This paper originated with a software quality panel workshop at the EIA Computer Resources Workshop. The panel group initially focused on requirements for nondeliverable software. It was felt that the requirements for deliverable software were well covered by Department of Defense standards but non-deliverable software issues were not adequately covered. After much discussion, including attempts to define nondeliverable by examples of software types, a surprising unanimous conclusion was reached by the participants. That is: the states of "deliverable" and "non-deliverable" had no bearing on the quality of the software, and thereby also had no bearing on the quality assurance requirements that should be applied for that software. Rather, the only rational application of quality assurance was primarily dependent on the impact that the software had, or could have, on the end product (¡.e. the ultimate mission).