[AccessD] OT: The Great Primary Debate

Hi Ken

> That's why I dumped autonumber and went with a Modified Julian Date PK field
> for my date dimension table. It meets the three basic requirements for a PK:
> not null, unique, and won't change. Plus, the MJD is trivially easy to
> calculate from any date; you don't have to lookup the PK from the date
> table.

> I wrote a long, boring post on this some time ago, which elicited a
> tremendous yawn from the AccessD community. This post is much shorter, and
> thus should generate only a tiny yawn :-o

Ahh, Ken, not so sensitive - at least you didn't get flamed, and - for
what credit it is worth - I didn't delete your message on "Using
Modified Julian Days" - it is there should I need it. However, I'm not
doing any data warehousing and have never felt the need for date
tables, though I can see the possible advantages of those for some
scenarios.

Having said that, I'm convinced that using anything else than a
meaningless PK should be reserved for very special cases.

/gustav


> In your data dimension table example, you are creating a meaningful field
> SequentialDateNumber (which you are calling ID) and are using it in data
> calculations.

> Thr real question in this situation is not whether you use this natural key as
> a PK, but whether you have a PK in the table at all - which comes down to the
> sub-debate about "what is a PK and what is it used for" :-)


















> --
> Lexacorp Ltd
> http://www.lexacorp.com.pg
> Information Technology Consultancy, Software Development,System Support.