FramerD: Representing knowledge in the large

Content-aware media applications require rich, interconnected descriptions of media content. FramerD is an object-oriented database developed at the MIT Media Laboratory to support just such descriptions in a scalable and distributed manner. Conventional databases, even the newer object-oriented ones, focus mostly on data structures whose components are scalar or literal values, rather than references to other objects. FramerD is optimized to support objects that have components consisting of references to other objects in an environment where the storage of and the operations over the objects are distributed over local- and wide-area networks. This paper discusses the design, implementation, and performance of FramerD, and sketches some current applications of the system.