Re: UML vs. ER modelling (long)

From: Silvio Bierman <sbierman_at_idfix.nl>
Date: Sun, 29 Dec 2002 18:03:45 +0100
Message-ID: <3e0f2ac8$0$138$e4fe514c_at_news.xs4all.nl>


Interesting discussion. Opinions seem to differ on this a lot.

Personally I prefer doing a solid information analysis resulting in an ER scheme representing the strictly data-oriented part of the database. The process analysis could be done using UML resulting in class- object- and activitymodels. Mapping the UML models onto the ER model is something that can be handled in the business layer. Usually there is need for process-support data persistance provided by the database as well, but this can be implemented as a separate addition to the database scheme extracted from the ER model.

I have seen many projects where a havy emphasys on process modelling results in a database scheme that is to much oriented towards process support, leading to semi-normalized structure and poor queryability. Going in from two separate directions can prevent this.

Bottom line I think that there is a principal difference in the way data is handled by a process (dynamic data handling) and the way it should be represented in a database (static data storage) where it can be accessed by other processes/tools(queries). In all but trivial systems this difference is to important to ignore. That is also part of the reason I am not a blind believer in object oriented databases and automated object-relational mapping provided by techniques like EJB-CMP/JDO.

Just my two euro-cents...

Silvio Bierman Received on Sun Dec 29 2002 - 18:03:45 CET

Original text of this message