New engineers in the software industry argue the profound impact of good requirements engineering on a project. Even fewer engineers can claim that they spend any tangible effort on requirements engineering. Most sheepishly smile and claim, ironically, that their management fears that requirement analysis might negatively impact the delivery of the final product. This, of course, hints to a devil in the details.
In general, the software industry is obsessed with the "end game," or shipping the "final" product. As evidence of this obsession, witness the tools most prevalent throughout our industry:
- Languages of all shapes, sizes, and colors with a myriad of accompanying development tools
- Impressively sophisticated debuggers tooled to focus on the machine's most minute details
- Libraries enabling weekend warriors to craft entire applications in the blink of an eye
- A deluge of research into more beautiful ways to quickly construct software.
Academic institutions groom us as solution-space engineers from the moment we express an interest in building software. We learn algorithms, language theory, computational theory, discrete mathematics, and calculus. Each of these lessons is meant to help us solve problems. But we never learn about the problem space.
How do we identify the problem? How do we extract information about the problem? How do we organize our findings? Ultimately, we never learned how to restrict ambiguity in the problem space. The disease manifests symptoms such as ill-defined requirements, unmet requirements, unknown requirements, and design-dependent requirements (one of the greater sins).
Many requirement specifications introduce design dependencies meant to restrict the creativity of the designers and developers. This is the exact opposite of what a specification should do. For a legion of engineers groomed with incredible technologies at their disposal, a specification must restrict ambiguity in the problem space while simultaneously introducing latitude into the solution space. Injecting latitude into the solution space enables our legion of engineers to bring the force of their creativity to bear by providing them with the gamut of possible solutions to what the specification must identify as the user's need. This strategy acknowledges that technologies change by the hour, while problems change by the day.
But why stop at the creativity of engineers? The power we demand from our solution-space tools has spawned stories of clients building portions of the final deliverable. The blurry line between developer and client suggests that we shouldn't concern ourselves as much with the solutions space. It follows, too, that the antiquated coursework offered by our academic institutions requires updating.
Of course, academic institutions are slow to respond and we need help now. Distressingly, the ocean of tools available to solution-space engineers rarely helps problem-space engineers. Some argue that CASE tools help capture data requirements. Unfortunately, the notations that CASE tools offer are often those used to capture design-level details, leading to bias and confusion. While many CASE tools enable problem space engineers to capture use cases and their exemplary dynamic behavior, they do not aid in succinctly explaining those diagrams in words. It is here we face the devil face-to-face -- and smell the Jolt and coffee on his breath.
Wang Bi, a third-century Chinese scholar, has this to say in "General Remarks on the Changes of the Zhou":
Images are the means to express ideas. Words are the means to explain the images. To yield up ideas completely, there is nothing better than images, and to yield up the meaning of images, there is nothing better than words.
In other words, he suggests that to capture ideas from the problem space, we need both imaging tools and descriptive tools. Tools for imaging ideas, such as UML, BON and the like (flawed as they are for modeling requirements) will suffice until some bright engineer provides a better solution. One possible descriptive tool is requirement patterns.
