Transitions in programming models

Summary form only given. The future of programming languages is not what it used to be. From the 50's to the 90's, richer, more flexible, and more robust structures were imposed on raw computation. Generally, new models of data and control managed to subsume older ones. But now, as programs and applications expand beyond a single local network and a single administrative domain, the very nature of data and control changes, and many long-lasting conceptual invariants are disrupted. We discuss three of these disruptive changes, which seem to be happening all at the same time, and for related reasons: asynchronous concurrency, semistructured data, and (in much less detail) security abstractions. We outline research project that address issues in those areas, mostly as examples of much larger territories yet to explore.