Re: Proposal: 6NF

David Cressey wrote:

> Paul,

> You either have a systematic way of dealing with missing data, or you deal
> with missing data in an unsystematic way, when the eventuality happens.
> Which is worse?
There is a systematic way of dealing with problem of missing data proposed by Darwen (Mad a typo it was not Mc Goveran). It consists of maintaining a repository that keep track of attempts of inserting missing data into tables then be able to *bring them back* into the table where they were supposed to be inserted, in case they get updated to a known value. I have to admit it creates an extra layer of administration as you have to maintain the repository, but this is nothing compared to work that has to be done when letting NULLS in. It is systematic and most of all insures accurateness. Accepting NULLS certainly does NOT bring systematic way of dealing with missing data. Quite frankly, I do not see HOW adding subjectivity into the system by interpretating results can be in any way *systematic*.

>
> I didn't intend the inference that you drew, that a company should design
> its data strategy around dependency on missing data. I do think that a
> system that a company depends on should not choke up and refuse to work if
> some item of expected data is not present.
>
> I also don't believe that "it's better manually". Manual systems also have
> ways of dealing with missing data. Sometimes they depend on "common sense"
> on the part of the people in the system. Sometimes that works. Sometimes
> it doesn't. I think a good, well designed system that incorporates a
> systematic way of dealing with missing data can be significantly more robust
> that a system that relies on uncommon levels of common sense.
Received on Wed Oct 04 2006 - 10:35:01 CEST