Niall Litchfield wrote:
> "Mark C. Stock" <mcstockX_at_Xenquery .com> wrote in message
> news:SbSdnR0ccbBME6rcRVn-sQ_at_comcast.com...
>
>>so, if you're offended, sorry. but if you're up for a lively discussion,
>
> why
>
>>not post a structure that you feel requires more than one column for the
>>PK -- either we'll see each other's side better, or you'll have the
>>satisfaction of proving me wrong ;-)
>
>
> A possibility for you. We have an helpdesk app. It contains, not entirely
> surprisingly, an inventory table. That table contains the fields MODEL and
> SERIAL#, we use (MODEL,SERIAL#) as the PK. I'm not prepared to guarantee
> uniqueness of the serial# across manufacturers and over time. I am prepared
> to gurantee the uniqueness of SERIAL# across the same model from the same
> manufacturer over time. I'll grant you that we could use an artificial key
> (and in fact it would be my preference), never the less for the design of
> the app the important thing is that it is an item of inventory (not that it
> is a laptop made by manufacturer y or a router made by company Z).
You do understand, I presume, that if it was done your way you could
easily get duplicate model, serial#s into your table or be forced to
create a unique constraint (with its index) do duplicate exactly what
it is that your PK does for you now.
--
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace 'x' with 'u' to respond)
Received on Sun Sep 05 2004 - 17:44:39 CDT
