Model MARC control numbers as identifiers in metadata schema

[barmintor]

This was originally brought over from the Palace slack and opened as issue #65, but needs discussion:

One of the obstacles to broader integration between Palace and our other systems is the difficulty in pulling the identifiers that our other workflows are based on through the circulation manager's ingested OPDS feeds to be published in an OPDS catalog. In particular, ISBNs and MARC organization-scoped control numbers (001+003, 010, 016, 035) without relying on them being encoded in the single URI identifier for a publication.

The use of schema.org properties provides a way to represent ISBN/ISSN as metadata/identifier property values in OPDS, and there's a merged PR for this in the parser Palace uses for the publication manifests (though the data is not yet operated on by the CM, to my knowledge). The organization-scoped control number are more complicated.

schema:identifier technically has a range of 3 data types: text (supported), URI (supported), and PropertyValue (no support). PropertyValue is an object value that could be used to model the MARC control numbers. We could pursue this by trying to get schema.org to have an identifier sub-property for bibframe:Local (which is how those MARC fields are modeled in BIBFRAME, with a value and associated assigner property pointing to the MARC org as a URI), or we could simply document a practice for including these control numbers as PropertyValue (with a nested assigner property and, ideally, a mechanism for indicating cancelled values.

[barmintor]

schema.org examples for identifier include control numbers as PropertyValue:

{
 "@context": "https://schema.org/",
 "@type": "Book",
 "name": "Library linked data in the cloud : OCLC's experiments with new models of resource description",
 "author": "Carol Jean Godby",
 "isbn": "9781627052191",
 "identifier": {
 "@type": "PropertyValue",
   "propertyID": "OCoLC",
   "value":  "889647468"
  },
}

I think we might make a stronger recommendation to clarify what the propertyID value might mean in such an example.

[barmintor]

Perhaps:

{
  "metadata": {
    "@type": "http://schema.org/EBook",
    "title": "Library linked data in the cloud : OCLC's experiments with new models of resource description",
    "author": "Carol Jean Godby",
    "schema:isbn": "9781627052191",
    "schema:identifier": {
      "@type": "schema:PropertyValue",
      "schema:additionalType": "http://id.loc.gov/ontologies/bibframe/Local",
      "schema:propertyId": "http://id.loc.gov/ontologies/bibframe/identifiedBy",
      "schema:value":  "889647468",
      "schema:valueReference": [
        {
          "@type": "schema:DefinedTerm",
          "identifier": "http://id.loc.gov/ontologies/bibframe/assigner",
          "schema:inDefinedTermSet": "http://id.loc.gov/vocabulary/organizations",
          "schema:termCode": "ocolc"
        },
        {
          "@type": "schema:PropertyValue",
          "identifier": "http://purl.org/dc/terms/replaces",
          "schema:value": "replacedNumberFromSubfield-z",
        },
      ],
    },
  }
}

001: identifier as text; bf:Local with valueReference from 003, else 040$a, else none
010: 010$a as bf:Local with dlc valueReference; 010$b as bf:Local with dlcnucm valueReference
016: 016$a value with blank first indicator as bf:Local with lac valueReference; 7 first indicator with valueReference from 016$2
035: bf:Local with marcorg valueReference; remainder of $a in value
all: cancelled values in subfield $z represented in additional valueReference

[barmintor]

This is very verbose - it could be made significantly more compact by adding an identifier type for MARC21 control numbers to the schema, and the assigner values are all representable by a URI - but I think it captures the necessary data. @HadrienGardeur I am curious what you think.