How do you translate a LIDO 1.1 museum collection record without changing the history, provenance or identity of the object it describes? Lightweight Information Describing Objects is an XML schema for sharing information about art, cultural heritage, natural science specimens and other collection objects with online services and aggregators. Its multilingual fields can help visitors discover a work in their own language, but the record also carries identifiers, object types, events, participants, dates, places, materials, rights and relationships whose meaning must not drift during localization.
People searching for LIDO 1.1 translation, multilingual museum metadata, LIDO XML localization, object-title translation, cultural-heritage record interoperability or event-based provenance need to distinguish display language from indexed concepts and documentary evidence. A painting title is not the same as an accession number. A named maker is not interchangeable with a collector. A production date differs from a museum acquisition date. Translating every visible string as plain prose can create a fluent catalogue that gives the public a false account of where an object came from or who made it.
This guide explains a careful international approach to LIDO 1.1 localization, including the descriptive and administrative parts of the schema, event records, actors, controlled concepts, rights, multimedia resources and validation. It is written for museums, archives, galleries, heritage portals, researchers, curators and language professionals. It complements the other eduKateSG translation guides on library and archival metadata: LIDO’s distinctive challenge is describing objects through well-defined events and relationships while offering meaningful language variants to different audiences.
How museum records preserve object identity across languages
1. LIDO is an interchange language for object descriptions
LIDO 1.1 is designed for publishing and exchanging museum and heritage metadata. It is not intended to replace every function of a collections management system, such as loan operations or conservation workflows. A record can describe a painting, historical instrument, archaeological artifact, scientific specimen or other object through structured information suitable for discovery portals. Translators should understand the export role before editing anything. The localized record should let a reader understand the same object and its relationships, not transform the source database into a new curatorial narrative. The schema helps with interoperability, but integrity still depends on decisions made by people.
2. Record the version and application profile
The ICOM-CIDOC LIDO Working Group published LIDO 1.1 in December 2021. Implementations may use local profiles, vocabulary rules or portal requirements in addition to the base XSD. Always note the actual schema version and any receiving aggregator’s constraints. A record that is valid under the general schema can still be rejected by a service requiring particular identifiers or controlled term references. Translation should follow those published requirements without inventing missing facts. Do not assume that because a collection database exports XML, every field in the export is free to change or every portal will interpret it identically.
3. Object identity survives changes of language
A museum object can have an accession number, persistent record identifier, collection identifier and other references. These are not display titles. Translate explanatory labels around them, but protect the identifiers and their exact relationships. Two objects may share a common translated title such as Blue Vase while having different dates, makers and locations. They must remain separate records. Conversely, a single object may have several names in different languages without becoming several physical objects. Identity control is the foundation on which multilingual discovery, citations and provenance research rely.
4. Distinguish the object from its metadata record
A collection object exists independently from the database record describing it. The record has a creator, modification history, source and administrative status of its own. LIDO can distinguish object information from record-level data. A translated cataloguer’s statement should not be presented as an inscription written on the object itself. Likewise, a record revision date is not an object creation date. Translate record notes with the same attribution and time scope as the source, keeping the documentary boundary visible. This helps scholars assess whether a statement comes from the artifact, a catalogue entry or a later correction.
5. Descriptive and administrative metadata differ
Descriptive metadata explains what the object is, how it may be classified and what events or subjects relate to it. Administrative metadata helps manage rights, record provenance and digital representations. A translator must handle both layers according to their roles. A public object description might be rephrased for natural reading, but a rights restriction cannot be softened just because the sentence is awkward. A digital resource identifier must not change because its accompanying image caption is translated. Classifying fields by function before extraction prevents the most damaging forms of accidental editing.
6. Language attributes must be consistent
LIDO supports multilingual descriptive metadata, including language identification using XML language attributes. A target-language record should declare its language correctly and distinguish embedded quotations or names in other languages when the implementation supports it. Simply adding a translation in a new field without correct language labeling can make search facets and display selection unreliable. Do not assume that a user who chooses French wants every Latin inscription transliterated or rewritten. Test the portal’s language selection and fallback behavior, especially when some fields have a translation but others retain their approved source-language forms.
