From: Bhushan Mandhani (bhushan@cs.washington.edu)
Date: Mon Apr 26 2004 - 11:41:39 PDT
Of Objects and Databases: A Decade of Turmoil
- M. Carey and D. DeWitt
This paper describes where the field of "Objects and Databases" was in
1986, compares it with the state in 1996 (when the paper was written), and
tried to predict how things would evolve over the next 10 years.
For each of the four different lines of thought being pursued in the
beginning, the paper attempts to explain why they have evolved the way
they have. It attributes the death of database toolkits to they requiring
too much expertise to use, and being inefficient. It also mentions that
the persistent PL's approach never took off commercially. It spends much
more time talking about the other two approaches: OODB's and ORDBMS's.
Despite a considerable amount of research effort put in, OODB's weren't
the (commercial) success many people expected them to be. This is
attributed to the lack of standards, the inferiority of OODB products, and
the inferiority of application development support for them. I feel one
reason whose importance is underestimated is the fact that when they came
along, relational DB's had already been around for some time, and were
being adopted on a large scale, and at that point the OODB products
weren't mature enough to change this trend.
It then describes object-relational DB's, which were the clear winner
among the four. It is easy to see why. They extend the familiar relational
model, and offer support for objects of new complex types, as OODB systems
do. Further, with SQL99 standardizing many of their features, we would
expect them to become the dominant technology, which is exactly what the
authors predict.
The authors further predict that high-function highly integrated clients
will become common. They identify the client/server interface, and the
functionality of the server and the client as important directions for
research. I guess, this is a trend that we have seen followed. They also
talk about the need of ways to continue to access legacy data sources.
One thing they miss completely in their vision is data exchange and the
Internet. This is surprising, considering the paper was written in 96.
Further, they give a very high-level view of the different technologies,
and use only commercial success as the metric, ignoring the possibility of
interesting technical ideas even in approaches which haven't been
commercially successful. Otherwise, the paper is well-written.
This archive was generated by hypermail 2.1.6 : Mon Apr 26 2004 - 11:41:41 PDT