Dynamic change management and Ada

Abstract

Many large real‐time systems are expected to have a long, non‐stop operational life. It cannot, however, be assumed that the code that initially started executing will remain appropriate for the complete lifetime of the system. Indeed it will more likely be the case that the system will be subject to evolutionary change. This can reflect bug fixes, the addition of new functionality, or redeployment of the original code. We introduce in this paper the term Dynamic change management to represent the process of controlling the modification of executing software. The Ada language will increasingly be used to program non‐stop systems. Unfortunately dynamic modifications to Ada programs appear to be invalid in terms of the Ada Language Reference Manual. This paper discusses dynamic change management within the context of distributed Ada execution. Our intention is not to provide a methodology within which such changes can take place; we believe this to be too difficult at the current time. Rather we wish to provide a focus for discussion of this important future problem area. However, we do offer a useful classification of software components, and suggest how change can be effected—the approach to be taken depending upon the structure of the component being replaced. The flexibility of dynamic change management depends upon the initial granularity of distribution and deployment. We also attempt to define rules for legally changing running Ada programs. Although this paper focuses on the Ada language, many of the issues are of general concern. Copyright © 1989 John Wiley & Sons, Ltd