Comment on More on Sub-DRDs by bruce
Blog: Method & Style (Bruce Silver)
I must not be explaining it well. When you ask why show the end-to-end logic as a single diagram, I assume you mean you could draw a diagram of, say, AutoEligibility Score, and another one of DriverEligibility Score, etc. I agree you can do that to create a flat model chopped up into little pieces. And maybe you even stub out the other pieces to acknowledge they exist. In BPMN that’s like using Link events as off-page connectors. I personally think that’s bad practice but it is allowed and some people do it.
I know you say that what is in the DRDs doesn’t matter, what matters is the model captured in a repository. OK that is your view. The Method and Style philosophy, which has been very successful in BPMN and which I am now applying to DMN, is that the only thing that counts is what is in the diagrams (DRD and boxed expressions). So if one diagram just omits another part of the decision model, or shows a stubbed out decision (absent its information requirements and supporting decisions), for me it is not properly representing the end-to-end decision logic. I think the end-to-end decision logic should be represented by a single decision model (definitions element) which is rendered visually as a single DRD. So the value of the sub-DRD is that it allows such a DRD to be constructed hierarchically without hiding anything or using unspecified filtering.
I don’t ask that all modelers adopt my methodology. The sub-DRD idea just makes understanding the end-to-end model easier. There is also a methodological benefit similar to what Alan Fish has put forward for BKMs, as a way to identify major logic elements (or “knowledge areas”) used in top-down decomposition. I just think my approach works better graphically, and does not require the extra step of parameter mapping when reuse is not required.