7. The object or work type is a controlled concept
An object might be categorized as a painting, drawing, textile, instrument, fossil or another type. These categories can have precise definitions and identifiers. A localizer should use an approved target-language label for the same concept rather than substitute a more familiar but inaccurate term. A print is not automatically a painting because both are artworks. A historical scientific apparatus is not simply a decorative object because a casual visitor might call it that. Preserve controlled concept IDs and vocabulary references, then verify that the target display text still identifies the same object type.
8. Titles often have several authorities
A work may have a title assigned by the artist, an inscription, a cataloguer’s descriptive title, a historical title and a translated exhibition label. These should not be collapsed into a single supposedly original target-language title. Preserve the source title and its attribution where documented, and identify which language variants are translations or descriptive inventions. For untitled works, do not turn a supplied cataloguing phrase into an author’s title. A sensible multilingual catalogue can show a familiar title for discovery while allowing researchers to inspect the exact historical or institutional wording.
9. Names are linked to roles and evidence
A name in a museum record can identify an artist, workshop, manufacturer, donor, collector, previous owner, publisher, photographer or restorer. Translation should retain each role and its degree of certainty. If a vase is attributed to a workshop, the target language must not claim an individual master made it personally. Match names against recognized authority records where available, distinguish translated or transliterated display forms and preserve underlying identifiers. A fluent translated label that assigns the right person the wrong role creates a serious historical error.
10. Events organize an object’s history
LIDO’s event-oriented approach can represent the making, use, discovery, ownership, exhibition and other episodes in an object’s life. Each event may have participants, places, dates, methods and evidence. Translators should first identify the event type, then translate descriptions without moving participants or dates between events. An object made in one city and later excavated in another cannot be described as made at the excavation site. An exhibition date is not a production date. The event structure prevents these confusions when localizers respect its relationships instead of flattening the record into a single story.
11. Creation and production are not always simple
A painting may have a named artist, but a machine, tapestry or historical instrument can involve designers, workshops, manufacturers and later modifications. The production event must retain the same attribution structure as the evidence. A translated account can explain the roles, but should not assign sole authorship when the source identifies multiple participants or uncertainty. Preserve production technique, place and approximate time. For a composite object, a curator may distinguish events affecting components. Translation should not erase those distinctions for the sake of a shorter public label.
12. Discovery and excavation must stay distinct
Archaeological and natural-history objects often have collection or discovery events that are different from their origin. A mineral found in a quarry may have a geological formation history far older than its discovery date. An artifact recovered during excavation may have been made elsewhere centuries earlier. Localized metadata should distinguish found at, made at, collected at and held at. A literal place-name translation can mask the role of the place unless the event type remains visible. These distinctions matter for research, cultural claims and the interpretation of historical movement.
13. Acquisition is not creation
A museum may acquire an object through purchase, gift, bequest, exchange or another route. The date of accession describes an institution’s relationship to the object, not when the object originated. Translate acquisition wording with care and avoid reducing all transactions to bought if the source records a donation. This matters especially in public provenance displays where the type and date of acquisition may be contentious. Record any uncertainty about transaction history rather than invent a clear chain. An accessible summary should help readers understand the institution’s evidence, not conceal gaps.
14. Ownership and custody are different
An object may have a legal owner, a current custodian, a lender, a deposit holder and a location where it is displayed. These are distinct roles. A loan exhibition does not automatically transfer ownership. Translating held by as owned by can misrepresent rights and institutional relationships. Preserve role identifiers, dates and the relevant event descriptions. Some records intentionally withhold current location for security or cultural reasons, so a target-language description should respect the source publication policy. In heritage metadata, a single preposition can change the meaning of a complicated institutional relationship.
15. Provenance is a series of claims
Provenance is often described as a chain of ownership, custody, transfer and documentary evidence. Gaps and qualifications are important. A translated description should preserve dates, named parties, transaction types and the difference between documented and inferred links. If the record says possibly in a private collection, the new language must not make it a confirmed ownership statement. Do not fill chronological gaps with an imagined journey merely to create a smooth narrative. A responsible multilingual catalogue is readable without concealing uncertainty, especially for objects connected with colonial history, conflict or disputed property.
16. Attributions and uncertainty
Museums frequently use phrases such as attributed to, workshop of, circle of, formerly attributed to and after. These are not decorative modifiers. They can distinguish a maker’s direct hand from a follower or a later copy. Translate them according to accepted art-historical and disciplinary terminology in the target language. A slight change may affect market interpretation or scholarly attribution. Record the source wording, concept if controlled, and basis of the statement. Native-language fluency is valuable, but preserving the exact strength of an attribution is essential for trust.
