Re: A Question on Integrety

"Mike Sherrill" <MSherrillnonono_at_compuserve.com> wrote in message news:t336uv8jpu7cl2oggbqm1j5p8j3a0lovjs_at_4ax.com...
> On Wed, 17 Dec 2003 12:21:45 -0000, "Tobes \(Breath\)"
> <tobin_dont_spam_me_at_breathemail.net> wrote:
>
> >In my limited experiece this is often true. I've often seen businesses
want
> >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.
>
> IME, the idea never bears fruit. Bad data never becomes good data
> without a *lot* of effort, and the effort is almost wasted when you
> put the good data into a database where "anything goes (in)".
>

This statement seems analogous to something like "it makes no sense to make a draft of a paper using word processing software before finalizing it". Word gives the option to the use to use the spell-check or grammar checking features.

A good implementation of a DBMS should provide similar features. As I understand it, in some implementations of an RDBMS (or perhaps only in Tutorial D) there is the ability to implement local constraints rather than only global constraints. If local constraints including data type and size are possible, then do not have all such constraints on data that is tagged (in an attribute) as a draft.

Include the draft number (effectively a version number for the data) as an attribute that is a partial candidate key. Then if you really can implement local constraints, place more constraints on subsequent versions of the data, so that the final version includes all data constraints.

One of the problems with the relational databases I have worked with was that I never saw or used local constraints, so all constraints were considered global, which does not resemble real world constraints.

The spell check is required before a draft of the document that is going out to readers, for example, but not before then (in the analogy).

--dawn

[remainder of original deleted] Received on Sat Dec 20 2003 - 15:42:51 CET