I've done studies on it this semester in my Database Design subject at uni. If you like, I could dig up the slides covering it if you'd like to read another perspective on it. Though if your familiar with the theory it probably won't do much good.
The way I see it, its about learning steps on normalising because although in some Universe's of Discourse your familiar with and can see the key's and efficiency issues immediately, in more complex UoD's it may be more difficult in identifying the most efficient databases.
That said, not all the steps are completely applicable in every situation, and depending on the DB's use, it may not be advisable to completely normalise in the first place. (completely atomic databases would be hell for large queries for example).
Just give me a heads up if you want those slides.
