As NASA starts to wind down their Space Shuttle activity in the next three years, the space agency's effort to return to the Moon has been ramping up quietly in the background. With their new Orion/Ares space vehicle combination, crew automation will definitely be on the top of software priorities for NASA. But with a much smaller budget and shorter timescale than the last lunar attempt, would NASA and its contractors embrace new approaches and techniques so our tax dollars are better spent? Can Dan Goldin's "Faster, Better, Cheaper" approaches be finally achieved?

A couple of months ago I was fortunate enough to join ThoughtWorks, a company that advocates the use of Agile software development practices (Extreme Programming, Scrum, TDD, etc.) to bring business value to our customers. I have been using Agile practices on my previous project for over three years and it had proved to be highly successful. And ThoughtWorks' experience in this area proves that Agile can also be applied successfully on large enterprise software projects. But can Agile be used on a highly mission-critical software project such as the one for the Orion spacecraft?

Over ten years ago my first programming job was for small software company developing real-time, safety critical software for controlling railway trains. The work we did was the embodiment of the Waterfall model. The system requirements were collected and analyzed. The model was designed and validated. Then we mere programmers set out to write code to realize the model. Huge amount of unit tests and integration tests were created to make sure our code did what the model said it should do. All the while, the project manager kept track of our progress to ensure that, hopefully, we delivered the product on time and on budget.

At first glance, Agile sounds like a good fit with this type of project where requirements are generally very well defined and correctness are paramount. Short iteration and test-driven development will ensure features are delivered often and proved to be working by the unit tests. Continuous integration means there will be fewer surprises as multiple systems are joined up to work with together. The costs of requirement changes will be reduced and can be implemented quickly, rather than in the next version.

But would the world of safety /mission critical software development, dominated by engineers and scientists, be receptive to the less rigid world of Agile development? Would they feel that without the top down approach, its highly structured development process, and the tightly prescribed set of delivery artifacts, the project delivery cannot be guaranteed?

I would love to hear from people who have more recent experiences in this area of software development with regard to Agile. Is it being used, is it being used widely, and how effective it has been?