Re: Sequence Numbers as Primary Keys
Date: 1995/09/24
Message-ID: <443tt8$r3r_at_news4.digex.net>#1/1
^^^^^^^ (jhagans_at_telerama.lm.com) wrote:
: Recently at work ,there has been a controversy between the application development
: group (IT) and the database administration group (DBA) concerning using sequence
: numbers as primary keys and hence as foreign keys in the appropriate table.
:
: The IT group claims that their C++ program will be much more efficient with
fixed-length
: unique identifiers.
:
: The DBA group refuses to use sequence numbers claiming that they are not
"business
: data" and do not belong in the database.
:
: The database has probably around 50 tables with an estimated 6 million hits a day;
: However it is only a few tables (7 or less) that will be getting the majority of the hits. The
: composite key in these tables is quite large, usually 5 or 6 columns to make the row
: unique.
:
: We are running Oracle 7 on a RS/6000 Unix box.
:
: Any insight to this matter would be greatly appreciated.
As nice as it is to try to stick to all the classic design approaches (no unnormalized data, no extraneous data, etc) the real world doesn't always let you do that.
However, the program itself should not define the data strucutre either. The relationship of the data, espeically between their foreign keys, should drive the decision.
Since you state that if sequence #'s arent used, then 5 or 6 columns comprise the key. This makes some things very difficult to work with (in any language) as well as harder to maintain. ("Which fields make up the key?")
My opinion here (flame=ON!) is that the DBA group is out of line in maintaining such a hard stance. The database has to be *useful* or else it doesn't matter whether it only has "business data" or not.
Joe Nardone
:
: Joel Hagans
: jhagans_at_telerama.lm.com
:
--
=------------------------------------------------------------------------=
Joe Nardone |
joe_at_access.digex.net |
Received on Sun Sep 24 1995 - 00:00:00 CEST