Re: Just one more anecdote

From: Kenneth Downs <knode.wants.this_at_see.sigblock>
Date: Tue, 02 Aug 2005 16:32:23 -0400
Message-Id: <0p15s2-5q2.ln1@pluto.downsfam.net>

dawn wrote:

> Marshall Spight wrote:

>> Kenneth Downs wrote:
>> >
>> > ERD has never appealed to me because it seemed to be trying to make
>> > something easy that was in fact already easy.  Because it was trying to
>> > make something more easy, it had to introduce elements that masked
>> > reality,
>> > such as a M:M relationship that masks a cross-reference table.  What's
>> > the
>> > point?  The x-ref itself is sometimes useful for direct querying, so
>> > why not show it?
>>
>> Great post! This does a bang-up job of articulating something I've
>> been feeling but unable to put into words for a while.

>
> I think this provides support for what I suggested in another post --
> that people will often create what someone might call a conceptual
> model or an analysis model, but where they are presupposing an
> implementation model when preparing it.
>
> Perhaps you and Kenneth would not call it a conceptual model, but what
> you are then suggesting is that you skip the step where you model the
> problem domain and go right for a relational design.

The step in question is not a valid step, and the insertion of this invalid step should be avoided. Speaking for myself, I simply proceed from brainstorming to the next correct step, which is table design.

>
> Just on the off chance that perhaps someday you will want to do an
> implementation with a different model than the one you currently use, I
> suggest that it really is a good idea to aim first to prepare a model
> without regard for the target implementation. Just a suggestion, of
> course.
>

That sounds wise, but it fails to inspire me. Since you can store anything in tabular data, why not specify it as such? If the past decades gave any suggestion that the Quasi-relational SQL products out there were a stop-along-the-way technology, I might fear the reaper, but to all appearances they are a long-term stable implementation platform.

But I can have my cake and eat it too. When somebody introduces a post-relational data model that gives all of the power of RM plus heretofore unknown nifty features, I'll just write some SQL to convert my dictionaries to that new format. Can you yet see why we're not worried?

-- 
Kenneth Downs
Secure Data Software, Inc.
(Ken)nneth@(Sec)ure(Dat)a(.com)

Received on Tue Aug 02 2005 - 15:32:23 CDT