FAIRsFAIR / FAIRSemantics

MIT License
7 stars 1 forks source link

P-Rec. 8: Define human and machine-readable persistency policies for semantic artefacts metadata #8

Open ghost opened 4 years ago

ghost commented 4 years ago

Description

Once published in a semantic repository, semantic artefacts are hopefully being reused by others to build their information system. In the eventuality where the semantic artefact, a concept/term or a relation is deprecated or simply replaced, the repository should offer persistency policies for the metadata (duration of archiving, …). These policies should be both humans and machines readable. Machine readable policies will allow building services which could automatically detect the change to either warn the user or to directly integrate the change whenever it is possible. For humans, repositories could use the classical tombstone page with redirect to the new page when the semantic artefact or the element has been replaced.

 

Existing recommendations: N/A

Stakeholder: Repository

alko-k commented 3 years ago

YES is the short answer NVS uses the owl:deprecated and skos:note properties to publish if a concept is deprecated or not. Most deprecated terms, if possible, will have a dct:replacedBy property that points to the accepted term. e.g. http://vocab.nerc.ac.uk/collection/P02/current/PREX/ The persistency policy for the metadata (duration of archiving, …) is not machine readable though We also have an atom feed here: http://vocab.nerc.ac.uk/nvs_atom.xml that notifies registered users on the changes/updates of NVS

graybeal commented 3 years ago

Existing recommendations: Cool URIs Don't Change (Berners-Lee)

I agree with the text but the title and the text are not aligned (title is about metadata, text is about content)

rob-metalinkage commented 3 years ago

OGC has defined a policy: namespace in absence of a standard, and uses ISO 19135 status terms from a controlled vocabulary:

http://www.opengis.net/def/status http://www.opengis.net/def/status/valid

we could upgrade to a canonical namespace/property - but wouldnt support overloading a generic property like skos:note or a too specific one like owl:deprecated.

rob-metalinkage commented 3 years ago

From OGC perspective we dont have a lot of churn, but can do better with metadata - what we'd like is a published profile for the policy metadata required - and SHACL validation for this.

rob-metalinkage commented 3 years ago

Further to discussions in FSRTG - OGC makes status metadata a responsibility of the service, not the ontology - and the aim is to automatically entail status flags to every object published in the registry base if not present in the ontology resource.