A methodology for classification of software errors
The problem of lack of reliability in computer software continuously plagues modern computer systems. Unreliability in software is directly related to the presence of software errors. Unfortunately, no comprehensive data base of software error data has existed for study. Futhermore, no theoretical methodology had been developed with which a comprehensive data base of software error data could be developed. The purpose of this thesis was to develop the theoretical basis for development of a software error classification methodology.
The methodology was developed from several primary sources of data. Previous attempts to classify software errors were studied for insight into the breadth and difficulties of the process. Analysis of measureable parameters and characteristics of software, software errors and reliability provided useful input to the classification structure. General information classification theory and the analysis of functionality in the computer and software structure provided the underlying, theoretical basis for the software error classification structure.
The resultant software error classification structure contains 52 comprehensive error parameters expressed in terms of a single unique error vector. The thorough parameterization of software errors through the development of this classification methodology will allow the collection of critically needed data about software errors. This data base, in conjunction with a data base of general parameters of software, will yield to analysts greater insights into the error processes which make modern software systems unreliable. These insights and new understanding will provide many new tools for increasing the reliability of software.
Thesis77.E925.pdf
2.87 MB
Adobe PDF
c90072462341daf8f7316cb3041f563c