[gruseom]

It isn't that simple. Excessive generality is one problem. Insufficient generality is another. Some things are easy to change later. Others aren't. It's part of being a good programmer to make good judgments about these things.

What people usually do (and this author, in my opinion, is doing) is observe one unhelpful behavior and conclude that the opposite behavior must be "correct". I think this is because building software is a complex activity, so we seek invariant principles to simplify it. Then we view the complex activity through the filter of our "correct" model (which we've identified with it) and get a feeling that we know what we're doing. In the end this causes a lot of problems.