DCAT-AP-UK: DCAT Application Profile for the UK
Help Site map

Context

Requirements stream

New This page is new since 1.2.0-rc.1 (2026-11-06). Everything that changed

Over the last year the same need has come up in many places: catalogue owners, departments putting the Data Asset Management Policy into practice, national discovery services, defence and security users, policing and geospatial teams. They are not asking for the same standard. They are asking for the same architecture: a common DCAT core, governed specialist layers, clear access and handling semantics, machine-readable validation, provenance that survives harvesting, and mappings that say what they lose.

How this is evidenced. Requirements come from cross-government engagement and from public sources. Engagement records are not published: people, meetings and documents are withheld, so each requirement says only what kind of community raised it. Public sources are linked and dated. A requirement is met only when a property, rule or test shows it.

Evidence: E Raised in cross-government engagement. The record is not published.; V Checked against a public primary source on the date given.; O A rule this profile sets for itself.

At a glance

12

Met A property, rule or test shows it.

12

Partly met Some of it is shown; the gap is stated.

3

Planned Not yet built. It is in the work programme.

2

Controlled Needs controlled material. Handled outside this public repository; no claim is made here.

Requirements by theme and status
ThemeMetPartly metPlannedControlledAll
Interoperability between standards14117
Security and access semantics314
Conformance and validation325
Provenance and mapping213
Adoption across environments224
Governance and openness1326

A family of profiles, on one core

From the foundation up, L0 to L4: each level builds on the ones before it, shares one validator, and exists only with its own tests. Layers marked + sit alongside the spine. Reuse before inventing; profiles, not forks; discovery stays separate from access.

  1. L0

    Foundation standards Met

    DCAT 3, DCTERMS, SKOS, PROV-O, DQV and the other W3C and OGC vocabularies every layer builds on.

    See it

  2. L1

    DCAT-AP-UK core Met

    The common spine for UK government catalogues, aligned with DCAT-AP 3.0.1. Required of every catalogue.

    See it · 15 requirements

  3. L2

    Geospatial (GeoDCAT-UK) Partly met

    Location, coordinate reference systems, resolution and lineage, reusing GeoDCAT-AP and mapped to UK GEMINI. Required for geospatial datasets, series and services.

    See it · 2 requirements

  4. L3

    Geodesy module Partly met

    A candidate module for reference frames, epochs, observation techniques and uncertainty, without loading them onto the core.

    See it · 1 requirement

  5. L4

    Controlled partner overlays Controlled

    Defence, security and allied exchange: stricter validation, handling and control semantics, applied only in partner exchanges. Specified, governed and tested outside this public repository.

    Why this layer is not here · 2 requirements

  6. +

    Data asset management layer Partly met

    What critical data asset reporting needs: ownership, criticality and provenance.

    See it · 3 requirements

  7. +

    Other sector extensions Met

    Health, statistics, transport, energy, high-value datasets, AI-ready and data spaces, each a module of its own.

    See it · 1 requirement

  8. +

    Platform and standard mappings Partly met

    Crosswalks and the weaver for catalogue platforms and other standards, each saying what it loses.

    See it · 5 requirements

From evidence to release

Every requirement should be traceable along this chain. The counts are taken from the profile and its tests at build time.

  1. Evidence and user need5 public sources, plus engagement
  2. Requirement29 requirements
  3. Property or pattern20 properties traced
  4. Shape or schema rule14 class shapes, 8 rules
  5. Positive and negative tests10 examples, 36 deliberate breaks
  6. Validation evidencethree validators, one verdict
  7. Decision or issue21 decisions
  8. Release status1.2.0-rc.1

Release gates

A release is ready when all four gates are met (RQ-26).

Release gates and where they stand
GateStatusEvidence
Reuse assessed for every new termPartly metEach property records whether it is reused, changed or new against the xGov MEM and DCAT; not yet a blocking check.
Every rule traced to a testMettests/test_profile.py fails the build if a rule has no deliberate break.
No open serious semantic-loss or disclosure defectPartly metThe defect register is complete and C-22 blocks protective markings; semantic-loss defects are not yet graded by severity.
Compatibility claimed only with evidenceMetConformance levels come from the kit; no allied or controlled compatibility is claimed (RQ-15, RQ-29).

