This work is a proposal regarding the validation of telecommunication
feature specifications. It is more precisely concerned with the detection
of feature interactions, which is a problem that hinders the rapid introduction
of new features in telecommunication systems. A feature interaction may
occur when several features, developed in an independent manner are simultaneously
available: this coexistence may alter the behavior of the features.
The validation of specifications requires modeling both the services and the underlying network. The proposed solution relies on a methodology for formally specifying and validating features. It is based on the use of a synchronous formal language for the specification part and of some specific testing methods for the validation part.
The thesis describes a new testing method called ``pattern-guided testing'',
which we embedded in Lutess. Lutess is a testing tool for synchronous software,
whose basic principle lies on a dynamic and random but constrained generation
of test data.
Our testing method extends the notion of constraints in Lutess with
guiding patterns. A pattern describes a class of environment behaviors
which one wishes to favor. In the context of Lutess, patterns are seen
as loose constraints: they influence the generation so as the behaviors
they represent are more often produced.
This method has been formalized then validated on several case studies.
The most important one has been conducted in the context of the ``First
Feature Interaction Detection Contest'', held during the 5th Feature Interaction
Workshop in 1998. On that occasion, Lutess won the Best Tool Award.