I have used many third party apps with single code tables and they have caused me many nightmares, performance being the least of them. The performance issue we solved by having 1 records per block reducing the hot block contention. One nightmare that I remember was pretty simple a developer decided to reorder a drop down list on the front end, unfortunately the application tool changed the codes values to accommodate this change and since the app enforced referential integrity, all 4 million members now lived in a different state and I had to clean it up, based on zip code for the members who had them (90%) and based on street and city for the rest of the records that didn't have them. Ken _____ From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Dennis Williams Sent: Tuesday, November 21, 2006 4:30 PM To: jkstill@xxxxxxxxx Cc: paulastankus@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx Subject: Re: Physical Database Design - Code Tables Paula, One issue I don't see mentioned is the number of columns. In my experience, some codes tables have a different number of columns than others, and those columns may have different properties (data type, column name, etc.). If you use a single codes table, then all codes must have the same columns. If at a later point, some codes are found to need an additional column, then all codes will get the added column. If you add that column, then you have to hope that all the application code will be unaffected. This isn't probably a big deal, but I think it points to the non-normalized situation that a shared code table gets you into. Dennis Williams