gs1 enforces strings vs things dichotomy

[VladimirAlexiev]

Looking at https://www.gs1.org/voc/PostalAddress:

• gs1:addressCountry is a resource
• gs1:addressLocality, gs1:countyCode and gs1:addressRegion are strings

This decision looks pretty random

• Datasets like Wikidata and Geonames have comprehensive gazetteers that include all levels, so after reconciliation I may want to link an address to resources at all levels (city, county, region)
• On the other hand, if I have a source with pretty dirty field "country", I may want to save it as string (as given) before reconciliation

Other organization or encyclopedic ontologies allow string or thing in such situations, but gs1 enforces a random dichotomy:

• almost every schema.org property allows string or thing
• The euBusinessGraph ontology (which uses w3 org, regorg, schema, locn) handles this by introducing parallel props for string or thing at the various geography levels (whcih are not 4 but 7 because also include EU NUTS)

[mgh128]

The GS1 random dichotomy you refer to is contained within the class gs1:PostalAddress Some of its properties expect string (e.g. for postal code). Many address fields expect rdf:langString gs1:addressCountry expects a gs1:Country class because ISO 3166 defines code values for countries as well as subdivisions within countries, so we're just reflecting that capability.

[VladimirAlexiev]

There's also a dichotomy between string vs langString (#22), and the assignment also appears random:

• https://www.gs1.org/voc/addressRegion: rdf:langString (i.e. free text): but "for example NJ" shows this can be a code?
• https://www.gs1.org/voc/countyCode: xsd:string (i.e. code): if the lower level is codified, why wouldn't the higher level be codified?

Rest of comments in new issue.