Standardize "identifier" usage

[jbrown-xentity]

Name: James Brown

Affiliation: GSA/data.gov

Type of issue: Various schema identifier usage

Issue: identifier is utilized in various classes and objects. There is also an Identifier class. Sometimes the implementations use the class, sometimes a simple string is allowed. We should document when/where one is used over another.

Recommended change(s):
Propose to utilize the Identifier class in all instances and usage of identifiers. This schema should allow for the more robust current definition, and allow for simple strings of any type. This will allow for consistent usage and allow for both complex and simple identifier usage implementation.

[jbrown-xentity]

@zopalmer14 I envision this by allowing the Identifier class to not just be an object with all the associated detail, but a literal string. I believe this would involve changing the type from object to an anyOf using the current structure, but also adding another option for type string. https://github.com/GSA/dcat-us/blob/main/jsonschema/definitions/Identifier.json#L6

[zopalmer14]

If I understand correctly, you're saying it should be something like:

"anyOf": [
        {
          "type": "object"
        },
        {
          "type": "string"
        }
]

This would eliminate the need to allow anyOf a string or an Identifier object in the identifier and otherIdentifier properties, which could instead simply accept Identifier objects as the flexibility is built into the class. Am I following your logic correctly?

Question, instead of the above structure do you think the following is valid?

"type": ["object", "string"]

[jbrown-xentity]

I'm thinking more along the line of the following. This means we can just say that Dataset has an identifier field that implements the identifier class, and that field can either be a simple string or the more complex original definition of identifier, allowing specificity of complexity and how the identifier was generated/defined.

This way we don't have the Dataset identifier field be a simple string and then the otherIdentifiers be a list of identifier class, and the usage super inconsistent. It allows for every identifier to be the same class, and that class can either have the actual identifier with a bunch of clarifying items in an object or just have the simple identifier. This does put more work on data.gov to check the JSON structure, but I think it's very manageable; especially for key fields.

Identifier.json

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://resources.data.gov/dcat-us/3.0.0/definitions/identifier",
  "title": "Identifier",
  "description": "Information about an identifier scheme",
  "anyOf": [
  {
    "type": "string",
  },
  {
  "type": "object",
  "properties": {
    "@id": {
      "type": "string",
      "format": "iri"
    },
    "@type": {
      "type": "string",
      "default": "Identifier"
    },
    "schemaAgency": {
      "title": "schema agency",
      "description": "The name of the agency that issued the identifier",
      "type": ["null", "string"]
    },...
  }]
}

[zopalmer14]

Thx for the additional comment. In hindsight, my JSON structure doesn't make any sense as the "properties" object needs to be grouped with the type object. I actually think we're on the same page I just screwed up the JSON and didn't explain my thinking very well.