Design principles

  1. A common core, with governed layers for specialist needs. No single profile carries everything.
  2. Neutral about technology and supplier. Platforms connect through mappings, never by becoming the standard.
  3. Being able to find a record never implies being allowed to have the data.
  4. The record and the data can be handled differently. Each can say how it must be handled.
  5. Provenance survives harvesting and mapping: who said what, and where it came from.
  6. Every rule is machine-readable, and three validators give one verdict.
  7. A mapping says what it loses.
  8. Existing catalogues connect through mappings and adapters, not by being replaced.
  9. Public artefacts never carry controlled semantics. Controlled layers are specified and tested elsewhere.
  10. Built in the open wherever security allows.

The requirements

29 requirements in 6 themes. Each says who raised it, by kind of community; how it is evidenced; what would show it met; and where it is traced. Select one in the lineage explorer to follow it to properties and tests.

The requirements catalogue
IDRequirementRaised byEvidenceStatusTraced to
RQ-01

Interoperability between standards · Platform and standard mappings

Records can be exchanged with international and allied DCAT-based profiles without bespoke translation for each pair.

Every bespoke translation is a cost and a place where meaning is lost.

Gap: Crosswalks cover CKAN, Dublin Core, UK GEMINI, HDR UK, Croissant and DDI, and the weaver also reads and writes DCAT-US and schema.org. Allied profiles that are not public are out of scope here.

Acceptance and validation
Acceptance
For each listed profile, a crosswalk exists and a round trip through the weaver keeps every mandatory value.
Validation
tests/test_integrations.py Weaver round trip
Public sources
Priority
Must
International and allied partners, Defence and security users, Catalogue owners and platform teamsEPartly met
RQ-02

Security and access semantics · DCAT-AP-UK core

Finding a record never grants access to the data. Who may have the data is stated separately from who may see the record.

Discovery across government only works if publishing a record does not expose the data.

Acceptance and validation
Acceptance
Datasets and data services carry access rights; catalogue records carry their own audience.
Validation
SHACL and JSON Schema shapes for accessRights and metadataAudience
Priority
Must
Defence and security users, Departments implementing data asset management, National discovery and data library servicesEMet
RQ-03

Security and access semantics · DCAT-AP-UK core

A catalogue record can say how the record itself must be handled, separately from the handling of the data it describes. New in 1.2.0-rc.1

A record about protected data is often itself OFFICIAL, and a shared catalogue needs to know that.

Acceptance and validation
Acceptance
Catalogue records take a security classification of their own, from the same list as datasets.
Validation
Generated shapes for CatalogueRecord; class page diagram
Priority
Must
Defence and security users, Catalogue owners and platform teamsEMet
RQ-04

Conformance and validation · DCAT-AP-UK core

A record can say why a property is empty (unknown, withheld, not applicable, not collected or not yet available), so a strict harvester need not guess. New in 1.2.0-rc.1

Stricter validators treat an absent value as an error. An explicit reason lets them tell a withheld value from a forgotten one.

Acceptance and validation
Acceptance
Missing values are structured, use a controlled list of reasons, and never satisfy a mandatory property (rule Q-02).
Validation
tests/test_toolkit.py MissingValues; tests/fixtures/invalid.yaml missing-value-bad-reason and missing-value-no-reason
Public sources
Priority
Should
Defence and security users, National discovery and data library services, Geospatial publishersEMet
RQ-05

Conformance and validation · DCAT-AP-UK core

Validation is machine-readable and gives one verdict wherever it runs.

A record that passes in one tool and fails in another cannot be trusted by either.

Acceptance and validation
Acceptance
pySHACL, JSON Schema and validator.js agree on every example and every deliberate break.
Validation
tests/test_profile.py every example passes all three; every break fails all three
Priority
Must
Catalogue owners and platform teams, National discovery and data library services, Defence and security usersE OMet
RQ-06

Conformance and validation · DCAT-AP-UK core

Conformance is claimed by level and by layer, and only after the matching tests pass.

A single yes or no hides what a catalogue can actually exchange.

