The Navigation Toolkit project revisited

Abstract

report is a follow-on to “The Navigation Toolkit,” which was presented at OOPSLA ‘94. The Navigation Toolkit was the first fully objectoriented project within our company. and was “completed” in May 1994. But its impact since then has been enormous. Its success was used as the primary hard data point for selling a much larger follow-on project to NASA. The basic Toolkit architecture has been selected as the architecture for all future systems, but changes will clearly be needed to expand into new domains. The Navigation Toolkit itself continued on in a somewhat independent M&M mode. Users gained some capability over existing tools (such as the ability to model any number of space vehicles) and lost others (such as the ability to model tethered satellites and other, similar forces). The lost capabilities have resulted in user frustration, and the new capabilities have not yet been exploited. Another influence on the evolution of the Toolkit is the continued slow-down in the aerospace business. This is leading our company to seek new business opportunities, and improved flexibility in software could be a key to success in this area. 1. Scaling up is hard to do The Toolkit was built with a 7person. mostly highly experienced team. As reported in our original experience report, despite the introduction of new technology, etc., the team performed nearly as promised, achieving an overall productivity rate of 300 SLOG/Person Month and low defect density rates. The success of the Toolkit (and an object-based C predecessor, built by the same team) was the primary hard data point used to sell the Reusable Object Software Environment, or ROSE, to NASA. ROSE is a proposed major rewrite of most of the design and analysis software used for space shuttle trajectory operations. Rockwell management was concerned that the kind of software development organization which proved successful for Toolkit development would not scale up to the full ROSE project. In particular, management felt it could not count on the availability of highly skilled developers and chose instead to rely on minimally-skilled developers (90 days of training). It also needed to work under an accelerated development schedule, meaning it needed a larger development team. ‘Current address: TRW Systems Integration Group, Rosslyn, Virginia 2Current address: McDonnell Douglas Aerospace Houston Division Addendum to the Proceedings OOPSLA ‘95 61 A raw size comparison of the two projects is contained in the following table: Code produced (KSLOC) Nav Toolkit ROSE actual s proposals