Testing interactive systems is known to be a complex task that cannot be exhaustive. Indeed, the infinite number of combination of user input and the complexity of information presentation exceed the practical limits of exhaustive and analytical approach to testing [31]. Most interactive software testing techniques are produced by applying and tuning techniques from the field of software testing to try to address the specificities of interactive applications. When some elements cannot be taken into account by the software testing technique, they are usually ignored. In this paper we propose to follow an opposite approach, starting from a generic architecture for interactive systems (including both software and hardware elements) for identifying in a systematic way, testing problems and testing needs. This architecture-driven approach makes it possible to identify how software testing knowledge and techniques can support interactive systems testing but also where the interactive systems engineering community should invest in order to test their idiosyncrasies too.