Equip, the Stanford EventHeap, L2imbo etc. are members of a family of distributed systems platforms based on the Gelernter's 'Tuple Space' shared dataspace paradigm. Increasingly such platforms are moving further away from the original application domain — large scale parallel processing on clusters of workstations — where tuples (or messages/ events) represent data to be consumed (processed) and replaced with the results of computations. Instead, the data exchanged in such platforms consists of sampled state (the status of real world things). Once represented in the data space, such information loses its relation and temporal coherence with the real world. A key challenge (e.g. in ECT) as we move forward to create higher level applications upon such platforms is thus to facilitate reasoning across data samples and through pipelines of components which may each rely upon data sampled at different moments and processed at different rates. Under the infrastructure challenge we aim to examine the underlying semantics of how data is processed in such platforms (and ECT in particular) with the aim of identifying a solution to this challenge. We are currently building on the work of G. Abowd and A. Dix, on 'Integrating status and event phenomena' into the ECT toolkit. We are also exploring the matching semantics required by ubiquitous computing systems through a series of prototypes based on the concept of a 'Distributed Event Heap'. |