This paper draws together the results of three separate projects (all falling under what we call the "Kate" project) in which, over the past seven years, we have been studying requirements and specifications for complex systems. We present a set of requirements issues, then motivate them by examining three Kate projects: the critic, the planner, and the resolver.

Our results help to clarify our understanding of the issues surrounding requirements engineering, and support the idea that semi-automated tools can be built to address those issues. However, the domains in which we have been a.hie to work effectively are relatively small. Several important specification issues remain to be addressed.