We examine how computer system problems can be derived from real application problems, with a particular focus on the relevance of some assumptions, especially those related to computational models. Then, we compare these models, ranging from pure synchronous to pure asynchronous semantics, to conclude that synchrony does not necessarily dominate asynchrony whenever one is concerned with real operational systems. The issue as to whether asynchronous solutions can be considered for designing and building real-time distributed dependable systems is addressed. A priori, time free solutions are antagonistic with proving timeliness properties. We show how to circumvent this apparent contradiction via the late binding principle. This principle, as well as drawbacks of synchronous solutions, are illustrated.