Gap: Levels exist for the core only. Layer claims (data asset management, sectors, controlled overlays) are planned.

Acceptance and validation
Acceptance
The conformance kit gives levels 1 to 3 for the core. Each further layer gets its own test records before anyone may claim it.
Validation
downloads/conformance-kit.json expected verdicts
Priority
Must
Catalogue owners and platform teams, Defence and security users, Departments implementing data asset managementEPartly met
RQ-07

Interoperability between standards · Other sector extensions

Specialist terms live in their own namespaces, as modules that can be required or left out as a whole.

Modules let a community be strict about its own terms without making everyone else carry them.

Gap: Extensions reuse their sectors' namespaces. New UK-only terms still go in the one dcatuk namespace; a namespace per module is not yet decided.

Acceptance and validation
Acceptance
Each extension is a module with its own mandatory and optional properties, using its sector's vocabulary, merged by one function.
Validation
Profile.with_extensions; extension tests in check_browser (break extension)
Priority
Should
Defence and security users, Geospatial publishers, Catalogue owners and platform teamsEPartly met
RQ-08

Provenance and mapping · DCAT-AP-UK core

The authoritative source of a record, and of each value changed on the way, survives harvesting and mapping.

Enriched or merged records that look like the original mislead people about who is accountable.

Acceptance and validation
Acceptance
Records carry publisher, creator and qualified attribution; harvested copies keep the original identifier, point to the record they came from (source record) and list the properties they enriched.
Validation
Generated shapes; weaver notes; worked example record-harvested
Priority
Must
National discovery and data library services, Catalogue owners and platform teams, Departments implementing data asset managementEMet
RQ-09

Provenance and mapping · Platform and standard mappings

Every mapping says what it loses.

Silent loss is worse than visible loss.

Gap: Notes are written by hand for each reader and writer, so a new mapping could drop a value without a note. A generic loss report is planned.

Acceptance and validation
Acceptance
The weaver returns notes on values it changes or drops, and the crosswalk matrix marks partial and missing matches.
Validation
tests/test_integrations.py Weaver
Priority
Must
Catalogue owners and platform teams, International and allied partnersE OPartly met
RQ-10

Adoption across environments · Data asset management layer

Asset, dataset, data product, data service and distribution have one meaning each, which policy and catalogues share.

Departments reporting under the same policy count different things when the words differ.

Gap: Data product has no class of its own and no agreed mapping yet.

Acceptance and validation
Acceptance
Each term has a definition, a class or a stated mapping to one, and an entry in the glossary.
Validation
Glossary and class pages
Priority
Must
Departments implementing data asset management, Catalogue owners and platform teamsEPartly met
RQ-11

Adoption across environments · DCAT-AP-UK core

Organisations without an enterprise catalogue can still describe their assets, from a spreadsheet, and be found.

Discovery should not wait for every department to buy a catalogue.

Acceptance and validation
Acceptance
The bulk checker reads a spreadsheet (CSV) and gives a verdict for every row.
Validation
tools/check_browser.py bulk (break bulk)
Priority
Should
Departments implementing data asset management, National discovery and data library servicesEMet
RQ-12

Adoption across environments · Platform and standard mappings

The profile works with the catalogue platforms departments already run, through configuration or adapters.

Replacing platforms to adopt a standard is the lock-in the standard is meant to prevent.

Gap: Product entries are background knowledge until each supplier confirms them (open register O-19).

Acceptance and validation
Acceptance
Each named platform has an integration pattern and a test route through the conformance kit.
Validation
integrations page; conformance kit
Priority
Must
Catalogue owners and platform teams, Departments implementing data asset managementEPartly met
RQ-13

Adoption across environments · DCAT-AP-UK core

Validation works in disconnected and air-gapped environments, with no calls to the network. New in 1.2.0-rc.1

Many protected environments cannot reach the internet.

Acceptance and validation
Acceptance
The validator, its profile file and the command line run offline; validator.js makes no network calls.
Validation
tests/test_toolkit.py Offline
Priority
Must
Defence and security usersEMet
RQ-14

Security and access semantics · DCAT-AP-UK core

The public profile describes OFFICIAL information only. Anything needing higher classifications is handled in controlled layers.

