Some of the basic assertions, such as the relative inefficiency of block compression in database engines, are true. I've seen material gains from using context/content-aware compression and some commercial OLAP databases exploit this extensively. They appear to be using many of the same kinds of techniques.

However, the assertions made around caching behavior, such as wasting memory due to double caching, are not generally true. While you will see this in simple/naive database engines, a sophisticated high-performance database implementation won't be designed this way.