System Software for Ubiquitous Computing

• Authors: Tim Kindberg;Armando Fox
• Venue: IEEE Pervasive Computing
• Year: 2002

The authors discuss how two key characteristics of ubiquitous computing systems-physical integration and spontaneous interoperation-affect the design of ubiquitous system software. Physical integration provides a connection between computing nodes (virtual world) and the physical world. A system that makes this connection must make clear where the "semantic Rubicon" lies-the line that divides physical-world semantic processing done by the system from processing done by human users. Spontaneous interoperation involves interaction among heterogeneous components that spontaneously appear and disappear. Events and tuple spaces are examples of models of interoperation that enable spontaneous interaction inside system boundaries, but we have yet to enable purely program-driven spontaneous interoperation across boundaries. Situations considered "failures" in nonubiquitous systems-network disconnection, a component abruptly leaving an environment, transient device or software faults-will occur much more frequently in a ubiquitous system. We must start thinking of such cases not as failures, but as part of normal operation, avoiding techniques that assume only rare failures and possibly entail ungraceful or expensive recovery. Additionally, we can leverage considerably from the mobile computing work that uses adaptation- and transformation-based middleware to address heterogeneity and dynamism, but some generalizations to the techniques are necessary to let them blend gracefully into the background, as Weiser suggests in his vision of calm computing.