If you invent a better software development paradigm, nobody will beat a pathway to your door. You may have to spend years tuning it to show that it works. For example, this paper describes a ten-year project to see how model-driven engineering (MDE) is adopted in practice. Modeling shifts programming to a more abstract level than even the highest-level languages by using diagrams. This paper has some useful nuggets. If your organization is thinking about changing its tools or methods, it may interest you. For instance, the author shows that layered architectures work--the layers are similar to the dominant model-view-controller (MVC) pattern. Another insight: software always changes after it is released. Indeed, technology and methodology also evolve. So there is a need for round-trip engineering, but this was not provided in this project. Instead, Kulkarni developed complex forms of meta-modeling to support changes and product lines.

The paper supports the maxim “One size doesn’t fit all.” It provides both heavyweight and lightweight tool sets, for example. Further, “a configurable extensible modeling language engineering workbench seems required for greater adoption of model-driven software development.” But could a more agile and iterative process work with simpler tools and languages to give better value faster?

The paper is not well written and omits details of process, organization, and scenarios. However, it provides some useful pointers for those adopting or inventing new methods.