CS262B Reading Summary

Of Objects and Databases: A Decade of Turmoil

Michael J. Carey and David J. DeWitt

Summary by Feng Zhou
2/23/2004

Strong points of the paper are:

  1. The reasons listed for the failure of database toolkits are reasonable.  Basically, it is really hard to divide a database system into components with standard interfaces and semantics, and to provide a toolkit for buildng ever component.  There are too many ways a component needs to behave in order to satisfy all users' requirements.
  2. The reasons for the failure of OODB approach are concrete.  It's hard to come up with a viable standard with such a complex system.  Without this standard, user acceptance is hard to achieve.  And the performance advantages of OODBs are often offset by the lack of sophisticated optimizations implemented as mature RDBMS systems do.  And the number of popular and appearing languages make good integration and compatibility hard to achieve.  Lastly, the lack of commitment from large vendors is also a very important reason of its failure.
  3. Object-relational databases, as an evolutionary approach, have enjoyed commercial success as the authors predicted.  The object-oriented client wrappers, taking the name of object-relational mapping tools, are becoming more and more popular. Another interesting category is mapping tools with language support, such as the JDO standard for Java.
  4. One interesting hypothesis from reading the paper is: if a technology a big inter-related whole that's hard to separate into clear-defined parts.  It's probably not a good idea, even if it has appealing properties like good performance.  Trying to building a whole new big system will often fail because the repeated investment needed to just bring the system on par with the old one.