Re: A Question on Integrety

From: Bob Badour <>
Date: Wed, 17 Dec 2003 10:43:11 -0500
Message-ID: <>

"Tobes (Breath)" <> wrote in message news:brphos$607ju$
> "Mike Sherrill" <> wrote in message
> > On 14 Dec 2003 14:04:52 -0800, (--CELKO--)
> > wrote:
> >
> > >>> How do I design a database for entities that are subject to
> > >different business rules throughout their lifecycle!?
> > [snip]
> > >I just sent in a three part article to on this topic. My
> > >recommendation was a series of "staging tables" which start off with
> > >the absolute nightmare of making all the columns VARCHAR(<max>) or
> > >NVARCHAR(<max>) and then move them in steps to other staging tables
> > >with more and more constraints and proper datatypes as I scrub the
> > >crap into some kind of usable shape.
> >
> > Some crap can't be scrubbed.
> In my limited experiece this is often true. I've often seen businesses
> to get at least *some* data, even if it's not entirely useful. The idea is
> then to come back and complete the data later. So, you have records that
> don't really conform to the "optimal" business rules, but at this time the
> business rules don't have to apply. At some point in time the business
> want to make use of this data, but perhaps only the items that validate
> against a certain set of rules.

One would ordinarily call that "restrict". It is one of the relational operations.

> From my viewpoint, it seems that the RDBMS
> is biased toward enforcing integrety on what goes in, rather than
> rules that change throughout an entities lifecycle.

Your viewpoint seems rather misinformed.

> Would it be totally
> ignorant/amateur of me to suggest that database tools should support the
> concept of entity lifecycle?

No, not at all. Database management encompasses all of database management. Thus far, the relational model best meets the need.

> > A lot of legal systems model "what is" differently from "what might
> > be". Think about your current address. Some people might know only
> > the state. (Or only "somewhere out west".) Some people might know
> > only the city and state. Some people might know only the street.
> > Some people might know only one or two digits of the house number.
> > Some people might be recalling the wrong address--the one you lived at
> > five years ago. (But that's still a fact in this context.)
> A few years ago I remember wandering, at what point does an online
> cart" become an "order". Could a cart be considered an order that is under
> construction? Or should the completed cart be "migrated" to an order?

It is mostly irrelevant with adequate physical and logical independence, and depends on the particular view an application sees.

Does it really matter all that much whether: "A = B UNION C" or "B = A WHERE p" and "C = A WHERE NOT p" ? Received on Wed Dec 17 2003 - 16:43:11 CET

Original text of this message