable-unit identity, target variant bindings, fallback declaration, terminology constraints and quality assertions to a subject that already exists, without becoming the store of that subject's content.
Host object model designated by the adopting Dimension - Attach language, locale, translation and terminology binding plus its governance to an existing object, field or resource without giving the binding an independent identity or lifecycle.

extends

Regional or organizational locale profile registry - Specialize the permitted-tag set, required and prohibited subtags, extension policy, default locale and fallback chain for a named region or organization, without duplicating the generic identity, authority or conflict machinery defined in this model.

neighbor

Localized content and resource models (documents, strings, media assets) - This model declares which language, script and locale apply to a value or resource; the resource model owns the text, its versions, its editorial state and its publication lifecycle.
Party, person and account profile models - A stated language preference is an attribute the party model binds to a party. This model defines the shape and provenance of the preference declaration only, and never treats a preference as an identity attribute or capability assertion.
Geographic region, territory and jurisdiction models - A region subtag identifies a locale variant. Territory geometry, political status, nationality and legal jurisdiction belong to those models and must never be inferred from a language or locale tag.
CLDR convention data and its implementations (calendars, collations, numbering, measurement, time zones) - This model records the selection keys and the pinned data version. The convention definitions and the formatting, sorting and parsing behaviour remain in the referenced data and the implementing library.
Translation production and terminology management systems - This model binds a resource to a locale and can record the source language of transformed content. Job state, vendor assignment, review gates and terminology approval are owned by those systems.
Access control, privacy and audit models of the adopting Dimension - This model classifies preference data as potentially linkable and supplies reproducibility fields. Policy evaluation, enforcement, erasure execution and audit-trail retention are owned and executed there.
Text rendering and bidirectional layout engines - This model supplies the paragraph direction and orientation hints as metadata. The Unicode bidirectional algorithm, shaping, line breaking and glyph selection are performed by the rendering engine.

What else AI and robots need to interact with it Filled

Capabilities and actions required Filled

Parse and canonicalize a language or locale identifier: Parses a supplied identifier into its subtags by role, applies the stated case and ordering normalization, and returns the canonical form together with the parse result and the list of normalizations applied.
Validate subtags against the pinned subtag registry: Checks every subtag of a well-formed identifier against a pinned subtag registry snapshot and reports validity, deprecation, the prescribed replacement, and the prefix and script-suppression advisories attached to the matching records.
Filter or look up tags against language ranges: Applies basic filtering, extended filtering or lookup to compare a language priority list against a set of candidate tags, returning either the matching set in priority order or the single closest match with its truncation steps.
Resolve declared preferences to a resolved locale: Resolves an ordered preference declaration against an available locale inventory and a configured default, producing the resolved locale, the outcome class, an ordered fallback chain with a reason per step, and the extension values actually honoured.
Bind target variant and translation resource: Create or update the binding between a translatable unit and one target language variant, pinning the source revision, the variant revision, the origin class and a digest-verified reference to the resource that holds the text.
Resolve declared fallback chain: Given a requested language priority list and the declared fallback chain profile, select the binding that supplies a value, or return the declared terminal outcome, and label the result as translation or fallback.
Validate placeholder, selection and markup parity: Compare a target variant against the source unit's declared selection and placeholder contract, checking placeholder presence and arity, option and operand type constraints, catch-all variant presence, inline code disposition and bidirectional isolation.
Apply terminology constraints to a variant: Resolve the concept bindings in scope for a unit, determine the preferred or admitted target term for the target locale, and record compliance or breach findings for the variant.
Record review and quality assertion: Append an immutable assertion about a variant covering machine confidence, quality issues, any overall rating against a named profile and threshold, coverage measures with stated exclusions, and the agents and times involved.
Check binding-operation role eligibility: Advises whether the requesting principal's asserted role is listed as eligible for a requested operation and returns the governing role or rule references for evaluation by the adopting Dimension's external authorization component. It never returns or enforces an allow/deny decision; identity proofing, policy evaluation, credential handling and audit remain external.
Read binding: Returns the localization binding for a host object at a requested version or effective instant, applying declared finding- and artifact-level disclosure rules to the response.
Bind locale to host object: Creates a localization binding on a host object with a recorded original tag, source and default locales, priority list, matching scheme and no-match behaviour, using a caller-supplied idempotency key so retries do not create duplicates.
Update binding: Applies a partial change to an unpublished binding version under optimistic concurrency, rejecting the call when the supplied version token does not match the stored one.
Unbind locale from host object: Detaches a localization binding from its host object while preserving referential integrity for anything that cites the binding.
Resolve locale for a request: Resolves a requested language priority list against the bindings available on a host object using the declared matching scheme, returning exactly one result for lookup or zero or more for filtering, and applying the declared no-match behaviour when the list is exhausted.
Validate binding: Runs well-formedness, validity, canonicality and profile-conformance checks against the pinned registry and profile revisions, and produces a dated validation report without mutating the recorded original tag.
Reconcile competing language claims: Compares the declared binding value with detected content language, transport-level claims and profile defaults, applies the declared precedence rule and emits a conflict disposition with rationale.
Review binding: Records a review outcome against a binding version, including declared limitations where a cultural or accessibility review could not be completed.
Approve and release binding version: Applies the release authority's approval, sets the effective interval and moves the binding version to a published state; subsequent change proceeds by supersession only.
Supersede or deprecate binding version: Issues a successor binding version that supersedes a published one, or marks a version deprecated with replacement guidance and a sunset date, preserving the superseded record intact.
Export binding with redaction: Projects the binding into a requested external carrier using the declared crosswalk, applies the disclosure profile's redaction rules for the requesting audience, and states what the projection loses.
Dispose binding and write tombstone: Evaluates retention class and hold status for a binding record, records the disposition decision and writes the tombstone, then hands physical destruction or transfer to the records system that owns execution.