17. Event dates can be approximate
Creation and use dates may be centuries, ranges or scholarly estimates rather than exact calendar days. A translator should not turn late eighteenth century into 1789 or circa 1900 into a definite date. Preserve machine-readable date values, earliest and latest boundaries, and explanatory wording. Distinguish a date attached to an event from a date attached to the metadata record. Local date formats may change for readability, but chronology and certainty must remain stable. A catalogue that erases approximation is misleading even if the numbers look more precise.
18. Places have names and historical contexts
Objects can cross linguistic and political boundaries. A location might have a historical name, contemporary official name, transliterated form and locally used name. Use a consistent policy and preserve authoritative place identifiers where available. Do not silently replace an historical place statement with a modern administrative unit that did not exist at the time. When a place is part of provenance evidence, distinguish the source’s wording from an explanatory current map label. This allows readers to understand both where an event occurred and how the repository has interpreted the location.
19. Physical materials are evidence
An object may be described as bronze, ceramic, silk, stone, wood, photographic paper or a mixture of materials. These are not interchangeable natural-language synonyms. In scientific collections, a misidentified material can change research interpretation. Use reviewed terminology and controlled concepts where supported. Translate free-text descriptions carefully while keeping technical substance. Distinguish material from technique: oil on canvas names a medium and support, while cast refers to a process. A shortened caption should not erase the distinction if it changes the reader’s understanding of how the object was produced.
20. Techniques and processes
A record may distinguish carved, cast, woven, printed, painted, glazed, fired, assembled or repaired techniques. Terms vary by craft and historical period. The target-language label should identify the actual process rather than sound superficially familiar. For example, a photographic printing technique is not the same as modern office printing, even if both use a common word. Preserve the concept ID and contextual description. Where experts disagree about a production process, state the uncertainty. Translators should review process terms with subject specialists because an ordinary dictionary often misses the relevant technical distinctions.
21. Measurements and units are protected facts
A museum record may list height, width, depth, diameter, mass and quantities for different parts of an object. Translation can adapt the presentation of units but must preserve numbers, dimensional roles and the measurement basis. A 20-centimetre diameter is not a 20-centimetre radius. If the object includes a frame or pedestal, specify whether dimensions include it according to the source. Avoid locale formatting that could change a decimal value. Test both structured values and reader-facing descriptions. Numerical metadata is especially easy to damage during otherwise fluent copy editing.
22. Subjects and depicted content
A painting of a historical event may have a subject concept, named people, geographical setting and a separate physical object description. These layers should not be merged. A subject label describing a battle does not prove that the object itself was made at the battle site. Translate narratives of depicted content and apply authorized labels for controlled subjects. For culturally sensitive subjects, consult appropriate terminology and community guidance. Keep the distinction between what is visibly depicted and what a curator has interpreted, especially if the identification of a figure or scene is uncertain.
23. Cultural context and contested terminology
Some objects carry names or classifications that have changed because of scholarship or community preference. A modern target-language label may differ from a historically documented catalogue term. Retain a traceable record of the source term and clearly explain any updated display convention rather than silently rewriting historical evidence. Avoid offensive legacy terminology in public-facing prose when an accurate contemporary equivalent is available, while preserving quotations or catalogue history where scholarly access requires them. This is an editorial decision requiring subject knowledge and sometimes community consultation, not a mechanical replacement rule.
24. Inscriptions are primary evidence
Words carved, painted, printed or written on an object belong to the object itself. A translated description should distinguish a transcription of that inscription from a translation of its meaning. Keep original wording, language, direction and uncertainty where recorded, and do not introduce corrected spelling as though it were physically present. A damaged inscription with missing letters should remain marked as uncertain. Public displays can add a readable translation, but it should be clearly presented as interpretation. Researchers may need to compare the original inscription against the localized summary.
25. Related works and linked objects
A museum may link sketches to finished paintings, objects in a set, archaeological fragments, casts and reproductions. The relationship type matters. A replica is not an original; a preliminary drawing is not the completed sculpture. Translate descriptive relationship labels while preserving IDs and links. Test that every target-language relation still points to the right object record. Collections portals sometimes combine records from multiple institutions, increasing the risk that similar translated names lead to accidental merging. Machine identity and relationship semantics must remain stronger than superficial similarities in titles.
