This position paper summarizes the most important problems that must be solved in order to establish a Component-Based Software Engineering (CBSE) discipline for resource-constraint systems. Resource-constraints are especially relevant for embedded systems (e.g. telecommunication systems and many modern consumer products), real-time systems and dependable safety- critical systems, but can be encountered in virtually all types of systems. In this paper, we define a number of research challenges in the form of requirements that are not yet met by contemporary component models. For each requirement, we explain its relevance and suggest directions for possible solutions. We concentrate on the architectural level, since it is here that CBSE has the highest benefits in terms of adaptability and reuse. We also do not consider non-functional constraints in general, but only dependability constraints (timeliness, performance 1 , reliability, availability and security). The reason is that the latter directly affect the feasibility of a given component configuration, while this is less obvious for general constraints like scalability, maintainability and interoperability.