Sources Filled

RFC 5646: Tags for Identifying Languages - Internet Engineering Task Force (IETF)
RFC 4647: Matching of Language Tags - Internet Engineering Task Force (IETF)
IANA Language Subtag Registry - Internet Assigned Numbers Authority (IANA)
UTS #35: Unicode Locale Data Markup Language (LDML), Part 1: Core - Unicode Consortium
UTS #35: Unicode Locale Data Markup Language (LDML), Part 2: General - Unicode Consortium
RFC 9110: HTTP Semantics - Internet Engineering Task Force (IETF)
RFC 6067: BCP 47 Extension U (Unicode Locale) - Internet Engineering Task Force (IETF)
RFC 6497: BCP 47 Extension T - Transformed Content - Internet Engineering Task Force (IETF)
ECMA-402: ECMAScript Internationalization API Specification - Ecma International, TC39
HTML Living Standard: The lang, xml:lang and dir attributes - WHATWG
Strings on the Web: Language and Direction Metadata - World Wide Web Consortium (W3C) Internationalization Working Group
Choosing a Language Tag (W3C Internationalization Questions) - World Wide Web Consortium (W3C)
ISO 15924 Code Lists - Unicode Consortium, acting as ISO 15924 Registration Authority
Unicode Locale Data Markup Language (LDML) Part 9: Message Format (UTS #35) - Unicode Consortium
Unicode Locale Data Markup Language (LDML) Part 3: Numbers - Language Plural Rules (UTS #35) - Unicode Consortium
Internationalization Tag Set (ITS) Version 2.0 - World Wide Web Consortium (W3C)
XLIFF Version 2.2. Part 1: Core - OASIS XML Localisation Interchange File Format TC
XLIFF Version 2.2. Part 2: Extended - OASIS XML Localisation Interchange File Format TC
Web Content Accessibility Guidelines (WCAG) 2.2 - World Wide Web Consortium (W3C)
TBX Dialects - Introduction to TermBase eXchange (TBX) - TBX-Info (companion documentation to ISO 30042:2019)
CLDR 36 Release Note - Unicode Consortium
Test your app with pseudolocales - Google (Android Developers)
XLIFF Version 2.1 - OASIS XML Localisation Interchange File Format TC
CLDR Development Process - Unicode Consortium (CLDR)
PROV-O: The PROV Ontology - World Wide Web Consortium (W3C)
RFC 3339: Date and Time on the Internet: Timestamps - Internet Engineering Task Force (IETF)
Universal Electronic Records Management (ERM) Requirements - U.S. National Archives and Records Administration (NARA)
ISO 15924 Registration Authority - Unicode Consortium (ISO 15924 Registration Authority)
ISO 639-3 Registration Authority - SIL Global
Language Tag Extensions Registry - Internet Assigned Numbers Authority (IANA)
Language Tags and Locale Identifiers for the World Wide Web - World Wide Web Consortium (W3C) Internationalization Working Group
ISO 30042:2019 Management of terminology resources — TermBase eXchange (TBX) - International Organization for Standardization (ISO)

