Date: 04 Jan 97 03:55:52 From: jfmezei <email@example.com> Organization: SPC References: 1 2 Followups: 1 2
View raw article or MIME structure
A few more comments on the "unreliability" of a "million lines of code" software. A single program with a million lines of code will be more prone to errors than 100 programs with a thousand lines of codes each. If your software runs on a proven operating system, or on different computers, modules are not likely to pollute other modules with memory leaks etc. So you are looking at certifying separate smaller modules instead of one huge module. The ods of errors in smaller modules is much smaller than in a larger module. Furthermore, in a "client-server" environment, you can test how the server will respond to requests (whether valid or invalid) and how the client will respond to responses (valid or invalid). This is not as easy to do whith one large program with multiple subroutines calling each other. Whether the software is bug free or not is not really the issue. What is the issue is how the software reacts to unanticipated inputs (eg: plane in the air but landing gear system tells the computer it is on the ground or vice-versa). If such conditions are known, they would be documented in technical manuals for the plane so pilots know to to handle such situations, so they can also be included in the programming so that the computers know how to handle this situation. If such conditions are unknown, then the pilots will have to respond to erroneous warning lights on a conventional planes, or erroneous behaviour or warnings on a FBW plane.