Rethinking structural computing infrastructures

Abstract

Structural computing asserts the primacy of structure over data. This has often been understood to mean that all levels of a structural computing system architecture should exhibit structure awareness, leading to data models centered around so-called "structural atoms." While systems based upon structural atoms do provide ubiquitous first-class structural abstractions, they also freeze the "granularity" of the structuring process throughout their architectures at design-time. That is, decisions regarding representations of structures in structural computing architectures based upon atoms cannot be recast at run-time. In this paper, we examine an alternative to atom-based models for structural computing systems that allows exactly such recasting. We demonstrate how this alternative model, which we call EAD, is superior to atom-based models for certain important applications, and describe our initial prototypical implementations.