Any object in the EKG is identified with at least one universally unique, opaque, permanent, non-reassignable and web-resolvable identifier in the form of an International Resource Identifier (IRI) for the EKG---i.e. an Enterprise Knowledge Graph IRI (EKG/IRI) aka EKG Identifier.
As said, the EKG/IRI identifier is permanent, will be proliferated across the enterprise's universe---including its ecosystem---and will be used for the expression of facts about the object including relationships between objects.
Additional non-EKG identifiers may also be assigned, and they may be human-readable, "external" to the enterprise's EKG and be transient and reassignable.
Resolving an identifier can be done in three ways:
- using it in a transaction---i.e. a query or update statement---submitted or routed via an internet protocol to a "lookup service" that translates one or more given "features" of an object to an EKG/IRI.
- constructing it via a standardized policy from key components and applying a hash and optionally signing it---where the object represented by the EKG/IRI may or may not already exist.
- constructing it by giving the object an EKG/IRI based on a random number in case the EKG is the authoritative source for the given object.
While the semantic web technologies---like for instance RDF---generally allow for many and varied IRIs, and this is still encouraged when integrating systems, there is benefit in being able to rely on one canonical and unchanging one, which can for example make the mapping of identities a many-to-one rather than a many-to-many task.
In addition to that, to enhance the likelihood that various EKGs, or smaller Knowledge Graphs within an EKG---across departments, organizations or ecosystems---can interoperate easily with each other, the use of standardized EKG/IRIs needs to be encouraged since various EKGs can come to the same identifiers independently, drastically increasing the number of links across EKGs and therefore enhancing potential interoperability.
Since EKG Identifiers will proliferate across the enterprise and its ecosystem and be used and stored for many years to come (could be decades) it is essential that the identifiers are always resolvable so that the need to make copies of data for long-term storage purposes decreases drastically.
- There should be a mapping or service to resolve other identifiers such as names, keys or other IRIs to the EKG/IRI.
- Since it is immutable, the EKG/IRI will have to be opaque i.e. not be a human-readable since even human names, company names, customer numbers, Social Security Numbers can change over time.
- Objects that already have a well-established RDF compliant and Linked Data compliant identifier may not necessarily need an additional EKG/IRI. In fact, they may already have one that is external to the company's EKG. It is in many cases recommended to even give those well-established objects, from well-established external datasets, an EKG/IRI of your own enterprise as well for various reasons. Examples of such objects are Legal Entity Identifiers (LEIs) and Financial Instrument Global Identifiers (FIGIs). The EKGF will maintain a list of these for convenience.
- The use of multiple EKG identifiers---often in combination with multiple "legacy identifiers"---generally means that an EKG should not use the Unique Name Assumption (UNA) (where the use of a different identifier would imply a different object).