DBMS, November 1997
I'm a data warehouse modeling consultant in Frankfurt/Main, Germany. Most of the issues in Ralph Kimball's August 1997 DBMS column, "Data Warehouse Role Models", are easy to understand and explained quite well. However, I have one question about an embedded location table. It is a "master location table." Why is the location table part of a Working Telephone Number, Billing Telephone Number, and Equipment Inventory? Does the master location table join with the Working Telephone Number Table or are the values within the table?
Kay zum Felde
KZF@CompuNet.de
The main requirement of a table playing a role model (like Location plays in the telco example, where it takes on the roles of Working Telephone Number, Billing Telephone Number, and Equipment Inventory) is that the single physical Location table is logically copied as many times as necessary. That means it is not joined into the schema in each of these locations. Joining would mean that all these locations would be identical in a given query, which may be nonsense. Instead, you must create SQL synonyms and then relabel each of these synonyms with views, as described in the article.
-- Ralph Kimball
I am a data warehousing consultant with Mastech Systems Corp. I have recently become involved in a data warehousing project where my role is to convert an RDBMS into a multidimensional database (MDDB) and apply OLAP methodology to the multidimensional database, and then use Cognos' PowerPlay to build PowerCubes. It was interesting to read Ralph Kimball's August 1997 DBMS article, "A Dimensional Modeling Manifesto,". It not only helped me to get more information, but also helped me to justify my design to the management. I thank you.
Sanjay KM
sanjaykm@usa.net
In Joe Celko's very interesting article about the Year 2000 problem (see "Year 2000 -- Threat or Menace?" DBMS, July 1997), he referred to the four-digit year as CCYY. Technically, this is inaccurate. It should be (CC-1)YY, if CC refers to the century. We are in the 20th century even though years begin with 19. The Oracle RDBMS, for example, returns 20 when using the CC format for a 1997 date.
Irv Cantor
icantor@prius.jnj.com
I want to thank Daniel Morgan for pointing out such a glaring mistake on Microsoft's part. (See "ODBC Security Compromised," DBMS, September 1997.) Where was the project manager? What were they thinking? There have inherently been suspect areas for snooping in ODBC, but to have the manufacturer supply such an open door is flabbergasting. At my company, we blindly installed SP1 for ODBC 3.0 three months ago so our recent tracing didn't uncover the problem. The KB article Q168879 is indeed scary. Anyone walking around with an original 3.0 SDK can glean the connection data. I did. Our current apps planned for client hacking with a SQL Server stored procedure permission scheme to all objects, but I've written for plenty that didn't.
John Beatty
johnb@intercomm.com