Interface-oriented middleware and distributed service inference

  • Authors:
  • L. Robert Varney;D. Stott Parker

  • Affiliations:
  • University of California, Los Angeles;University of California, Los Angeles

  • Venue:
  • ARM '04 Proceedings of the 3rd workshop on Adaptive and reflective middleware
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Distributed systems that dynamically adapt to changing environments should be developed based on incrementally defined, loosely coupled services. But current middleware platforms tend to encourage monolithic services that are tightly coupled with their environments. We suggest that the non-incremental nature of middleware services is due in part to the lack of support for implementation inheritance, and that tight coupling is due to insufficient support for interface abstraction. Interface-oriented programming (IOP) is an extension of object-oriented programming in which all program inter-dependencies, including inheritance dependencies, are expressed using interface abstractions. This encourages development of partial classes, and requires a new program linking mechanism called representation inference to integrate them. We propose that a distributed implementation of IOP including partial services and distributed service inference could address some of the limitations of middleware systems, providing new opportunities for performance reflection and transparent adaptation.