26. Resource representations are separate assets
A collection record may link photographs, scans, conservation images, 3D models, videos or audio files. These are representations of the object, not the object itself. Translating the image caption does not authorize altering the source image or changing its rights information. Keep persistent resource references stable and verify that each translated description matches the actual media. A detail photograph may show only a maker’s mark, not the entire object. Good localization tells readers which representation they are viewing and avoids treating the photograph’s creation date as the object’s origin date.
27. Rights statements need precision
Rights can apply differently to the physical object, metadata record, photograph and other digital resources. A public-domain object does not necessarily mean every photograph of it has identical reuse conditions. Translate rights explanations accurately, preserve license identifiers and URLs, and distinguish copyright from access or privacy restrictions. Do not promise unrestricted reuse because a translated caption sounds open. Some cultural collections also have ethical access expectations that are not captured fully by copyright law. Publish clear, source-supported statements without adding permissions that the collection has not granted.
28. Controlled concepts and multilingual terms
LIDO can support concept-based indexing along with display terms. The stable identifier for an object type, material or event category should not be translated. Instead use authorized labels in the target language, preserving the relationship between term and concept. If no approved label exists, document a provisional display form and consult relevant specialists. Treat near-synonyms carefully: bronze and brass, manufacture and excavation, donation and loan are not interchangeable classifications. A robust multilingual portal supports different searches for the same concept without changing the meaning of the underlying dataset.
29. Separate indexing forms from visitor descriptions
A searchable term may be concise and controlled, while an exhibition description can explain context in natural sentences. Translators should not force the same wording into both. A category label such as archaeological object may be appropriate for indexing but unhelpful as the only explanation of a complex artifact. Conversely, a friendly description may be too broad to serve as a controlled concept. Maintain distinct fields and purposes. This keeps search facets consistent while giving visitors enough context to understand what a record means. The record becomes more useful when technical precision and accessible explanation coexist.
30. Entity names and transliteration
Object records contain personal names, institutions, places and cultural names written in different scripts. A multilingual portal may need transliteration, translation of an institutional title or retention of an original name. Apply an authority policy that differentiates these operations. The translated name must still point to the same actor or place identifier. Avoid inventing a conventional English name for an individual who has none. If multiple romanizations are established, store approved variants as search aids without pretending one is a different person. Proper-name consistency reduces confusion when records are harvested by other services.
31. Missing data is not permission to invent
A collection may have unknown maker, uncertain production date or incomplete provenance. Translators may be tempted to fill gaps with plausible contextual statements so the label reads smoothly. That turns translation into unsupported cataloguing. Preserve unknown, not recorded and disputed statuses, because they may mean different things. A missing date is not the same as an object that is known to be undated. A clear target-language description can explain the limitation without speculating. Accurate silence is often more valuable to researchers than a confident but unsupported narrative.
32. Extract only language-bearing fields
A structured extraction should distinguish display text, controlled term labels, authority identifiers, XML keys, numerical values, event references and resource URLs. Provide translators with element paths and record context. A bare phrase such as studio of may describe an attribution, an institution or a physical location depending on where it appears. Protect technical keys and links. Reassembly should restore language text into its intended nodes without changing the surrounding structure. Compare the same source record and target record field by field, not as two flattened paragraphs. Structural meaning is lost when context is stripped away.
33. Plan language fallback intentionally
Not every record has every field translated. A portal may show the target-language title but leave the material term or provenance summary in another language. Decide a clear fallback order and indicate when source-language material remains. Avoid pretending the entire record is translated if key interpretive fields are not. Test mixed-language results with screen readers, search facets and mobile displays. A visitor should be able to tell whether an untranslated phrase is a historical quote, a proper name or a missing translation. Transparent fallback is better than a misleadingly homogeneous interface.
34. Terminology needs institutional review
Museums differ in their preferred labels for materials, techniques, historical periods, object types and attribution formulas. A translator should use a controlled glossary with authoritative references, approved local-language variants and examples from the collection. Include distinctions between professional cataloguing terms and visitor-friendly explanations. Do not simplify a technical label so much that the object changes classification. When a cultural term does not have a satisfactory equivalent, retain the source term with a contextual explanation. Effective multilingual terminology work requires a partnership between language experts and curators, not the assumption that a dictionary is sufficient.
