G-Rec 3 Crosswalks (collections of mappings), individual mappings, and their versions must be identified with a GUPRI

[linikujp]

Each individual mapping contains two objects/entities/terms and the mapping relationships and a direction between them. Additionally, associated metadata describing this mapping provides context of the mapping. A globally unique, persistent and resolvable identifier (GUPRI) must be provided to this individual mapping along with its associated metadata. Usually, mappings were provided as a set, or a collection of individual mappings. In this case, not only a GUPRI should be provided for the set or collections of mappings, but also each individual mapping component formed the set or collection should be assigned to a specific GUPRI. Similar to data or other semantic artefact, mappings will evolve, becoming obsolete, or being discarded. Therefore, the GUPRI must direct to different versions of mappings for use. The benefit of identifying individual mappings is to enable the composition or combination of collections or crosswalks leveraging full or partial mappings from other existing collections.

[graybeal]

The phrase "the GUPRI must direct to different versions of mappings for use" is ambiguous here.

I agree that GUPRIs should exist for each version of a mapping set, and ideally for each version of a single mapping.

But it is not the job of the metadata specification to specify that if you visit a GUPRI you should be able to learn about all the other versions of that GUPRI. Yes that's a great spec for the repository, but it requires that all the other metadata elements to reconstruct that knowledge be present in the mapping(set) metadata.

See also #64, #66, #71 for related points.

[matentzn]

Fully agree. Note right now SSSOM does not enforce resolvable version iris - it is up to the user to decide how exactly to identify the mappings set.

I personally have used mapping_set_id to identify the series (like what ontology iri does) and mapping_set_version is simply a textual string.

For resolvable mapping_set_versions we need to do some work on the standard.