Re: normalization question and one to one relations
- From: "Rick Brandt" <rickbrandt2@xxxxxxxxxxx>
- Date: Thu, 31 May 2007 08:03:44 -0500
giladp1@xxxxxxxxx wrote:
Thanks so much to all the replies.
I feel better now to leave things as they are without adding extra
tables.
Rick, thanks greatly for taking the time to answer me.
As for your first reason, you didn't say why half a dozen or more
fields that belong to an identified subject would likely push you to a
second table. Is it a performance issue for example?
Nope. Having a table with 15 fields where 10 of them are mostly empty just
doesn't (to me) look like a table that properly models the real world entity
that the table should represent. It's not a black and white issue though.
Much would depend on what the tables held, how the data was used, etc..
As for your second comment, I saw the same situation in another big
organization a couple of years ago. A university actualy. they told me
they spent millions and ten years listing contstraints in their
database, and were afraid to implement new technology for fear that
all this investment will be lost. I don't realy understand that. isn't
listing a contsraint in SQL Server a matter of a couple of seconds? If
all the logic is already there you only need to translate it to the
new technology, why not move on to SQL Server or to Oracle?
I can't speak to that concern, but on the surface I agree that it sounds
like a weak reason to resist change. The legacy system I am talking about
uses "tables" in a relational database (UDB400), but it doesn't use them as
proper relational tables. They are used as "files" would have been under
older main frame type systems. The programs (Mostly RPG) that use these
tables do not (generally) use SQL or any other traditional database access
mechanisms and all objects are very highly interdependent on each other
right down to the "last design change".
For example, objects on the system have a "level" value that is a number
assigned when it is compiled and all RPG programs perform a "level check"
when executed. Any object that has been altered since the program was
compiled will cause the program to fail until it is recompiled. Of course
recompiling that program changes its level triggering a domino affect of
programs and other objects that all have to be identified and recompiled
before the system runs again. Changes can be done of course and they are,
but only with a lot more trouble, time, and expense, so they are not taken
as lightly as they would be in a more database-oriented system like we are
used to.
--
Rick Brandt, Microsoft Access MVP
Email (as appropriate) to...
RBrandt at Hunter dot com
.
- References:
- normalization question and one to one relations
- From: giladp1
- Re: normalization question and one to one relations
- From: Rick Brandt
- Re: normalization question and one to one relations
- From: giladp1
- normalization question and one to one relations
- Prev by Date: Re: Unique Client ID
- Next by Date: Re: Tracking Notes
- Previous by thread: Re: normalization question and one to one relations
- Next by thread: Converting MS Access to WEB
- Index(es):
Relevant Pages
|
Loading