Service Oriented Programming - which combines Distributed Object Computing, Component Based and webbased concepts - has recently emerged as a promising approach to develop dynamic and heterogeneous service provision environments. Such systems are referenced in literature as Service Oriented Architectures. Classic strategies to address dependability in Distributed Object Computing middleware may not be straightforwardly applied to Service Oriented Architectures (SOAs) for pervasive computing, since they operate in quite different contexts compared to traditional DOC middleware. In this paper focus is on dependability issues of SOAs. In particular, we identify dependability requirements of such systems during their life cycle showing how these requirements may change depending on the time phase (discovery, lookup, setup, delivery). We also explore the suitability of Jini technology as an enabling infrastructure to improve availability and reliability, describing both the benefits and drawbacks of a Jini-based solution in the context of a case study application, namely PRINCEPS.