From: Stavan Parikh (stavan@cs.washington.edu)
Date: Mon May 03 2004 - 08:18:28 PDT
Ullman presents a look at the two different information integration systems - Information Manifold and Tsimmis - presenting their key similarities and differences.
To start of the author presents an overview of conjunctive queries and describes how views can represent queries. This was an extremely useful discussion that helped understand the later part of the paper. My main complain about this was that the author jumps into describing these parts without an introduction. While the concept of information integration systems might be known to people in the area, as a novice it was annoying to try and figure out what the goal of the paper was while reading the background.
The need for information integration systems is motivated by the need to combine information from various sources. The Information Manifold approach is the closest to the CQ basis. Data integration is carried out by generating a set of views and queries are then processed on a subset of these views. This makes addition of views easy and does not require rewriting of queries when new sources are added. In contrast in Tsimmis there is a mediator that exports data by combining the original data. These new sources are then queried. This requires that when new sources are added all affected data has to be recomputed.
One thing to note that other reviewers also pointed out was how the differences in IM and Tsimmis can be though of in terms of DBMS and XML. Tsimmis in particular almost seems like a precursor to XML just because of the way it uses semi-structured data.
In conclusion while this was a good first paper to read on the subject I didn't think the work itself was that useful. The author just summarized the main points of two different systems and compared them. There were no take away messages from the comparison or suggestions either way.
This archive was generated by hypermail 2.1.6 : Mon May 03 2004 - 08:18:29 PDT