review

From: Alexander Moshchuk (anm@cs.washington.edu)
Date: Mon Apr 26 2004 - 01:19:03 PDT

  • Next message: CR: "Of Objects and Databases: A Decade of Turmoil"

    Of Objects and Databases: A Decade of Turmoil

    The paper provides a survey of the work related to incorporating
    object-oriented support into traditional database systems. The
    authors identify four main approaches to this problem, and tell us how
    well the approaches fared during the decade before the paper's
    publication.

    In extended RDBMS, known as object-relational databases today, we
    start with a traditional RDMBS and SQL and add support for objects to
    both of these. The objects can be defined as user-defined ADTs along
    with an implementation in an external language, or row types, an
    extension for tuples. The authors advocate this approach as the
    winner among the rest - major commercial vendors are pushing in this
    direction because they can extend their existing RDBMS technology
    rather than commit to a new system.

    The other three approaches have been less successful. Unlike the
    previous one, they all involve adopting a new system. Persistent
    programming languages add features to support data persistence and
    atomic program execution. Although active in academia, they don't
    have any commercial implementations, leading authors to declare it as
    a dead-end. Another casualty is the database system toolkit approach.
    Here, similarly to the idea of microkernels for OS design, we keep a
    set of core DMBS facilities which can be extended at any level to
    provide domain-specific functionality. This proved to be inflexible,
    awkward, and too hard to use. The final approach - Object-Oriented
    Database Systems - basically redesigns a DMBS in an object-oriented
    fashion; this can be viewed as essentially adding database
    functionality into an OO language. OODB failed to live up to
    commercial expectations, with problems like no agreement on standards
    and some functionality that lags behind RDBMS.

    I thought the paper was interesting to read, and all of the arguments
    seemed plausible. The paper assumes an understanding of the outlined
    approaches, so sometimes I felt they could have given more background
    for those not as familiar with the topic. I don't have enough DB
    knowledge to tell if the authors' predictions for the next ten years
    indeed came true, but it seems ORDB hasn't yet taken over the world.
    In any case, the authors obviously missed the emerging importance of
    the internet and related services in their vision of the future.
    Also, they wrote their evaluation from a highly commercial standpoint,
    where success of an idea ultimately depends only on a major vendor
    implementing the idea.


  • Next message: CR: "Of Objects and Databases: A Decade of Turmoil"

    This archive was generated by hypermail 2.1.6 : Mon Apr 26 2004 - 01:17:45 PDT