Michael Stonebraker
Adjunct Professor
Massachusetts Institute of Technology

Title:

Where Should We Go From Here?

Abstract:

In the 1970's and early 1980's, the research community served as a visionary beacon on future DBMS technology.  The theory of relational systems, early implementations, access methods, and transaction management all came from our community.  At that time, we clearly led industrial activity, which looked to us for guidance.

Since then the track record of our community has been spotty.  The concepts we have worked on (e.g. Object-Relational DBMS, Object-Oriented DBMS, Distributed DBMS) have not taken off.  Moreover, important concepts in the marketplace (e.g. OLAP, XML, app servers) have evolved with little and/or late participation by our community.  Lastly, our community has become a master of least-publishable units (LPUs), writing papers on "safe" topics with lots of mathematics and graphs.

Even the most forgiving in our community will admit that the SIGMOD industrial presentations are way more interesting than the research ones. Without much of a stretch, one could claim that we are following industry, not leading them, at the current time.

Moreover, some of the ideas we roundly criticized during our "golden years" (e.g. hierarchical data bases) have suddenly become respectable again (XML Schema).  As such, we appear to have lost our historical perspective.

Of course, commercial DBMSs work much the same way they did two decades ago.  Whenever technology is moving slowly, it is difficult for research to lead industry.  To lead industry we require a fundamental "sea change" in our technology.  This talk focuses on a collection ideas, each of which could be such a sea change, and argues that we should work on this agenda rather than the current collection of safe topics.

This talk also argues that most new ideas are not, in fact, new but have been proposed before.  As such, it is important to have a strong historical perspective, so we avoid reinventing the wheel and repeating historical mistakes.

back