From: Neva Cherniavsky (nchernia@cs.washington.edu)
Date: Mon May 03 2004 - 11:26:07 PDT
This paper describes the theory behind conjunctive queries as background,
and then goes on to describe the information integration problem.
Information integration is hard because of legacy databases (cannot alter
the database), different meanings in databases that are closely related,
and time-varying schema for web-based data. Mediators are used to solve
this problem by communicating between different sources. There are two
types of mediator research projects the author describes: information
manifold and Tsimmis.
In information manifold, all queries are expressed in terms of global
predicates. Each information source is associated with constraints and at
least one view, but views may give only partial information. The solution
to the query is the union of all minimal CQs.
Tsimmis creates a hierarchy of wrappers and mediators that talk to one
another. Components communicate via a special data model and query
language. The model is object oriented.
I found this paper quite difficult to read. Because the author
immediately dives into background, for a long time I had no idea what the
purpose of the paper was. Once I understood what he was comparing, it was
still difficult to understand the importance of the systems, or why he
chose those two in particular to examine. I think this is a good
background paper, but I would have preferred to know before I started why
I was reading the paper and what knowledge he was trying to impart.
This archive was generated by hypermail 2.1.6 : Mon May 03 2004 - 11:26:08 PDT