Model-Based Testing gives the possibility to use models when evaluating requirements. These can be used as a evaluation option in the analysis phase, this in order to determine whether the requirements are fully covered in terms of functionality. But how you do you evaluate models? Because these models are the basis for the test cases generation. One flaw in the models is a similar error in the test cases, with as a result the possibility of missing defects. The result then is greater risks when the system is taken into production.

OORT
Evaluating models is one thing, but how can you review the models as effectively and efficiently as possible. Because the models used for MBT are UML models and diagrams and these are not so easy to read. In 1999 a working group at the University Of Maryland (USA) did a study to evaluating models [Travassos, 1999]. Their result was: OORT, Object-Oriented Reading Techniques. In this study, the group members developed special techniques to compare UML diagrams against each other and against the requirements, with the purpose to find defects. They’ve created a dichotomy between horizontal and vertical reading. With horizontal reading the designs of the various UML diagrams are compared in the design phase. The focus is on the consistency between the diagrams. In vertical reading some UML diagrams are compared with earlier products (from the analysis phase). It focuses on the traceability and completeness of the models. It is also possible to use OORT for evaluating static or dynamic components in the models. The relationship between horizontal and vertical reading technique and static and dynamic components are expressed in 7 official reading techniques:

• * OORT-1: Sequence Diagram vs. Class Diagram (horizontal, static)
• * OORT-2: State Diagram vs. Class Description (horizontal, dynamic)
• * OORT-3: Sequence Diagram vs. State Diagram (horizontal, dynamic)
• * OORT-4: Class Diagram vs. Class Description (horizontal, static)
• * OORT-5: Class Description vs. Requirements (vertical, static)
• * OORT-6: Sequence Diagram vs. Use Cases (vertical, static/dynamic)
• * OORT-7: State Diagram vs. Requirements/Use Cases (vertical, dynamic)

When evaluating models for MBT mainly OORT-2, OORT-4, and OORT-5 and OORT-7 are important.

The above mentioned techniques consist of several stages with specific questions and allow the reviewer to focus on the different design aspects on consistency and completeness. However, manageability and testability are not affected by these techniques.

Defects found by these techniques are classified as:

• * omission, missing part;
• * additional information, is not in the model;
• * interpretation, misinterpretation of the concept;
• * ambiguous, no clear concept;
• * inconsistence, no agreement, or
• * other, every other defect.

The severity is expressed as:

• * High, no further reading possible;
• * Average, this part of the document is invalidated, or
• * Low, should be studied.

Tip: for further information: http://sel.gsfc.nasa.gov/website/sew/1999/program.html

Result
The use of OORT at Ericsson [Conradi, 2003] is compared to normal (ad-hoc) reading. It has proved that OORT is both effective and efficient and is also more practical. At Ericsson, they have proven that the cost savings of 20% resulted in the development of applications!

Bibliography
[Travassos, 1999]
Travassos, G.H., Shull, F., Carver, J, Basili, V.R., Reading Techniques for OO Design InspectionsTec, Proc. Twenty-Forth Annual Software Engineering Workshop, NASA-SEL, Greenbelt, MD, December 1999, http://sel.gsfc.nasa.gov/website/sew/1999/program.html
[Conradi, 2003]
Conradi, R., Mohagheghi, P., arif, T., Hegde, L.C., Bunde, A.A., Pedersen, A., Object-Oriented Reqding techniques for Inspection of UML Models (An Industrial Experiment), 2003