Program Design as a Systematic Activity (1987)
In this article I tried to reformulate, in a more concentrated and internationally communicable variant, my ideas about program design as a systematic human activity. The text starts from the general characteristics of design — its prospective, iterative, creative, controlled, paradigmatic, and collective character — and tries to show to what extent these features can be captured by a formal model sufficiently simple to serve as a basis for the architecture of a CAD system for program development.
The article is important because it makes explicit the connection between design theory and the construction of a software environment to assist the designer. In the final part, the text leads directly to the idea of the man-machine system based on interforms, explicitly invoking the INTERFORM system as an already realized implementation of this research direction.
Navigation
Paper Details
Bibliographic References
Author: L. D. Șerbănați
Title: Program Design as a Systematic Activity
Year: 1987
Content: program design as a human activity, design schema, formal model, and the transition to an interform-based CAD system
Internal link in the series of papers: the text explicitly refers to my 1983 article about the design model and to the paper about INTERFORM presented in 1987.
Structure of the Work
- Introduction — design as an activity difficult to systematize and the need for cooperation between man and computer;
- Program design as a human activity — the list of defining features of design;
- Program design as a schema — the schema of the feedback loop between synthesis and analysis;
- Program design as a formal model — the framework
<S, L, P, N; U, ε>; - Program design as a man-machine system — interforms, specialized processors, database, and interface;
- Concluding remarks — reference to the already implemented system.
Conceptual Contribution
The article has great synthetic clarity. It concentrates in a short space several previously formulated ideas:
- design is a model-generating and multi-view activity;
- program design must be viewed both synchronically and diachronically;
- a simple formal model is useful even if it does not capture the full richness of the real process;
- this model can structure the architecture of a design assistance system.
Theme of the Article
From the very introduction, the article poses the problem in very clear terms: design is the most difficult stage in the life cycle of any product and, precisely because of this, any attempt to make it more systematic is valuable. For programs, the stake is even higher since the software product has an abstract nature, and any progress in the design activity means increased productivity and improved program quality.
The declared goal of the paper is twofold: on the one hand, to identify the fundamental features of program design as a human activity; on the other hand, to use these conclusions to structure the architecture of a CAD system capable of taking over some of the designer's tasks. The article is thus a bridge between theoretical reflection and the actual construction of software tools.
Personal and Historical Context of Publication
The paper was submitted to the international congress on cybernetics and general systems held in London in 1987. I was unable to attend: I did not obtain the visa required to travel. I hoped that Professor Solomon Marcus, whose trip had been authorized, would be able to present the paper on my behalf, but I never learned for certain whether it was actually presented. I do know that the paper was included in the congress proceedings. For a researcher working in Romania at that time, publication in such an international setting carried a personal significance beyond its strictly scientific value.
In the sequence of my research, this paper continued the design model presented in Human Action Logic (1983) and the direction of intelligent CAD/CAM systems discussed in my Tokyo paper (1985). It was also connected to the implementation of INTERFORM, presented at ICSE in 1987. The London paper sought to explain the conceptual framework behind these efforts in a concise form: design as a cognitive activity that can be formalized only in part, in which the computer should support, rather than replace, the designer's judgment.
Design as a Human Activity
One of the most successful sections of the article is the systematic enumeration of design characteristics. It is described as an activity that is:
- prospective — oriented towards a product that does not yet exist;
- iterative — based on successive returns and improvements;
- creative — synthesis is not reducible to an algorithm;
- strictly controlled — every synthesis is followed by verification;
- model-generating — each step produces increasingly precise models;
- multi-view — the model has multiple perspectives on the product;
- paradigmatic — each model is an instantiation of a meta-model;
- empirical — paradigm selection still relies on experience, not on a complete theory;
- multi-solution — for the same requirements there can be multiple designs;
- approximate — initial requirements are incomplete and refined over time;
- collective — designing complex systems is a team activity.
For program design, the article then insists on the necessary double vision: a synchronic one, on the current structure of the project, and a diachronic one, on its evolution over time. This double perspective is one of the constants of my reflection on design.
The Schema and Formal Model
After characterizing design as a human activity, the article proposes a simple schema: a feedback loop in which the synthesis of a new representation is followed by analysis and verification. The schema is non-deterministic or, more precisely, must allow backtracking, because design can unfold along several possible paths.
On this basis, a very condensed formal model is also introduced. The following are defined:
- U — the universe of computations;
- L — the language of computation descriptions;
- S — the specification language, with
S ⊂ L; - P — the standard project language, with
P ⊂ L; - ε: L → U — the evaluation function;
- N — the set of design norms, formulated deontically.
With these components, design is described by the formal system
<S, L, P, N; U, ε>. A terminal theorem models the project, and its proof models
the design steps starting from the specification. However, the article also highlights the limitations of the model: the real process
can fail, initial requirements are not completely formal, and the model does not directly capture the organization of the design team.
Design as a Man-Machine System
In the second part, the article moves from the formal model to the practical question: what functions could be taken over by the computer? A distinction is made between higher-level functions, difficult to implement efficiently — deduction, recognition, intuition, creation, association — and lower-level functions, already accessible to computers: data management, input-output, data processing.
The central representation of the project here is the interform, defined as a piece of code or a part of the project interrelated with others and described by three main components: identification, relationships with other interforms, and body description. The body description must offer multiple views of the interform: text, syntactic structure, data interfaces, computation description, and other information useful to the designer.
Based on interforms, the article clearly outlines the system functions:
- textual, syntactic, semantic, and data-oriented processing;
- text editor and syntax-directed editor;
- syntactic and semantic analysis;
- symbolic execution;
- code generators;
- relational database with access rights for team members;
- input-output interface as complete as possible, based on menus or even natural language.
The explicit conclusion of the article is very important: such a system had already been implemented at the Polytechnic Institute of Bucharest and presented in another paper, the one dedicated to the INTERFORM system. This article therefore appears as a short and clear formulation of the theoretical justification for that system.
Contribution of the Paper
- condenses in a very clear form the main features of program design as a human activity;
- explicitly links the formal model of design to the architecture of a CAD system;
- emphasizes the role of meta-models and multiple project representations;
- introduces the interform as a practical unit of project organization and man-machine cooperation;
- acts as a synthesis text between previous theoretical articles and the practical work on INTERFORM.
Viewed retrospectively, the article does not bring major mathematical novelties compared to my previous works, but it has the merit of formulating very clearly, in English and in an international setting, why program design must be seen as a systematic activity and why a CAD-type environment can become a real partner for the designer.
Links
- The 1981 article: A model of program product design (bibliographic reference; the dedicated page has not yet been identified).
- The 1983 Article: Towards a Model of Design Methods for Program Products
- The Article about INTERFORM (1987)
- Back to "Order in Chaos"
- Back to Articles Index