An OFFICIAL catalogue must not become a list of protected holdings.

Acceptance and validation
Acceptance
The security classification guidance says records about SECRET or TOP SECRET data do not belong in an OFFICIAL catalogue; the portal carries no protective markings.
Validation
tools/check_site.py C-22
Priority
Must
Defence and security users, Departments implementing data asset managementE OMet
RQ-15

Interoperability between standards · Controlled partner overlays

Where allied exchange needs it, DCAT-AP-UK maps to the NATO Core Metadata Specification and its confidentiality labelling.

Defence data shared with allies is described in NATO's own core metadata.

Acceptance and validation
Acceptance
A mapping reviewed against authorised sources. No conformance claim is made until then.
Validation
Outside this repository
Public sources
Priority
Could
Defence and security users, International and allied partnersE VControlled
RQ-16

Interoperability between standards · Geospatial (GeoDCAT-UK)

Geospatial records reference GEMINI rather than repeat it.

Two sources of truth for the same geospatial terms drift apart.

Gap: Review of the crosswalk by geospatial specialists is planned.

Acceptance and validation
Acceptance
The geospatial extension is aligned to GeoDCAT-AP and UK GEMINI, with a crosswalk to GEMINI, and defines nothing they already define.
Validation
Extension tests
Priority
Should
Geospatial publishers, Catalogue owners and platform teamsEPartly met
RQ-17

Governance and openness · Data asset management layer

Catalogues and information asset registers work together without becoming two registers of the same things.

Duplicate registers disagree, and people stop trusting both.

Acceptance and validation
Acceptance
A stated mapping between register fields and DCAT-AP-UK properties, tested on a real register extract.
Validation
Planned
Priority
Should
Departments implementing data asset management, Catalogue owners and platform teamsEPlanned
RQ-18

Governance and openness · DCAT-AP-UK core

Public artefacts never expose controlled semantics, and never name people or meetings from internal engagement. New in 1.2.0-rc.1

Openness is the default, but aggregation can reveal what single facts do not.

Acceptance and validation
Acceptance
A check fails the build if requirements or pages carry protective markings or controlled-specification terms.
Validation
tools/check_site.py C-22 (deliberate break registered)
Priority
Must
Defence and security usersOMet
RQ-19

Governance and openness · Data asset management layer

Missing metadata is collected from the business areas that hold it, without a central team re-keying it.

Departments often do not hold all the metadata centrally.

Gap: The bulk checker exists; the commissioning guidance does not.

Acceptance and validation
Acceptance
Guidance and a bulk route for business areas to supply and verify their own entries.
Validation
Planned
Priority
Should
Departments implementing data asset managementEPartly met
RQ-20

Interoperability between standards · DCAT-AP-UK core

Policing and justice systems can describe shared data with the same core, alongside their own interoperability layers.

Interoperability in policing is organised in layers that a common metadata core can sit inside.

Acceptance and validation
Acceptance
A worked mapping from the core to the semantic layer of policing interoperability guidance.
Validation
Planned
Public sources
Priority
Could
Policing and justiceVPlanned
RQ-21

Interoperability between standards · Platform and standard mappings

Metadata held in a catalogue platform's own model, such as the Apache Atlas types behind Microsoft Purview, converts to DCAT-AP-UK through a mapping layer, not by re-keying. New in 1.2.0-rc.1

Departments adopting a platform should not have to choose between its tooling and cross-government discovery.

Acceptance and validation
Acceptance
The weaver reads Atlas v2 entities, turns epoch times into dates, keeps GUIDs only as identifiers, and names every DCAT-AP-UK property the entity cannot supply.
Validation
tests/test_integrations.py AtlasReader; weaver sample
Priority
Must
Catalogue owners and platform teams, Departments implementing data asset management, National discovery and data library servicesEMet
RQ-22

Provenance and mapping · Platform and standard mappings

A mapping never invents a value. No default description, licence, language, contact name or classification; a missing value stays missing and is reported. New in 1.2.0-rc.1

A plausible default is worse than a gap, because nobody goes back to check it. Defaulting an unlabelled asset to OFFICIAL is a security decision nobody made.

