From: CR (chrisre@cs.washington.edu)
Date: Mon May 03 2004 - 08:59:58 PDT
This paper begins with an introduction to conjunctive query containment
and then shows how this applies to views and information integration. It
is a really beautiful application of theory. It is straightforward and
shows how much formalizing the problem can help. This really also shows
how databases take advantage of their strong theoretical model to make
practical improvements. Not only in terms of performance - though this
is certainly central - but also in terms of new functionality. This
paper deals with the difficult problem of heterogeneous source
integration through views.
Two systems are compared, Tsimmis and Information Manifold. Both
share direct roots with this theory. The IM architecture uses a query
synthesis strategy versus a rule based approach. Tsmimmis requires that
the data be organized into objects - to help alleviate the problem of
creating wrappers they claim that almost all of it can be generated.
What they are proposing as objects, even in their own words, seems to be
a precursor to semi-structured data.
I could not really get a feel for IM from this paper. However the
general approach of using indirection to hide the changes in the
underlying model is one of the central ideas from RDBMS
(logical/physical independence). The model is elegant because it allows
us to specify our indirection in a manner we are already familiar with -
queries.
I liked the presentation of this paper, especially the first section.
This archive was generated by hypermail 2.1.6 : Mon May 03 2004 - 09:00:00 PDT