Towards supporting malleable architecture models

Engineers commonly use informal diagrams for sketching, brainstorming, and communicating initial system designs. Diagramming is accessible, new concepts can be added freely, and diagrams can be specifically adjusted to communicate at the exact right level of abstraction depending on the audience. However, the information carried by informal diagrams is most often not precise enough for automation purposes, e.g. analysis. Consequently, there is a risk that these diagrams and the knowledge carried by them are lost in slide decks or buried somewhere in a documentation space. Diagrammatic models, which, unlike informal diagrams, have well-defined syntax and semantics, can be used for automatic analysis of consistency, architectural conformance, change impact, and others. Given the preference for informal diagramming, complete adoption of modeling for architectures seems unlikely in industrial settings that we have studied so far. Therefore, we propose to support engineers to benefit from both these seemingly opposites, as well as to ensure that diagrams convey the right amount of information for the particular task at hand. We envision more malleable models, in which engineers can freely navigate along two orthogonal spectra: (1) horizontal, from informal diagramming to modeling and (2) vertical, from less to more details exposed by a diagram. This paper describes our vision for working with such malleable models, use cases for it, and research needed to make it possible.