Open questions

Evaluate whether the locale profile release and stewardship charter artifacts should become standalone registry-type Vercy models once real multi-Dimension reuse evidence exists, rather than remaining artifacts nested inside this mixin.
Retrieve ISO 15489-1 and at least one applicable data-protection statute to close the self-declared retention-and-deletion and privacy gaps, replacing the current single-anchor NARA-only retention basis.
Determine whether Vercy already defines or should define a generic cross-cutting role/authority-assignment pattern, and check the six localization-specific roles here for duplication against it.
Confirm against other registered Vercy mixins whether full lifecycle, approval, supersession, disclosure and disposition service machinery is the established pattern for entry_kind mixin, to keep cross-registry consistency.
Research a cross-engine machine-translation confidence calibration reference and a translation-supply-chain security threat model, both explicitly flagged as gaps with no primary source located in this pass.
The ISO 639 and ISO 3166-1 code sets were not consulted directly because they are not openly published; they are reached only through the registry records that quote them, so no claim is made about clauses of those standards.
CLDR supplemental script metadata, including per-script right-to-left flags and sample characters, was not retrieved. Direction guidance rests on the layout orientation elements and the HTML direction attribute instead.
CLDR language matching distance data and its weights are referenced but not enumerated, so no matching quality thresholds are specified here.
The full -u and -t key and type vocabularies change with each locale data release and are referenced rather than listed; the keys named in questions are illustrative, not exhaustive.
Sign languages, constructed languages, historical orthographies and transliteration schemes are handled only through registered variant subtags and the transformed-content extension; no consulted source justified additional structure.
Text segmentation, character encoding, Unicode normalization form, font selection and shaping were excluded because no consulted primary source places them inside a locale declaration.
Terminology and glossary assets are bound by reference only; the internal structure of a term base was not researched and is left to the referenced terminology model.
Segmentation rules themselves are not modelled; text boundary determination and segmentation-rule exchange were not consulted, so unit-to-node correspondence assumes a segmentation the owning system already performed.
Translation memory leverage and fuzzy-match scoring algorithms are excluded; only the resulting binding and its recorded origin are held.
Grammatical features beyond plural and select, such as grammatical gender and formality dimensions available in locale data, are not given a first-class finding.
The normative text of ISO 30042:2019 for TBX and of ISO 5060:2024 for translation evaluation is paywalled and was not read; terminology structure is taken from first-party companion documentation and quality vocabulary from ITS instead.
Certified, sworn and legally attested translation, and the identity of the attesting authority, are not modelled.
Sign-language variants, Easy-Read and plain-language renditions are treated only as bound alternatives, not as first-class language variants with their own category systems.
Character encoding, font availability, text shaping and line-breaking are excluded, so a variant can pass every check here and still fail to display.
Cost, vendor commercial terms and service-level obligations attached to translation work are excluded.
No records-management standard text was obtained as a primary source; ISO 15489-1 and the ISO 30042 full text were not retrievable, so retention semantics rest on a national archives authority's requirements plus the ISO 30042 catalogue record only.
No data-protection legislation was retrieved live in this pass; subject-rights erasure, restriction of processing and storage-limitation obligations are referenced by delegation rather than modelled.
Sign-language, oral-only and unwritten-locale governance is under-specified: script subtag policy and the notion of a canonical written form do not transfer cleanly to these cases.
Machine-translation confidence calibration has no cross-engine primary basis; the model requires a declared scale but cannot make scores comparable.
Bidirectional text, script-shaping and font-availability constraints that materially affect whether a localized value is usable are excluded as rendering concerns and are not represented even as a declared limitation field.
Pluralization rules, gender agreement and message-format placeholder semantics are locale-data implementation concerns and are not bound here, although they routinely break naive translation bindings.

Machine files

Card as JSON Specification AGENTS.md Source Card backlog vercy.card/1

Provenance

world-models research · reviewable-draft

Built from: models/wm-xct-031-localization-language/spec.yaml, ver-cy/world-models/card-supplements/wm-xct-031-localization-language.json