Software systems are central for the infrastructure of modern society. To justify the huge investments such systems need to live for decades. This requires software which is highly adaptable. Software systems must support a high degree of (spatial) variability to accomodate a range of requirements and operating conditions, and temporal evolvability to allow these parameters to change over time. Current approaches to reusability and maintenance are inadequate to cope with the dynamics and longevity of future software applications and infrastructures, e.g. for e-commerce, e-health and e-government. At the same time, we rely increasingly on systems that provide a high degree of trustworthiness. Thus, the major challenge facing software construction in the next decades is high adaptability combined with trustworthiness. A severe limitation of current development practices is the missing rigour of models and property specifications. Without a formal notation of distributed, component-based systems it is impossible to achieve automation for consistency checking, enforcement of security, generation of trustworthy code, etc. Furthermore, it does not suffice to simply extend current formal approaches. We propose to take an empirically successful, yet informal software development paradigm and put it on a formal basis. Specifically, we will turn software product family (SWPF) development into a rigorous approach. The technical core of the project is an Abstract Behavioural Specification language which will allow precise description of SWPF features and components and their instances. The main project outcome is a methodological and tool framework achieving not merely far-reaching automation in maintaining dynamically evolving software, but an unprecedented level of trust while informal processes are replaced with rigorous analyses based on formal semantics. This includes the perspective of designing self-adapting software systems.