In this chapter, I’ve given some strong arguments in favor of not denormalizing; in effect, therefore, I’ve given arguments in favor of normalizing.[85] And it’s certainly true that good designs are usually fully normalized. But it’s important to understand that the opposite isn’t necessarily true! That is, a design can be fully normalized and yet still be bad. For example, the projection ST of relvar S on attributes SNO and STATUS is certainly in BCNF—in fact, it’s in the highest possible normal form, as we’ll see in Part III of this book—but it’s clearly not a good design, as we saw in Chapter 6.
[85] One reviewer noted that another advantage of normalization is that it tends to simplify the formulation of complex queries, in that it makes expressions more modular through nesting of simple subexpressions.