Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!wuarchive!sdd.hp.com!caen!uwm.edu!linac!att!bellcore!porthos!taichi!haim
From: haim@taichi.uucp (24122-Haim Kilov(L028)m000)
Newsgroups: comp.object
Subject: Re: Object Oriented Database Design
Message-ID: <1991Apr28.052610.5317@porthos.cc.bellcore.com>
Date: 28 Apr 91 05:26:10 GMT
References: <1991Apr22.073328.7858@ira.uka.de>
Sender: netnews@porthos.cc.bellcore.com (USENET System Software)
Reply-To: haim@taichi.UUCP (24122-Haim Kilov)
Organization: Bellcore, Livingston, NJ
Lines: 30

In reply to:
The problem with object oriented modelling is the integration of
  structural and behavioral properties. That is why classical semantic
  data models, i.e., the Entity Relationship Model, cannot be used for
  object oriented design.

--the first sentence is correct, whereas the second is not. ER model, 
actually, deals with behavioral properties, although they are very simple
ones: after all, consider the precise definitions of such concepts as
"relationship", "dependent", "composite", etc. These definitions will be based
on primitive (CRUD) behavior of the particular entity and associated entities,
i.e., on operations jointly owned by these "clusters" of entities.
I have published a paper in TOOLS'91 on this topic, and we have also 
published an extended Bellcore Special Report about this (I mentioned this
some time ago in comp.object).

On the other problem -- graphical representation of object classes -- I can
only say that it is difficult: the ER approach provides only a partial answer, 
and you still need different graphical representations for different kinds
(meta-types) of entities. Still, it's a reasonably good approach.

The problem of more application-specific behavior is more difficult. Are you
absolutely sure that a graphical representation will help? After all, I agree
with Bertrand Meyer: A picture is worth thousand words, but did you try to
find seven errors in a picture?

Hope this helps.

-Haim Kilov
haim@bcr.cc.bellcore.com