Acceptance and validation
Acceptance
No reader in the weaver sets a value the source does not hold; each gap becomes a note.
Validation
tests/test_integrations.py AtlasReader nothing is invented
Priority
Must
Catalogue owners and platform teams, Defence and security users, National discovery and data library servicesE OMet
RQ-23

Interoperability between standards · Geodesy module

Precise positions say which reference frame and epoch they are in, how they were observed and how uncertain they are, at a stated confidence. New in 1.2.0-rc.1

The same point has different coordinates in different frames and at different dates; mixing them silently is an error of centimetres to metres.

Gap: A candidate for review by geodetic specialists. Rules linking uncertainty to confidence, and frame to epoch, are not yet enforced.

Acceptance and validation
Acceptance
A geodesy module with frame, epoch, vertical reference, technique and uncertainty with confidence, and deliberate breaks.
Validation
Extension breaks geodesy-without-frame and geodesy-bad-technique
Priority
Should
Geodesy and positioning specialists, Geospatial publishers, Defence and security usersEPartly met
RQ-24

Conformance and validation · Geospatial (GeoDCAT-UK)

Geospatial records state their coordinate reference system, give resolution with a unit, and say how the data is represented; geospatial services state the OGC standards they conform to.

These are the facts a geospatial user checks first, and the ones generic discovery metadata leaves out.

Acceptance and validation
Acceptance
Shapes in the geospatial extension for each of the four, with deliberate breaks.
Validation
Extension breaks geo-without-crs and geo-bad-representation; a data service without conforms to fails under the extension
Priority
Should
Geospatial publishers, Catalogue owners and platform teamsEMet
RQ-25

Conformance and validation · DCAT-AP-UK core

A record declares which profile, layers and versions it was written to.

A harvester cannot apply the right rules, or migrate a record, if it does not know which ones it was made against.

Gap: The property exists, but there is no published identifier per layer version yet, and no rule checks that the identifier is a known one.

Acceptance and validation
Acceptance
Catalogue records give the profile and each layer as versioned identifiers in conforms to.
Validation
Generated shapes for CatalogueRecord; worked example record-harvested
Priority
Should
Catalogue owners and platform teams, National discovery and data library services, Defence and security usersEPartly met
RQ-26

Governance and openness · DCAT-AP-UK core

A release passes four gates. Reuse is assessed for every new term; every rule is traced to a test; no serious semantic-loss or disclosure defect is open; compatibility is claimed only with evidence.

Gates make "ready" a fact anyone can check rather than an opinion.

Gap: Traceable tests and the defect register are enforced; reuse assessment is recorded per property but not as a gate, and semantic-loss defects are not graded by severity.

Acceptance and validation
Acceptance
Each gate has a check in the release workflow or a recorded decision, shown on this page.
Validation
Release gates table below
Priority
Must
Catalogue owners and platform teams, Defence and security users, Geospatial publishersE OPartly met
RQ-27

Governance and openness · DCAT-AP-UK core

UK profiles agree which properties are mandatory, recommended and optional, so a record that conforms to one is not rejected by another.

Profiles built for statistics and for cross-government catalogues drifted apart on obligations, so the same record passed one and failed the other.

Acceptance and validation
Acceptance
A published comparison of obligations with each UK profile, and a decision on every difference.
Validation
Planned
Priority
Must
Official statistics producers, Catalogue owners and platform teamsEPlanned
RQ-28

Governance and openness · DCAT-AP-UK core

Every requirement can be followed from its evidence to the properties, rules, tests, decisions and release that meet it. New in 1.2.0-rc.1

A requirement that cannot be traced cannot be shown to be met, or safely changed.

Gap: Decisions and releases are not yet linked to individual requirements in the lineage.

Acceptance and validation
Acceptance
The lineage explorer links each requirement to what traces it; the chain below counts each link.
Validation
tests/test_integrations.py RequirementsStream every requirement is in the lineage
Priority
Must
Catalogue owners and platform teams, Defence and security users, Departments implementing data asset managementOPartly met
RQ-29

Security and access semantics · Controlled partner overlays

Partner overlays attach to the core without changing it. Core discovery stays open; overlay terms apply only in partner exchanges.

