Our focus on representation issues revealed the underlying beliefs and rationale of different communities in a surprisingly powerful way. We kept our own thrashing to a minimum by side-stepping questions phrased in a de-contextualized, hence unresolvable way (e.g., "What is a domain model?"). The difficulty of that process revealed something of the challenge of domain modeling process itself in our own experience.
It also made clear that domain modeling gets confusing in part because of implicit "ontology shifts" that get made. Domain modeling is hard because we are using models in very different ways from a system modeling sense; yet these models are intermixed with system models and keeping them distinct can be quite difficult. The same term or notational device may be used at one time to denote a specific entity within a specific system; at another time it may represent a "generic" or type entity. We speculated that there is a "contextual shift" (or "ontology shift") that takes place when you move between domain and system models, and that confusion over the semantics of this shift may be the source of many breakdowns in the domain modeling process. Keeping the ontological level straight, and developing the capacity to shift levels intentionally and traceably, seems to be one conceptual skill required in domain modeling that is not inherently part of the system modeler's repertoire.
Viable notations for domain modeling must address this problem in some way. The representation selected can help or hinder this conceptual clarity. The closer the domain modeling representation is to the system modeling representation, the easier is the technology transfer challenge; but the more severe is the danger of confusing domain with system model.
An example might help. One of our scenarios involved the use of a system modeling notation for both domain and system modeling. Those diagrams that were deemed part of the generic architecture, anticipated to be part of multiple systems in the domain, were differentiated with a distinctive border; other diagrams were essentially the diagrams of a single system. The two kinds of diagrams were inter- mixed into a hierarchical structure which, except for the differentiated "generic" diagrams, represented one system's configuration.