35. Validate with the 1.1 schema
Run XML well-formedness and LIDO 1.1 XSD validation on the output. Also apply local profile checks or Schematron rules when the receiving service requires them. A valid record can still contain wrong event attribution or a misleading translation, so validation is a necessary first stage rather than a final verdict. Compare protected IDs, dates, measurements, concept references, links and structural relationships against the source. Then inspect multilingual fields in their intended portal. Technical validity and museum truth are separate obligations that must both pass.
36. Check event relationships directly
An event-centric metadata record can fail without any broken XML link if language text accidentally assigns a person, date or place to the wrong event. Create an event-level comparison listing event type, actor, role, place, date and descriptive prose in source and target editions. Read each event as a coherent claim. Did the object get created, discovered, collected, acquired or exhibited at this point? Does the translated wording preserve that action? This review catches errors that title-only quality assurance misses, and it makes provenance stories much easier to verify.
37. Test in discovery portals
A museum XML file may be consumed by aggregators that present it differently from the source catalogue. Test target-language titles, facets, dates, images, event summaries and link navigation after ingestion. A portal may choose one display label among several or flatten a complex event into a short snippet. Check that the result remains accurate under those constraints. If an application drops necessary qualifiers, coordinate a display fix or improve the available concise wording without altering the underlying evidence. A data record is successful only when the audience receives the intended meaning.
38. Use culturally responsible review
Museum language can carry social, ethical and political significance, especially with human remains, sacred objects, colonial collections or disputed ownership. Translators should not impose a simplistic conclusion, erase source terminology or state institutional possession as uncontested legal ownership. Work with qualified curators and, where appropriate, communities connected to the object. Clearly distinguish museum records, interpretations and documented claims. A responsible multilingual record makes provenance and uncertainty easier for readers to understand. It does not resolve historical disagreements by choosing a more comfortable target-language phrase.
39. Keep a record of editorial decisions
For each target-language release, record the source snapshot, translated fields, glossary versions, material review decisions, legal or cultural restrictions, validation results and date of approval. When provenance research changes, identify which language variants and public descriptions must be revised. This prevents old certainty levels from persisting in a translated record after the source has been corrected. Maintain links to earlier versions when possible so scholarly citations remain intelligible. Traceability lets an institution improve its catalogue over time without pretending the record has always said the same thing.
40. Publish with transparent status and authority
A multilingual museum record should tell readers what language edition they are viewing, which institution supplied the record and whether labels are official, translated or interpretive. Include source references and restrictions where they help comprehension. Do not suggest that a target-language description is an artist’s original wording unless there is direct evidence. Provide clear paths to the authoritative object record and relevant images. International access is strongest when readers can use familiar language while still inspecting provenance, source claims and institutional responsibility.
41. A complete export and handoff
When handing LIDO records to an international portal, include the exact schema version, application profile, language choices, concept mappings, persistent identifiers, rights data and quality-control results. Test samples from different object categories, not just one familiar artwork. Include records with uncertain attributions, multiple events, mixed scripts, long titles and several images. Provide a process for reporting wrong labels without allowing uncontrolled edits to historic evidence. The objective is that a reader in any supported language reaches the same object and understands the same documented facts, even where wording differs.
Worked LIDO translation scenarios
A donor becomes the artist
A target-language catalogue describes a historical textile as created by the person who donated it to the museum. The name was correctly translated, but the actor’s role was wrong. Check the creation and acquisition events separately, restore the correct participant-to-role relationships and review the wording against the authoritative source. A visitor may remember a fluent label more readily than a complex database field, making the error especially damaging. Event-based translation QA should prove who performed each action, not just that all names are present.
Excavation place becomes manufacture place
An artifact is recovered from an archaeological site far from where it was probably produced. A short translated description says made at the excavation site because the localizer saw the place name in the record. Reconstruct the event chain: the archaeological discovery has a supported location, while the manufacturing place may be unknown. Correct the target-language statement, retain uncertainty and verify the event IDs. This distinction is essential for historical interpretation and can affect how audiences understand trade, migration and technology.
