Re: UML vs. ER modelling (long)

Hi again,

first of all let me thank all of you guys out there for your prompt and most welcome replies.

My remarks to your observations follow, in no particular order:

USDP is the Unified Software Development Process, devised by the inventors of UML (the "three amigos"). The "bible" is http://www.amazon.co.uk/exec/obidos/ASIN/0201571692/ref=pd_sim_b_dp/202-8196628-5683868

I'll examine naked objects (an open source project BTW, even more appealing!) as soon as I find some time. Thanks for the pointer Stephan.

I entirely agree with Silvio when he says that "Mapping the UML models onto the ER model is something that can be handled in the business layer." I could have written this sentence! Actually, this is precisely what I meant when I was talking about using UML only where it's really needed.

I also completely agree with AlexV about "If there is no java class for some tables, than it is NOT business domain and not UML."

Again, I agree with Silvio in not believing in automated object-relational mapping. I don't like CMP and I will not use it in this century! Just kidding of course, but I think the technology for an automated such mapping is definitely not mature yet.

I do not agree with AlexV when he says that "99.99% of applications are data-centric and data-model will be more complicated than business model." While this may be the case for business information systems
(assumed we mean the same thing by that), it is definitely not the
case for other systems e.g., text editors or UML editors. In these cases, business logic is prevalent wrt persistency (which in this case is limited to just opening a file and writing your text into it).

In a perfect world, AlexV is right saying that UML and ER should not overlap. In this world, however, I'm afraid they will. I do agree with him, however, about his points i) ii) and iii)

One point that emerges from this very interesting discussion (and one that we all seem to agree upon) is that a good, robust, "native"
(i.e., not UML) ER model is key to the whole project.

All the discussion simply confirms that my initial choice (undeclared to avoid biasing the audience in one way or another) was right: "Wouldn't it be better to just model those directly in E-R and use UML only where it's really needed?" Yes, it definitely would!

Thanks again,
andy Received on Mon Dec 30 2002 - 23:47:26 CET