Letters to the Editor

Abstract

Software Safety vs Software Reliability
 While looking back through Vol. 56, No. 1 (Summer 2020) of Journal of System Safety, I finally took the time to read Nathaniel Ozarin’s article “Lessons Learned in a Complex Software Safety Program.” The article is quite interesting and thought provoking, comparing what actually occurs while implementing a system safety program to the idealized descriptions found in documents such as MIL-STD-882, JSSSEH and AOP-52. While I found the article interesting and informative, I noted that the author consistently characterizes the “software safety problem” as a “reliability” problem, focused on finding and preventing “failures” and ensuring high “reliability.”
 Some Thoughts on the Probabilistic Criteria for Ensuring Safe Airplane-System Designs
 We have been employed in the risk sciences for a total of 86 years, including 62 years in reliability engineering and safety engineering positions at The Boeing Company. For many of those years, Yellman was the designated “Risk-Analysis Focal” (person) for Boeing’s 707, 727, 737 and 757 airplane models. For several decades, the United States government has published the same criteria, created by the U.S. Federal Aviation Administration (FAA), intended to ensure that the systems on large (transport-category) aircraft have been designed to be safe [Refs. 1 and 2]. But we believe that the criteria have failed to prevent certain aircraft accidents, and we think that the reasons for that should be better understood. We hope that this discussion will contribute to a better understanding by examining the part potentially played in those accidents by the FAA’s criteria that are defined probabilistically.