Every year we witness acceleration in the availability, deployment, and use of distributed applications. However building increasingly sophisticated applications for extant and emerging networked systems continues to be challenging for several reasons: - Abstract models of computation used in distributed systems research often do not fully capture the limitations and the unpredictable nature of realistic distributed computing platforms; - Fault-tolerance and ef?ciency of computation are dif?cult to combine when the c- puting medium is subject to changes, asynchrony, and failures; - Middleware...