A related resource from which the described resource is derived.
The described resource may be derived from the related resource in whole or in part.
In plain terms dc:source answers one question: what was this made from? A scan of a printed atlas plate has the atlas as its source. A clip cut from a broadcast recording has the broadcast. A table published in a report and extracted as a CSV file has the report. The definition calls it a related resource from which the described resource is derived, and the comment adds that the derivation can be whole or partial.
Writing it in RDF
Simple Dublin Core in RDF has two serialization habits that both appear in the wild. The older pattern, from DCMI’s 2002 note on expressing simple Dublin Core in RDF/XML, writes every element as a literal: the value sits between the tags as a string. The newer linked-data pattern uses rdf:resource to point at a URI.
<dc:source>urn:isbn:9780000000002</dc:source>
<dc:source rdf:resource="https://example.org/catalog/atlas-1856"/>
The fifteen elements in the http://purl.org/dc/elements/1.1/ namespace have no formal range, so both lines are valid RDF and valid Dublin Core. They are not equivalent, though. The first produces a string; an RDF processor sees text that happens to look like a URN. The second produces a link to a node that can carry its own description, title, date and identifiers. For any project that publishes linked data, the second form is the useful one, and dcterms:source is the property most profiles choose for it.
OAI-PMH is the main exception. Its required oai_dc format is an XML schema, not RDF, and every element there is a plain string. Records that hold a URI internally simply print it as text in that export, which is harmless as long as the string is the full, resolvable URI and not a local shorthand.
Derivation and other relationships
Several neighboring properties look similar, and choosing well makes records easier to query.
| Situation | Better property |
|---|---|
| The resource was made from another one | dc:source |
| Same content, different file format | dcterms:isFormatOf |
| A later or earlier edition | dcterms:isVersionOf |
| Part of a larger whole that still exists | dcterms:isPartOf |
| Any other connection | dc:relation |
The second row is the frequent overlap. A PDF exported from a word processing file is arguably derived from it, and it is also the same content in another format. dcterms:isFormatOf states the second fact more precisely and is the choice in format-centered collections. dc:source suits cases where the content changed in the process: selection, excerpting, transcription, reformatting from analog.
Identifiers over citations
DCMI recommends identifying the source with a string from a formal identification system. An ISBN, a DOI, a call number with an institution prefix or a URI all qualify. The sentence “Taken from the county atlas of 1856, held in the map room” is helpful to a reader and useless to a harvester trying to connect the scan with its original.
A common compromise keeps both: the identifier in dc:source, and a human-readable citation in a note or in dcterms:bibliographicCitation on the source’s own record. The identifier conventions are the same as for dc:identifier, which helps: a source value should match an identifier that exists somewhere as a real record.
Source in digitization records
For digitized collections, dc:source carries more than bibliographic interest. It ties the surrogate to the original whose physical condition, rights status and description may all differ. Copyright status in particular follows the source: a scan of a public domain book inherits that status in most jurisdictions, and a rights statement on the scan should reflect the original. When the original is later re-cataloged or re-photographed, every scan that points to it by a stable identifier picks up the connection without edits.
Example in RDF
<?xml version="1.0" encoding="UTF-8"?>
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:dc="http://purl.org/dc/elements/1.1/">
<rdf:Description rdf:about="https://example.org/scans/atlas-1856-plate-12.tif">
<dc:title>Plate 12, county atlas</dc:title>
<dc:source>urn:isbn:9780000000002</dc:source>
<dc:source rdf:resource="https://example.org/catalog/atlas-1856"/>
</rdf:Description>
</rdf:RDF>
Questions about dc:source
What does dc:source mean?
dc:source points to a related resource from which the described one is derived, in whole or in part. Typical cases are a scan made from a printed book, an excerpt taken from a longer recording or a dataset extracted from a larger database.
How is dc:source written in RDF?
In RDF/XML it can be a literal element value, such as an ISBN string, or an rdf:resource attribute pointing to a URI. Because the dc elements have no declared range, both forms are valid, but the URI form gives linked-data tools something to follow.
What is the difference between dc:source and dc:relation?
dc:relation states any relationship. dc:source states a specific one: the described resource was derived from the other. In DCMI Metadata Terms, dcterms:source is defined as a sub-property of dcterms:relation, which reflects that narrowing.
Should dc:source hold a citation or an identifier?
The DCMI comment recommends an identifier from a formal system, such as an ISBN, DOI or URI. A full citation is readable but hard to match automatically; where one is needed for display, a note field can hold it.
Is dc:source the place for provenance history?
No. Chains of custody, donors and previous owners are provenance and go in dcterms:provenance or an archival description. dc:source names the resource the content came from, not the people who held it.