Why Use NIEM with JSON?

JSON data often appears as a simple, flat list of (key, value) pairs, as in the example on the left. JSON data can also be complex, with nested objects and arrays of objects or values, as in the example on the right.

{                                {
  "lname": "Wimsey",               "name": {
  "fname": "Peter",                  "lname": "Wimsey",
  "date": "1893-05-04",              "gname": [ "Peter","Death","Bredon" ],
  "lat": 51.835,                   },
  "lon": -0.417,                   "date": "1893-05-04,
  "cdate": "1921-01-05",           "lat": 51.835,
  "charge": "Wild Driving"         "lon": -0.417,
}                                  "cdate": "1921-01-05",
                                   "charge": "Wild Driving"

Regardless of whether the JSON data is simple or complex, the data scientists and software developers who work with it need to understand what it means. This involves answers to questions like:

  • What is the meaning of the date key string? Birthday or something else?

  • What is the format of the date value? Is it always an ISO 8601 date field, or might we sometimes see Tuesday, or Fred? Is it the same format as cdate, or something different?

  • Does lname have the same meaning in both resources?

  • Will gname always be found within a name object, or is it optional?

Such questions are easily answered when the JSON data is only exchanged between instances of the same program, or only processed by software written by a single small team of developers, or is one of a few sources accessed by the data scientists. In these cases there is little difficulty in understanding the JSON data, and little need for NIEM.

However, understanding the data resource becomes much more difficult when the developer team is large, or when there are many independent teams, or when the members of a team change over time – in short, when developer cohesion is low. Also, data understanding takes more work when scientists and developers work with many data resources (all at once, or over time) that have different data element definitions for the same shared concept (fname, gname, givenname) – in short, when data resource overlap is high. These conditions – low cohesion, high overlap – are almost always found in large enterprises with many participants or extensive data sharing needs.

NIEM is especially useful in those circumstances. The NIEM approach applied to JSON data offers:

  • A mapping between each JSON key string and a global identifier for the corresponding data element.

  • A framework for machine-readable documentation of each data element that is used in a data resource.

  • A standards-based convention for canonical data element names used as JSON key strings (while still permitting local names like lat within a data resource).

These three combine to make NIEM JSON data self-documenting, at least to the extent that any data scientist or software developer can look at any key-value pair and find out how it is defined. In addition, the NIEM approach also offers:

  • A simple way for developers who are creating a resource to reuse data elements defined by others, instead of creating new definitions. This encourages a degree of harmonization between different resources.

  • A NIEM data model, partitioned into a NIEM core of concepts that might be reused in any exchange, plus several submodels extending the core, each defined by a subject-area or functional NIEM domain community. This also encourages harmonization among resources partly defined in terms of that model.

  • A governance approach in which the NIEM core is controlled by the entire NIEM community, while each submodel is independently controlled by its domain, and which allows any community to create and control its own extension of the NIEM model.

  • A version-control architecture that encourages reuse by insulating each community from external changes; this allows each community to change at its own pace, deciding if and when it will adopt changes made by others.

  • A framework for data resource specifications that contain all the artifacts required to document the required and optional elements that may appear in a conforming resource, together with the meaning of those elements.

These features of NIEM combine to form a standards-based approach for constructing any number of data resource specifications, which document the data resources found at runtime and allow a compatible understanding of that data among the scientists and developers who work with it.

Although NIEM was first created to work with XML-based data exchange, it can now be applied to JSON data, without requiring knowledge of the XML technology originally used. Data exchange designers can use NIEM to create a resource specification that is all-JSON, or one that is all-XML, or one that defines both a JSON and an XML serialization, depending on the needs of the producers and consumers.