ODBMS.org

Operational Database Management Systems

Transcript of OOPSLA 2006 Panel on Object Databases Published

by · Published · Updated

Representatives from ODBMS and ORM vendors incl. BEA, Microsoft, and Sun explored the “State of the Union” in 2006

SAN FRANCISCO, Calif., Nov. 8, 2006 – ODBMS.ORG, a vendor-independent non-profit group of high-profile software experts lead by Prof. Roberto Zicari, today announced the release of an exclusivetranscript of last month’s OOPSLA panel on “Objects and Databases”, which discussed the difficulties and different strategies of dealing with persistence in object-oriented programming environments.

The panelists were:

  • William Cook (Moderator), University of Texas at Austin
  • Robert Greene, Versant
  • Derek Henninger, Progress Software
  • Patrick Linskey, BEA
  • Erik Meijer, Microsoft
  • Craig Russell, Sun Microsystems
  • Bob Walker, GemStone Systems
  • Christof Wittig, db4objects

A podcast of the entire 1.5 hrs. session is available onwww.odbmsjournal.org, a full transcript of the panel discussion is available here for immediate and free download. It is a unique document of this first high profile panel related to object database technology for many years, underlining the growing importance of solutions to handle the object-relational mismatch, including object databases (ODBMS) and object-relational mappers (ORM).

Robert Greene from Versant explained the superiority of object databases in cases of very large and complex models, e.g. in massive multi-player games, where object databases were, in one instance, able to reduce the number of servers from 150,000 to 70,000. At the same time he admitted the need for interoperability with relational databases.

On a similar note, Derek Henninger from Progress, the company that provides ObjectStore as well as ORMs, gave examples where object databases clearly excelled over relational solutions by avoiding excessive Joins over 20 or more tables, or by providing navigational capabilities which are important to GIS systems, for instance.

Patrick Linskey from BEA, formerly Solarmetric, explained the organizational benefits of ORMs, because they would enable negotiations between developers and DBAs over data model decisions, rather than forcing one party into the needs of the other. While he admitted the benefits of ODBMSs in data federation and partitioning, he saw them only in a niche, not serving the heterogeneous data requirements in a modern enterprise, which go beyond objects and tables.

Microsoft’s Erik Meijer refuted attempts to build Uber models and made a point for stability in data models: “I think data should not evolve… I really think it’s a bad idea if it’s easy to change your data.” The other panelists unanimously disagreed with this and pointed at several examples where agility is a must or where it brings a competitive advantage. Erik, on the other hand, saw Monads, first explained by mathematicians in the 1920s, as the solution to the problems at hand, which would make data accessible in ever changing ways, while the data itself would not be changed.

Craig Russell from Sun Microsystems, who was the lead for the JDO ORM specifications, likened the problem to a restaurant, where the food preparation is separated from its consumption: “There’s a real separation in my mind between what is in the kitchen and what shows up on your plate. My focus is what is on the plate.”

Bob Walker from Gemstone Systems explained that GemStone Smalltalk was the first to integrate persistence into the language and, at the time, saw Java going a step back, which haunts OO developers today. With Native Queries, LINQ, and other language integrated persistence APIs he saw the original approach being validated and coming back. He sees the future in a “distributed in-memory live object cache that has transactional attributes but it doesn’t deal with disk space storage what so ever.”

Christof Wittig from db4objects said that ODBMS have found their place as an embeddable component, be it in software-enabled devices or in SOA application silos. ORMs are only a “band aid” – good enough as a compromise -, but not a solution to all object persistence problems. Where appropriate, he said, developers should reclaim the data layer, and their open source object database db4o gave them the power to do so.

 

About ODBMS.ORG
ODBMS.ORG (www.odbms.org) is a vendor-independent, not-for-profit educational program on object database technology and the integration of object-oriented programming and databases. Designed by Prof. Roberto Zicari of Frankfurt University, the program’s goal is to promote and further the use of object databases – by offering free resources for students, faculty and researchers at universities and research centers, as well as for JAVA and .NET developers in the commercial and the open source world. Content is provided by a panel of internationally recognized experts, who share research articles and teaching materials with the community via the Web portal.

Press and analysts contact: editor@odbms.org

You may also like...