The Feature Interaction Problem: Automatic Filtering of Incoherences &
Generation of Validation Test Suites at the Design Stage

Nicolas Gorse

Thesis submitted to the School of Graduate Studies and Research in
partial fulfillment of the requirements for the degree of Master of
Computer Science under the auspices of the Ottawa-Carleton Institute
for Computer Science.

Abstract

This thesis addresses the feature interaction problem in the research area
of telephony systems design. We start with a review of this problem and
present our definition of a feature and of a feature interaction. We
enumerate and discuss some known approaches with respect to their
strengths and limitations.

We present our initial approach and propose some refinements in order
to automate the detection of incoherences and derivation of validation
scenarios at the design stage using predicate logic and Prolog. The
identification of incoherences corresponding to potential feature
interactions allows the designer to refine the requirements and to
produce a better specification.

A UCM model is derived from requirements. This model allows the
designers to have a general point of view of the system and to detect
design defaults. The construction of the UCM model is followed by the
generation of a LOTOS specification.

The derivation of validation test suites and their application against
the specification permits to check whether or not the incoherences
identified in the requirements lead to feature interactions.

The results obtained over two case studies show that the application of
this approach improves the design and validation of a specification.