During September 25th to 27th the second Model-based testing (MBT) user conference was held in Tallinn, it was arranged by the tool vendor Elvior and the European standards union ETSI. One of the definitions of model-basted testing states that it is an umbrella term for approaches that generate tests from models.
After attending this conference I would characterize it as a vendor type conference, which was shown in the presentations where the tools were almost always present, but there was also an academic touch to it. There were case studies including the tools of course but also some presentations aimed at pointing out what MBT is and how to use it. The main focus was not really on conferring but everything was nicely set up with possibilities to mingle during breaks. Unfortunately I could only attend 2 out of 3 days and therefore missed the social event on the second day where I believe there might have been more opportunities to confer.
Throughout the conference I listened for anything that could give me a hint regarding how people construct their models, for example what to/not to model, abstraction level, type of model or how to think when doing the modeling. Some recommendations were: to analyze the requirements and model them, model functionality not test cases, do not model the whole application in one model and do not model states but instead what the user is doing.
We learned about UTP, the UML Testing Profile is based upon UML and provides a notation that introduces concepts for test like for example test case specifications, test environments and test data specifications.
The concept of Fuzz testing was explained which includes sending unexpected invalid data into the system using a model-based approach. This is a cheap way to find security and dependability issues. Also listened to a keynote by Robert V. Binder regarding Model-based testing - today and tomorrow. An entertaining story that caught my attention was the description on differences between hardware and software development. The hardware is always getting 1) cheaper, 2) better and 3) faster, in software development one has to choose between two out of the three. Olga Grinchtein shared MBT experiences from Ericsson. When using MBT it could take time to analyze failed tests and there was a risk for test case explosion but introduce changes in a model was on the other hand easy.
The conference was held at the Estonian IT college in Tallinn in a functional building where it was possible to put bags and outdoor clothes in a guarded wardrobe. Outside of the building there was some kind of large piece of art standing on the grass looking like a space capsule or time machine. I have no idea what it’s supposed to be and I haven’t googled it either simply to keep my imagination working on it.
To get the most out of this conference I believe one would preferably be using one or several of the tools that were in focus or be in a situation where these tools are evaluated. In my case I currently use only freeware but this was a great conference that acted as an introduction where I am now more aware of what tools are out there and how they are used. We have also been shared tool-independent experiences regarding MBT.