Happy Birthday, RDBMS!

04/30/2010 by Sylvain Hallé

This article in the latest Communications of the ACM retraces the rise of relational databases over low-level and homegrown systems that were available at that time. I highlight two factors to this success:

  1. RDBMS are declarative rather than imperative: the model concentrates on specifying what users want, and not how it should be computed.

  2. In the 1970s, three projects translated these concepts into working implementations that real people could use.

What lessons can we learn from this success story? In my view:

  1. Declarative is good. Some people might counter: "yes, but you have to sit down to write a declarative specification of a system/module/etc., and this takes time". Yet this is time well invested. Among the nice side effects of the relational model's declarativeness, the article cites IBM's Ronald Fagin: "For theorists like me, it was much easier to develop theory for it. And we could find ways to make the model perform better, and ways to build functions into the model. The relational model made collaboration between theorists and practitioners much easier."

  2. Implementations matter, although many in the academic world dismiss them as "mere tool papers". Translating an idea into a reasonable prototype is less straightforward than it appears, and the amount of work required to find real examples and make it work on them should be regarded as a contribution in its own right. Stated otherwise, inventors of new concepts will not be enticed to implement them if their work does not translate into academic currency: one must be able to earn good papers in good conferences/journals for case studies and implementations, or that work just won't get done.

comments powered by Disqus