Re: OO and relation "impedance mismatch"

From: Kenneth Downs <firstinit.lastname_at_lastnameplusfam.net>
Date: Wed, 06 Oct 2004 18:48:46 -0400
Message-ID: <fos1kc.6fc.ln_at_mercury.downsfam.net>


Laconic2 wrote:

>
> "Kenneth Downs" <firstinit.lastname_at_lastnameplusfam.net> wrote in message
> news:fac1kc.83b.ln_at_mercury.downsfam.net...
>

>> Here is an example of how different they are.  Question: if you were
>> designing a dd, and had complete freedom to do it anyway you wanted, how
>> would you specify a foreign key?  Well, we always start with the absolute
>> simplest way and then embellish, so we get:

>
> Well, if I had complete freedom, I'd substitute one of those fancy data
> modeling tools like Erwin or Data Architect for the DD, to begin with.
>
> DA already maintains three different kinds of models, a Conceptual Data
> Model, a Physical Data Model, and an OO Case Model. Starting from any
> of
> them, you can derive the other two.

Very tough call, buy or build? I don't like to be dependent upon a third party, which is why I'd like to be in control of the authoritative format for the dd. It has also seemed to me that the makers of these tools are usually trying to accommodate models such as ERD that I tend to ignore, owing to the fact that "People Understand Tables Just Fine." I think in terms of tables and want to specify things in terms of tables. And because I can type, I have no reason to use a GUI to make pictures of tables whose essential information I'm just going to have to type anyway.

But they do make really nice pictures, which it would be extremely silly to re-implement.

So I haven't figured out yet how to relate to them. I have always assumed an import/export system.

>
> Then I'd add whatever features necessary to automate the trigger
> generation
> you were talking about, together with suppressing the inclusion of the IR
> constraints in the DDL. That way I could have my cake and eat it too.
> The
> biz rules, including Integrity rules, would be in "one master place",
> but I could choose which of them got enforced in the DBMS.
>

This is the fun part, but it is also where you spend most of your time as the toolmaker. This is another reason though why I tend to ignore the GUI analysis tools, because writing the code generator and DDL generator is the real cool stuff, and frankly I just type the dd specs into text files, because that's the fastest way to get them created.

When I have it ready for others to use though, it will be pretty important to allow GUI editing, so a bridge to freely available open source tools will become much more important.

> Next, I'd add support for some network and hierarchical DBMS products.
> Maybe throw in the Nelson-Pick model to see if it's as great as it's
> touted by the Pickies in here.

Hmmm, had no plans, I'm kind of SQL bigoted. But since it will be GPL, perhaps you can add that?

-- 
Kenneth Downs
Use first initial plus last name at last name plus literal "fam.net" to
email me
Received on Thu Oct 07 2004 - 00:48:46 CEST

Original text of this message