It turns out that my co-author Doug Rosenberg has a few things to say on the subject of use case style. So this week I'm going to leap in the back of the Dodge Ram pick-up truck and let him drive.
Over to Doug:
There's often a fair bit of confusion in the early stages of a software project about what exactly needs to be built, Doug says. So it's worthwhile investing some effort in getting to a point of clarity about the requirements before leaping into construction.
One of the chief points of confusion is the mangling-up of "active voice stuff" with passive voice stuff. Scenarios (or stories) are "active voice stuff", e.g. "the user clicks the cancel button and the system erases the disk".
In the UML universe, we call the active voice stuff (which describes runtime behaviour) "use case scenario text" and the passive voice stuff "functional requirements".
It's useful to separate the scenarios from the functional requirements, not least because you must understand the runtime behaviour unambiguously before you code the system - and this is much harder to do when it's "intermangled" with the functional requirements.
At ICONIX, we often work with projects that are contractually required to demonstrate compliance against a formal, written set of requirements - banking systems, aerospace/defence projects, etc, and we've run into this "intermangling" phenomenon more times than we'd like to recall.
Here's an example: Once, when I walked into a client's building to conduct a training workshop, I was handed a hefty use case document. As I flipped through the use cases, I noticed that the scenario text was interspersed with lines that started with [REQUIREMENT]. So there would be a couple of lines of active voice scenario stuff, followed by a couple of lines of [REQUIREMENT]s, and so on.
Several of these uniquely numbered [REQUIREMENT]s were repeated on multiple pages (all with unique numbers, of course). So I took a deep breath and had the following conversation:
DOUG: It looks like some of these functional requirements repeat across several use cases...
CLIENT: Oh definitely! Some of them are in there 70 or 80 times.
DOUG: Oh... (silently emits "primal scream")
DOUG: Pauses (looking calm while still screaming silently)
DOUG: And they all seem to be uniquely numbered...
CLIENT: (brightly) Yep!
DOUG: So... how do you know whether you have 320 distinct requirements, or four requirements repeated 80 times each?
CLIENT: (smiling) Don't know... never thought about that one.
DOUG: What did you say your name was?
I had recently returned from teaching another class in Texas, where they're fond of taking liberties with the English language. Thus, the word "disintermangling" was born.
I'll wrap up with this thought: Unambiguous scenario descriptions are key to understanding what you're building before you code it. But before you can disambiguate the scenarios, you need to disintermangle them from the functional requirements.
Doug Rosenberg is an author and noted OO instructor, and runs public training classes worldwide.