Latte, M. (2004):
A Framework for Deductive Traders of Context Information
Context aware services often need derived and higher level information about users and their environments. Sensors and databases mostly offer low-level information only. The need to bridge this gap demands for refinement and ennoblement of contextual information. Therefore the concept of conventional traders known from distributed systems is adapted. Their core, typically a database, is equipped with its deductive closure. A complete deductive closure is ineligible for practical purposes because of a fairly high runtime complexity. Therefore a representation is used which is simple enough to maintain the deductive closure but which is still strong enough to model our environment adequately. All three key functionalities of conventional traders offered at their interface are emulated, namely to add and to withdraw service offers and context information as well as to enquire the availability of a, maybe derived, service. Algorithms for adding and removal are efficient as far as time and memory accesses are concerned since they are polynominal with low degree and measured in those changes that are imposed by deductive closeness. This avoids complete and expensive rebuildings of closures. Inquiries for services can be treated efficiently as well. For inquiries the requester needs to provide a construction strategy and a representation of the answer. This differs from conventional traders since structurally multiple answers are possible due to deductive closeness.