An identifier is a typed value that names a resource unambiguously within a stated scheme, such as a registry key, a media type or a file extension.
Format registries model an identifier as a pair of type and value so that one format can carry many identifiers from different schemes.
PNG has at least five identifiers in common use: the media type image/png, the extension .png, the Apple uniform type identifier public.png, a PRONOM PUID for each version, and a Library of Congress format description ID. None of them is wrong, none is complete, and each serves a different system. That is why format registries treat the identifier as a structured object with a type and a value, and not as a single text field.
Common identifier types
Registry keys. PRONOM assigns PUIDs in the form fmt/<number> (and x-fmt/<number> for an older series). The Library of Congress Format Descriptions use IDs beginning fdd, and Wikidata gives formats Q-numbers. These keys are opaque, stable and maintained by one authority.
Internet media types. Registered with IANA, written type/subtype (application/pdf, audio/flac). Broad and widely understood, but often too coarse: application/pdf covers every PDF version and profile. See dcterms:IMT for how vocabularies type these values.
Platform type codes. Apple’s UTIs (public.jpeg, com.adobe.pdf) and the older Mac OS type codes, which were FourCC values like TEXT and PDF .
Extensions. Easy to read, easy to break. They are identifiers only by convention, and a registry records several per format.
Codec and internal codes. FourCCs, WAVE format tags such as 0x0001 for PCM, and similar values identify encodings inside containers.
Modeling identifiers in OWL
In an OWL ontology, every individual already has an IRI, and the IRI is its identity in the strict sense. The identifiers people care about in format work are literals from other schemes, so ontologies take one of two routes.
The simple route is a data property per scheme: ex:mediaType "image/png". It is easy to query but needs a new property each time a scheme appears. The structured route defines an Identifier class whose instances carry an identifier type and a value, as in the example above. It handles any number of schemes and lets the type itself be a controlled term with its own definition. PREMIS follows the same structured pattern for its own identifiers.
OWL also offers ways to say an identifier is unique. owl:hasKey declares that individuals with the same key value are the same individual, and an inverse functional property says the same for a single property. Both support merging records from different sources, and both should be used carefully: two registries that disagree about which format a media type belongs to can, through such axioms, force a reasoner to conclude that two distinct formats are one.
Identifiers versus names
An identifier is not a label. “Tagged Image File Format”, “TIFF” and “TIF” are names; the first is the formal name and the others are alias names. Names are for people and may change or collide. Identifiers are for machines and should not.
In PREMIS, the same split appears in two places. An object’s own identity is an objectIdentifier with objectIdentifierType and objectIdentifierValue. Its format is pointed to with formatRegistry, which carries formatRegistryName, formatRegistryKey and formatRegistryRole, while formatDesignation holds the human-readable name and version. A repository that stores only the name will struggle to match its holdings against registry updates; one that stores the key can follow a format record as it changes. The file format identification guide shows where each identifier type enters an identification workflow, dc:identifier covers the general metadata element, and the vocabulary index lists the other registry terms.
Example in RDF
@prefix ex: <https://example.org/vocab#> .
<https://example.org/formats/png-1.2>
ex:name "Portable Network Graphics" ;
ex:identifier [ ex:identifierType ex:MediaType ; ex:identifierValue "image/png" ] ,
[ ex:identifierType ex:AppleUTI ; ex:identifierValue "public.png" ] ,
[ ex:identifierType ex:Extension ; ex:identifierValue "png" ] .
Questions about Identifier
What is a format identifier?
A value that names a file format within a defined scheme. Examples include a registry key such as a PRONOM PUID, an Internet media type like image/png, an Apple uniform type identifier like public.png, or a file extension.
Why model an identifier as a class and not a plain property?
Because the value only makes sense with its scheme. 'png' as an extension and 'image/png' as a media type are different kinds of identifier. A class with a type and a value keeps them apart and lets new schemes be added without new properties.
How does OWL handle identifiers?
In OWL every resource is already named by an IRI. Literal identifiers are modeled as data properties or as identifier objects; owl:hasKey and inverse functional properties can declare that a value picks out exactly one individual.
Is a file extension a reliable identifier?
It is the weakest one. Extensions are easy to change, shared between unrelated formats (.dat, .bin) and missing on many files. Registries record them because tools and people use them, but signature-based identification is more trustworthy.
How does PREMIS record a format identifier?
PREMIS uses formatRegistry with formatRegistryName (for example PRONOM) and formatRegistryKey (the PUID), plus a formatRegistryRole that says how the registry entry relates to the object. Object identifiers follow the same type-and-value pattern with objectIdentifierType and objectIdentifierValue, so every identifier stays tied to its scheme.