CIMI RM Requirements

Back to CIMI WG page

This page is for the collection, review, and agreement on requirements for the CIMI Reference Model. The RM is abstract enough, and the differences among the various proposals subtle enough, that it is helpful to have a set of criteria by which to judge them.

Proposed requirements

1. The model shall align with ISO 13606.
   1. This is firm for upper level classes (compositions, content, clusters). We will not implement item and item_group: see requirement 2.
2. The model shall include classes specific to the domain in order to reduce the overhead for implementers of constructing every property as a class.
3. The distinction between clinically relevant (meaningful, searchable) collections of data and sub-meaningful parts (address, finding site) is the distinction between 13606 Composition & Content (meaningful) and Cluster (sub).
4. The reference model will support all specialization, i.e., addition of properties
   1. The constraint model will support rules and cardinality and range constraints.
   2. Semantics are asserted in the constraint model, as BMM has no facility for doing so.

Open questions

1. We need to find heuristics for when an element should be a property rather than a parameterized element (i.e., supporting observation).
2. In order to support inference of DL expressions, we need a way to identify domain concepts for attributes. There is some question about how to do this; "is-a" may be inappropriate for instances.