Abstractions

I'm an old database hand—and I think that using such technologies effectively is still an aspect of "good IT governance". I can remember times before RDBMSs (Relational Database Management Systems) and deciding that the speed of IBM's IMS Fast Path database outweighed its lack of flexibility, and that the new Oracle thing had problems with its optimiser, some 20 or 30 years ago. How things have changed and RDBMSs are now—almost—ubiquitous (in the public consciousness, anyway; IMS is still a current product). But you can still make me lose my rag by saying things like "RDBMS is all very well, and very flexible but it is a bit slow". Speed is simply nothing to do with the RDBMS specification (which is a logical implementation of relational set theory)—if an RDBMS is slow, it is simply because it has been implemented badly (usually, because someone thinks that tables have to exist in physical storage as well as in the logical view—you can make such implementations fast, but it's a struggle).

There are a few counterexamples of fast RDBMS implementations with complete abstraction between the logical view and the physical storage. Intersystems Cach