Model-Based Testing Community | Connecting the MBT world…

Jul/13

1

Domain-Specific MBT

Hi folks,

we are all aware of the basic idea of MBT: there is some kind of formal specification of the system or the environment and this specification is used by a program to automatically compute a set of test cases. There are several influencing factors and different approaches to the test generation technique like, e.g., using model checkers, constraint solvers, or the like.

A major issue of these aspects, however, is that they are focused on a very technical level of this model-based testing. It seems like one has to be an expert in formal modeling and constraint solving theory to create models and use them for automatic test generation. Many users, however, are neither. As a consequence, modeling the input has to move closer to the domain of the users.

There are several approaches like modeling
* in a programming language that the users are already familiar with, and using an MBT tool on top of these models (e.g. C# for Spec Explorer‘s model programs or Java for Conformiq Designer‘s action language);
* with the help of a DSL and/or BDD on top of a standard general purpose programming language,
* in a simple DSL that is based on data-driven testing or keyword-driven testing.

We invite you to discuss these approaches and also name further if you happen to know some.

Cheers
Stephan

No tags

2 comments

  • Stephan Schulz · July 5, 2013 at 7:18 am

    Thank you for an excellent topic. Actually I have been pointing out for several years (see Dagstuhl 2010) that the MBT community has been overlooking the “T” in “MBT” in its tool chains.
    Fact is that especially in the later phases of testing (integration/functional, system, end2end)people working in testing generally have no or little background in programming (or even plainly refuse to work with programming). On the other had these testers have 5 to sometimes all the way up to 20 years (!) of experience in testing these type of systems – an incredible source of knowledge that we loose when we can not reach these testers.
    Another issue with general purpose programming approaches is that models can for a large part only be reviewed by developers. Domain experts mainly review generated testing artifacts.
    Another trend you can not overlook is that _conventional_ testing is tools for automtic test execution have been moving away for years now from programmatic approaches and towards spreadsheet or fully UI based test specification. Even standardization is now moving in this direction (see the uptake and intensity of work on tehir new Test Description Language (TDL)).
    All of the above is why Conformiq has reacted and is has since May 2013 productized a new modeling frontend called “Conformiq Creator” that allows such domain experts to create models without any need for programmatic specification – and at the same time draw the full benefits of running on top of (the same) test generation engine that enables fully automated test design.
    Finally I not resist to invite everybody interested in this topic to a joint tutorial (not a tool demo BTW) on this topic with DSM guru Juha-Pekka Tolvanen on Domain-Specific MBT at ETSI’s UCAAT 2013 Conference in October in Paris http://ucaat.etsi.org/2013/program.html .

    Reply

  • Holger Funke · July 16, 2013 at 8:06 pm

    Hi guys,

    thanks you, Stephan, for your thoughts on Domain-Specific MBT. In a research project on behalf of BSI we are currently setting up a way to model smart card protocols as used on German ID card. We are focussing on protocols in this project and use our models to generate test cases on the one hand and source code for a simulation on the other hand.

    Your are invited to find more infos in my blog post concerning this project:
    http://blog.protocolbench.org/2013/07/next-generation-of-epassport-testing

    Cheers,
    Holger

    Reply

Leave a Reply

<<

>>

Theme Design by devolux.nh2.me
© by MBT Community 2011