I have taken also other MOOCs (Machine Learning, Probabilistic Graphical Models, Functional Programming Principles in Scala, ...) and Widom's databases MOOC was by far the best planned, best designed and best executed.

There was quite a lot of homework, but it was all well planned and served a purpose, you could do all the SQL, Xpath, JSON, ... query exercises in the browser without needing to install anything, and the system gave good feedback when you did something wrong. That environment was very well implemented.

The course was packed with theory and practice, there was minimal extra chatter in the video lectures, almost every minute of the lectures was used to cover the material, and the lectures provided just the right amount of theory for doing the exercises, and the exercises covered almost everything from the lectures. Good video lectures are more dense, and proceed faster than live lectures, because when you miss something, you can always rewind and watch the last minutes again.

It's crazy how many pages it took me to find the actual class: https://lagunita.stanford.edu/courses/DB/Recursion/SelfPaced...

(those select boxes are definitely non-intuitive, and the whole interface in general)

Anyway, great content!