The file format, physical medium, or dimensions of the resource.
Recommended practice is to use a controlled vocabulary where available. For example, for file formats one could use the list of Internet Media Types [MIME]. Examples of dimensions include size and duration.
Every Dublin Core record that describes a file sooner or later has to say what kind of file it is, and dc:format is where that answer goes. The element covers three different things at once: the file format, the physical medium, and the dimensions of the resource. Simple Dublin Core does not separate them, so a single record can legitimately carry image/tiff, 35 mm slide and 4000 x 3000 pixels as three values of the same element.
Media types as the default vocabulary
The DCMI usage comment points to the list of Internet Media Types, the registry maintained by IANA and formerly known as MIME types. Values take the form type/subtype: application/pdf, image/tiff, audio/flac, text/csv. In qualified records the scheme can be declared explicitly with dcterms:IMT, so a parser knows the string is a media type and not free text.
Media types have parameters, and they are sometimes useful in a format statement. text/plain; charset=UTF-8 says something that text/plain alone does not, and character encoding is exactly the kind of detail that decides whether a text file is readable in thirty years.
Where media types run out
A media type is a broad label. application/pdf covers every version of PDF from 1.0 onward and every profile built on it, including PDF/A, PDF/X and PDF/UA. image/tiff says nothing about compression, bit depth or whether the file is a GeoTIFF. For access and display that rarely matters. For preservation planning it matters a great deal, since migration and validation decisions depend on the exact version.
Archives handle the gap in a few ways:
- adding a second
dc:formatvalue with a human-readable version (“PDF/A-2b”) - recording a format registry identifier, such as a PRONOM PUID, in a technical metadata schema like PREMIS
- describing the format itself as a resource and linking to it, which is what the dcterms:FileFormat class is for in linked-data settings
Identification tools such as DROID, Siegfried and the file utility can supply both the media type and the registry identifier automatically; the format identification guide covers how they work and where they disagree.
Dimensions and physical carriers
The second and third parts of the definition tend to be forgotten in digital work. Dimensions include size and duration: page counts, pixel dimensions, file size in bytes, running time. Physical medium covers the carrier for non-digital or hybrid collections: a reel of 16 mm film, a 3.5-inch floppy disk, a vinyl record.
DCMI Metadata Terms splits these apart. dcterms:extent takes size and duration, and dcterms:medium takes the material or physical carrier. Both are sub-properties of dcterms:format, which in turn refines dc:format. A system that only reads simple Dublin Core sees them all collapsed back into format statements, which is the intended fallback.
Common errors in harvested records
Several patterns show up repeatedly when records from different repositories are merged:
- Format used as type. Values like “Text” or “Image” in
dc:formatbelong in dc:type. - Extensions instead of media types. “.docx” or “JPG” are readable but not controlled, and they split one format into several facet values.
- Invented media types. Strings like
application/wordorimage/jpglook plausible yet appear in no IANA registration; the registered forms areapplication/mswordandimage/jpeg. - Everything in one string. “PDF, 12 pages, 2.3 MB” defeats any attempt to filter by format. One value per statement keeps each part usable.
The fourth is the cheapest to prevent. A cataloging form with separate fields for media type, extent and carrier produces clean repeated values without anyone needing to remember the rule.
Example in RDF
<rdf:Description rdf:about="https://example.org/oral-history/interview-07.flac">
<dc:title>Interview with a retired ferry pilot</dc:title>
<dc:type>Sound</dc:type>
<dc:format>audio/flac</dc:format>
<dc:format>48 minutes</dc:format>
</rdf:Description>
Questions about dc:format
What goes in the dc format element?
dc:format holds the file format, physical medium or dimensions of a resource. For digital files the usual value is an Internet media type such as application/pdf or image/tiff; for physical items it can name the carrier or give size and duration.
Is a file extension a valid dc:format value?
It is allowed but weak. Extensions such as .doc or .dat are ambiguous and easy to change. A registered media type is the recommended value, and a format registry identifier can be added where the record needs more precision.
What is the difference between dc:format and dc:type?
dc:type describes the nature or genre of the resource, such as Text, StillImage or Dataset. dc:format describes how it is encoded or carried. A scanned letter is of type Text even when its format is image/jpeg.
Does dc:format record the version of a file format?
Only indirectly. A media type like application/pdf covers every PDF version and profile. Preservation systems that need to tell PDF 1.4 from PDF/A-2b record a format registry identifier or a technical metadata field alongside it.
How does dcterms:format differ from dc:format?
dcterms:format has the same definition but declares a range, MediaTypeOrExtent, and has two refinements: dcterms:extent for size or duration and dcterms:medium for the physical carrier. It is a sub-property of dc:format.