This page presents the general framework of software design as I formulated it in the 1980s: design as an iterative process of synthesis, analysis, correction, and refinement. The emphasis here is on the design model and on the abstract machine. The concept of interform is only introduced as a transition point toward the page dedicated to it.
In its simplified form, design can be described as a feedback loop: synthesis → analysis → correction, repeated until the result becomes sufficiently precise. In practice, the designer works with mental models and with internal “libraries” of meta-models, selected and instantiated according to the problem.
A distinction that seemed essential to me was the double perspective on the program:
This duality led me to define a basic unit for the design process: the interform — an intermediate description of a part of the program under development, linked simultaneously to other interforms (synchronic) and to its own versions (diachronic).
To model design as a deductive process, I formulated a framework in which design means deriving a final design from an initial specification by applying norms (rules) to interforms, within a semantic universe in which they can be evaluated and compared.
In particular, comparison between intermediate forms becomes possible through a preorder relation (≼), which expresses the idea of “progressive refinement”: one form may be more detailed, more coherent, or more adequate than another, without necessarily being identical to it.
To anchor the formalism in a verifiable universe of computational processes, I used the concept of the abstract machine as the meta-model of the entity that “executes” the behavior described by an interform. In a formulation inspired by automata theory, an abstract machine is defined by states and actions that move the system from one state to another.
On many occasions I explicitly stated that formalism alone does not carry the process to the end: design remains a human activity, and where rigor becomes insufficient, knowledge and heuristics must intervene — an idea that, in the 1980s, naturally led me to ask: what can an “intelligent” system do as a partner of the designer?
The design model required a name for the intermediate states of the software product. From this need emerged the concept of the interform, treated separately on the next page. Here it is only the natural consequence of the model: if design passes through successive descriptions, then these descriptions must be identified, preserved, compared, and transformed.