I sat in on all three of the Stanford courses last year, thinking AI was going to be awesome and Databases the one I'd drop if I ran out of time because I've dabbled in databases as a quaternary function of my job. It turned out quite the opposite. While AI felt like I was watching Sebastian Thrun's hand draw exercises, nearly everything in the Jennifer Widom's class was useful.
She covers a broad overview of database-related technologies (list posted above by another person). The material consists of video lectures (with ungraded, in-video quizzes), exercises and programming assignments. The exercises and programming assignments are designed to let you retake them (within the scheduled time window). The material certainly has some academic treatment, but what I found most useful, were the hands-on exercises with a variety of technologies. Their test harness was great, and I learned a lot from debugging my stuff.
There are optional "screenside chats," which I found entertaining. The forums were helpful when I was really stuck on relational algebra, but the signal to noise ratio is a bit discouraging - there are a lot of people whining about credit on the exercises or the certificate of achievement, as if learning wasn't a prime goal.
One word of caution: There are several exercises due at the end of the course that take a non-trivial amount of time to work through. It is thus important that you pace yourself, even being ahead of schedule through the fourth week.
To recap: the course was very useful and I've already applied some of the things I've learned. I'd definitely encourage you to take it.