Some people have pointed out that Open Implementation (OI) can be seen as just a refinement of black box abstraction. Previous realizations of black box abstraction tended to leave some important information implicit in the interface. Black box abstraction can be extended to encompass OI by simply ``widening'' the module interfaces to encode this previously implicit information concerning performance, data representation, resource consumption, et cetera.
However, this completely misses the point.
I think the point of all this is that the way people build software is changing. The cost and complexity of constructing high-quality software, combined with the difference in performance between an untuned and a well-tuned implementation, have led people to develop techniques for making software modules adaptable to a wider range of usage patterns. We call this trend ``Open Implementations.'' Whether it's a new paradigm or a refinement of existing principles is irrelevant. The important thing is to recognize and understand this recent and allegedly ubiquitous change in the craft of software.
Thus, the purpose of this workshop should be to catalog examples of OI, and figure out what it is about each of them that makes them an open implementation. Only when we have enough well-understood examples, can we begin to say what an OI is and how people should go about building them.
Moreover, OI does not represent any new technology; the open problems in Operating Systems, Programming Languages, etc. are still unsolved. The only thing something like this workshop can do is to help people better understand what they are already doing, and maybe serve as a source of ideas from other disciplines that can be applied to our respective areas.
Was that inflammatory enough?
Back to Alphabetical List of Responses
(Last Revised October 1994)