Project on Diverse Software — An Experiment in Software Reliability

Abstract The Project on Diverse Software (PODS) is an experiment which has attempted to quantify ehe impact of a nunber of commonly used software development techniques on software reliability. The main objectives of the project were to: – Provide a measure of the merits of using diverse software. – Evaluate the formal specification language X. – Compare the time and effort expended using high-and low-level languages and using formal and informal specification techniques. – Compare the reliability of software written in high-and low-level languages. – Evaluate testing methodologies. – Observe the software development process This paper describes the experimental design, organisation and docunentatlon of the project and presents an analysis of the results. Some of the major conclusions of the experiment are that: – Diverse implementation was effective in revealing faults not discovered by normal development methods. Testing diverse programs ‘back-to-back’ Is a powerful method for discovering residual faults. – The functional specification contained the most persistent program faults and was the only source of comncxi mode failure. Better methods of establishing the customer requirements are needed. – Assembler programming required about twice as much coding effort and generated about 4 times as many coding faults as the equivalent Fortran program, but the nunber of residual faults was about the same. – Current ‘best practice’ software development methods are effective In removing imploaentatlon faults.