On 14 Oct 2004, wizofoz2k_at_yahoo.com.au wrote:
> "Mark C. Stock" <mcstockX_at_Xenquery .com> wrote in message
>> My comments were specific to the data modeling phase
>> (conceptual model) but discovery does continue during initial
>> implementation as well as ongoing use of the data. No analyst,
>> and no data model, perfectly captures every nuance of a
>> business at the first take -- new discoveries typically
>> clarify earlier incorrect, arbitrary or misconceived ideas.
>
> There is a recommendation then to use surrogate keys if I ever
> saw one. If you can't be 100% sure you have discovered all the
> nuances and possible relationships, how can you then define
> your tables with a potentialy complex relationship tree with
> natural keys? How can you now use PKs based on those natural
> keys in extensive FK/PK trees? What will happen to the FKs
> first time you need to change the PK because your analysis
> discovered a hidden natural key?
Here here. You explain my thoughts much better as I disagreed
with Daniel in another posting as he counters with not starting
until specs are signed off on. Even if one were to have signed
off specs the data architect should be flogged if his design
doesn't allow for corrections and additions to it. One should
most definitely be able to start the project without knowing
everything. One should most definitely architect a solution
which is able support new discovery during the lifecycle of
design, development and production deployment and also after it
is in production.
--
Galen Boyer
Received on Thu Oct 14 2004 - 07:43:11 CDT