An uncertain attribution becomes definite
A painting is listed as workshop of a well-known artist. The translated title card removes the qualifier to avoid an unfamiliar phrase. The new label now asserts direct authorship that the institution never accepted. Restore the attribution formula using an approved equivalent, link the artist authority correctly and ensure public search snippets retain the qualification. The case shows why attribution language is substantive evidence rather than an optional stylistic detail. A strong translation preserves the curatorial level of certainty even when the target expression is longer.
A title creates two object records
An aggregator receives one artwork under its original title and again under a translated title, then treats them as separate objects because the local export replaced a persistent record identifier. Fix the underlying identity field, retain both titles with language and title-type information, and re-ingest the corrected record. Verify that searches in either language resolve to the same object. Multilingual discovery should increase access paths, not multiply records. The exact source identifier is more reliable than title similarity for establishing identity across catalogues.
An image license is misread
A sculpture is out of copyright, but an associated photograph has separately described reuse conditions. The translated public record incorrectly states that the photograph may be reused without restriction. Compare rights at object, record and resource levels; restore each applicable statement and ensure the license URI points to the correct resource. A convenient simplification can mislead visitors and partners. Translation should communicate the actual permissions rather than infer them from the age of the physical object.
A colonial-era term is mishandled
A cataloguing record includes an outdated historical designation alongside a contemporary preferred name. In translation, one editor removes the historical term entirely; another presents it as the preferred modern classification. Both approaches lose an important distinction. Keep the evidence and its provenance accessible where appropriate, choose responsible public wording with qualified curatorial or community guidance, and explain the relationship between historical and current terminology. The aim is neither to repeat harmful language uncritically nor to rewrite the documentary past without acknowledgement.
Measurement changes during localization
A natural-history specimen is recorded with a length in millimetres. A target-language caption uses the correct number but labels it as centimetres, multiplying the implied size tenfold. The LIDO file remains valid. Compare structured quantity, unit and measured dimension with the visible text, then ask a subject specialist to confirm the result. This is a factual translation failure, not a schema problem. Good multilingual museum metadata makes measurements easier to read without ever converting them by accident.
A portal drops the word ‘possibly’
A record notes that an object was possibly held by a named private collection. The receiving portal displays a shortened summary that omits the qualifier, making a speculative provenance link look proven. Fix the aggregation or display rule where possible and provide a concise target-language phrase that retains the uncertainty. Then review other records with similar qualifiers. This is why localization quality must be tested after harvesting, not solely in the institution’s XML export. Readers judge the history from what the portal actually shows.
A natural science record uses an art label
A shared portal imports a scientific specimen and displays a target-language classification intended for artworks because a broad translated category swallowed two distinct controlled terms. Restore the concept IDs, create clear language labels for each category and test filters across collection types. A museum metadata schema can cover more than paintings and sculpture; its labels must reflect that breadth. The fix belongs in controlled-term mapping and display QA, not in changing the object record to fit a familiar art-focused vocabulary.
Frequently asked questions about multilingual LIDO 1.1 metadata
Is LIDO 1.1 a collections management system?
No. LIDO is a schema for delivering and exchanging descriptive object metadata. Institutions may use other systems to manage acquisitions, loans and internal collection processes.
What distinguishes LIDO from MODS or EAD?
LIDO is particularly suited to object-oriented heritage metadata, including event-based descriptions. MODS focuses on bibliographic description, while EAD describes archival finding aids. They solve different discovery tasks.
Can I translate the accession number?
No. Keep identity-bearing identifiers intact. Translate labels and explanatory content in designated language fields.
How should uncertain artist attribution be handled?
Preserve the exact evidential status, such as attributed to or workshop of, using reviewed target-language terminology. Do not upgrade an uncertain claim.
Should an old place name be replaced with today’s name?
Not automatically. Preserve documentary context and authority identifiers, and distinguish historical wording from a modern explanatory label.
Does valid XML prove a correct museum translation?
No. Schema validation checks form and allowed structure. Event roles, provenance, rights, measurements and uncertainty need independent semantic review.
Can several language titles refer to one object?
Yes. Language variants and alternative titles can improve discovery while pointing to the same stable object identity.
Is a digital image the same as the collection object?
No. The image is a representation with its own resource information and possibly distinct rights or dates.
What if the source catalogue has an unknown maker?
Keep that absence or uncertainty explicit. Translation does not justify guessing a maker or silently completing provenance gaps.