Loading partner controls onto the civil core would make it harder to adopt and leak semantics that should stay controlled.

Acceptance and validation
Acceptance
A controlled overlay specified, tested and governed outside this repository, using only the extension points the core publishes.
Validation
Outside this repository
Priority
Must
Defence and security users, International and allied partnersEControlled

Requirements (CSV) Requirements (JSON) Blank template (CSV)

Public sources

Public sources behind the requirements
SourceWhat it saysCheckedRequirements
Using metadata to describe data assets in a data catalogue (Data Standards Authority)DCAT version 3 is endorsed for describing datasets and data services, for example when connecting departmental catalogues to the central Government Data Catalogue, and can be used with other open standards such as DQV.VRQ-01
NATO Core Metadata Specification, ADatP-5636 Edition A Version 1A NATO standard, accepted in 2022 and listed in the NATO Interoperability Standards and Profiles, for core metadata used to find and share data; published detail is limited.VRQ-15
Data Strategy for Defence (Ministry of Defence, 2021)Sets Defence data rules and the aim of curated data flowing between users and platforms, nationally and internationally.VRQ-15
Interoperability Handbook (Police Digital Service, 2026)Guidance on interoperability across policing and law enforcement systems, in layers from organisational to foundational.VRQ-20
NilReasonType vocabulary (GML 3.2, as published by Eionet for INSPIRE reporting)Standard reasons for a void value, including unknown, withheld, inapplicable, unpopulated and template.VRQ-04

Controlled layers

Some requirements can only be specified or tested with material that is not public: allied specifications, control-set models and test records for protected environments. This repository holds none of it. It holds the open core and the extension points those layers attach to, and it lists such requirements with status "Controlled" so the gap stays visible. Their detail belongs in a controlled repository, governed by the right authorities.

What the open core gives those layers to attach to:

Proposed work programme

This project's proposal. Nothing here is commissioned or endorsed.

  1. Confirm mandate and governance Planned · Weeks 1 to 2
    Terms of reference, decision rights and an escalation route, including how controlled requirements are governed.
  2. Agree the design principles Done · Weeks 2 to 3
    The ten principles above, recorded as a decision.
  3. Consolidate the requirements Done · Weeks 3 to 4
    This catalogue, version-controlled, with evidence and acceptance criteria.
  4. Run requirements workshops by community Planned · Weeks 4 to 8
    Validated requirements for the core, data asset management, controlled domains, geospatial, platforms and governance, each in its own session.
  5. Produce the minimum viable profile Started · Month 3
    A candidate release with scope, conformance statement, vocabularies, examples and migration guidance from the xGov MEM.
  6. Build validation alongside Started · From month 1
    Shapes, positive and negative test records, mapping tests and loss reports, written with the profile, not after it.
  7. Pilot with real catalogues Planned · Months 3 to 5
    Harvesting, mapping, provenance and validation measured on a representative group of catalogues, including one controlled-domain case under appropriate controls.
  8. Settle the repository and publication model Planned · Before the candidate release
    An organisation-owned public repository for the open core, and controlled repositories for restricted mappings and test records.

The case against

This is a list of good intentions, not evidence.

Every requirement says how it is evidenced and what would show it met. Engagement evidence is labelled E and not overstated: it shows a recurring need across communities, not that any one body has asked for this profile.

Withholding who raised what makes it impossible to check.

True from outside. The engagement record exists for the owning team and the governance route; publishing it would aggregate information about protected communities, which the project avoids (ADR 0015).

A profile family will fragment into many profiles nobody can track.

Layers share one core and one validator, and a layer exists only with its own tests. The family is smaller than the set of bespoke translations it replaces.

Security semantics do not belong in a civil profile at all.

Only the minimum is in the core: handling of the record kept apart from handling of the data, and withheld values. Anything stricter is a controlled layer.

Search the portal

Keyboard shortcuts

/ or Ctrl K
Search
?
These shortcuts
Esc
Close menus, tooltips and dialogs
Arrow keys
Move between tabs and search results