Abstract This paper explores the apparent paradox that, although performance improves with experience, the maintenance of a software system becomes more difficult as experience with it accumulates. It is asserted that this phenomenon is a side effect of the environment used to maintain the software. A decade of experience with a large and complex clinical information system is summarized to demonstrate that software maintenance may be inherently less difficult than original development. It is shown that a highly integrated, million‐line, operational system, used in life‐threatening situations, has been maintained effectively by a staff of six. During the past five years, the size of the system increased by 167%. In addition to maintaining a growing number of programs, the staff produced new programs at a rate of 0.64 programs per effort‐day. Moreover, the case study also demonstrates that this level of performance can be transported to other sites. The paper concludes with some observations regarding changes to the software process that will resolve the paradox.

This article presents a model of the software process that offers considerable potential for improving both quality and productivity. It has been promoted by many researchers under a variety of names. The contribution of this paper is twofold. First, it offers a framework and unifying name, “adaptive design,” for a set of features, common to many paradigms, that exploits some of the unique characteristics of software. Second, it provides an existence proof that adaptive design is within the current state of the art; empirical data from commercial-grade experience with a particular implementation demonstrate the economic benefits available from adopting this model. The new process model is contrasted with several currently favored models, and suggestions for both researchers and practitioners are offered.

The paper describes and evaluates an alternative representation scheme for software applications in which the requirements are poorly understood or dynamic (i.e., open). The discussion begins with a classification of requirements specification properties and their relationship to the software process. Emphasis is placed on the representation schemes that are most appropriate for projects with open requirements in a flexible development setting. Fragment-based specifications, which capture a conceptual model of the product under development, are recommended for such applications. The paper describes an environment, in production use since 1980, that employs this form of specification. Evaluations of the environment's economic benefit and of the specification scheme's properties follow. A final section contains observations about the nature of future software applications and the environments necessary to support their development and maintenance.>

The software process is characterized as a transformation from a need to a software product that responds to that need. An application development environment is described that manages this process. It builds a conceptual model for the desired response (in the form of a fragment-based design) and automatically generates commercial-grade, module-based programs that implement the response. A decade of experience with a large application demonstrates the viability of this approach. Some observations on the software process are made.>

This paper describes how an application development environment, which maintains all design knowledge in a holistic application database, provides a window into its design knowledge. First an overview of the processing model is presented, then the application database is characterized. Finally, the approach to navigating through the design knowledge is described. The point is made that if an environment manages the design in a holistic fashion, then a window into that design knowledge is essential.>

This paper examines the future of software engineering with particular emphasis on the development of intelligent and cooperating information systems (ICISs). After a brief historical overview, the applications of the 1990s are characterized as having open requirements, depending on reuse, emphasizing integration, and relying on diverse computational models. It is suggested that experience with TEDIUM, an environment for developing interactive information systems, offers insight into how software engineering can adjust to its new challenges. The environment and the methods for its use are described, and its effect on the software process is evaluated. Because the environment employs a knowledge-based approach to software development, there is an extended discussion of how TEDIUM classifies, represents, and manages this knowledge. A final section relates the experience with TEDIUM to the demands of ICIS development and evolution.

The role of information and knowledge in medical informatics is discussed and expert system technology is examined in that context. The scope of computer-based applications in medicine is presented. This is followed by a review of the current state of the art. It is shown how the information- and knowledge-oriented applications are merging. Some of the key concepts in both information and knowledge processing are discussed in the context of medical computing.>

This paper presents an overview of a representation scheme being developed to define static data structures, the restrictions that establish the data model's integrity constraints, and the computational models that operate on the data. The context for this discussion is the role of a uniform structured representation for application generation in which semantically-rich expressions of the application's behavior are transformed into operational models with efficient performance.

A panel session concerning the use of knowledge-based techniques is reported. The discussion covers the following topics: knowledge-based machine vision architectures; experiences from a knowledge-based integrated-circuit testing system; artificial-intelligence (AI) tools for requirements capture and analysis; an approach to knowledge modeling in expert systems design; and a homogeneous approach to bringing AI and database management systems together.>

Current query languages, such as SQL, assume that the user is familiar with the database schema including the attribute names, types, and relation associations. When a user has imperfect knowledge of this information (or when he balks at the data-processing orientation of the required statements), he normally asks an experienced analyst to perform his and hoc query. The Intelligent Navigational Assistant (INA) was developed for the U S Army as a prototype query tool that permits the users to specify requests using only domain terms familiar to them. Once a request is made, it is converted into SQL for processing1,2

Although it is common to view software maintenance as an activity that is distinct from software development, it is often true that the development it is often true that the development of novel products is mixed in with the evolution of existing products. Examples are families of similar applications, such as communications processors, that share considerable functionality, and a large scale information system that acts as a base for new tasks. In these situations, the maintenance process involves the adaptation of previous experience to new requirements. The author explores this type of environment, first by examining the software process, and then by relating the key activities to maintenance. An information system case study is used to illustrate how the experience of the past can be structured to facilitate future modifications.>

A design for a hypertext system that manages software documentation for maintenance purposes is presented. A brief overview of hypertext is given, followed by a model of a software maintenance documentation system based on the hypertext concept. This is accomplished in three steps: the software process is reviewed, the essential information required for maintenance is identified, and the data model is described. Experience with this documentation model in a nonhypertext implementation is examined.>

This paper describes a paradigm for implementing software in which 1) a database is used to maintain virtually all recorded information about the application and 2) the executable product is generated directly from the formal specifications. A specific environment, TEDIUM®, defined for a specific application class, the interactive information system, is presented as a proof of concept. The benefits of this paradigm are evaluated briefly, and the characteristics associated with its in productivity improvement are identified.

Abstract This paper makes some observations about the applicability of iterative development for system implementation. After a general discussion of the software process, the types of projects suitable for this approach are identified. A system designed to support the iterative development method is then briefly examined, and a case study of its use is presented. Conclusions about the process are drawn, and some research issues regarding productivity are defined.

A one day workshop on the topic “A Framework for Medical Information Science” was held in October. Three issues were considered: what is Medical Information Science, what are the key research issues, and what are the educational needs. Because of the brevity of the workshop and scope of the topic, no summary conclusions were developed. Individual papers and position statements will be published. This paper reviews the workshop and some of the principal issues discussed.