Exploring a problem

There are not many problems in software development that can be fully specified easily. Even the few that seem like it leave some room for interpretation. When working on a project to implement a database adapter, I recently faced exactly this. There was a specification that needed to be implemented, and a set of unit tests making sure the implementation conforms to the specification. However, as I implemented it, I found myself asking questions along the way. The main question was very similar to what I would have asked if I hadn't had the specification: How are people going to use this piece of code? In a lot of cases, there are multiple ways to implement a certain feature, but picking one often means weighing different tradeoffs against each other, such as speed, extensibility, and readability.

In our orc dungeon, we have to ask the same basic question: How is the client going to use our software? Sadly, this question by itself will not yield the results we have...