Code Forums - View Single Post

technobard · 06-22-2002, 07:08 PM

You're right. Not all apps have to use databases. Go beyond a certain size, however, and most do. The value of using any kind of modeling increases with the size of the project. It's a lot of overhead for smaller projects. And whether you intended to or not, you just proved my point.

People who are trained to model data see a problem with UML. Most developers usually don't. That's not a criticism, just an observation.

The whole point of UML was to model software not data. It was designed by three computer scientists, object modelists specifically, who tried to take the best of their respective object oriented methodologies and come up with a "unified" solution...as in Unified Modeling Language. (i.e. Booch, Jacobson, and ?)

I've worked on a couple of decent size projects as a data architect that used UML. In one case, the software architect didn't get it. It was extremely difficult trying to reconcile the data modeling world with the programming-centric world of UML. Before I could make any real headway, I was pulled onto a higher priority project for 12 weeks. When I went back to the original project, they had ignored all of the things I told them not to do. None of it was rocket science, but things were screwed up. I was pissed, because I had to fix them.

The second project, the 12 week design gig, was exactly the opposite. We used UML, but the software architect had an open mind. We managed to work out a collaboration that made for a great design process. In short, we found a way to address the data needs outside of the UML, while making sure that the class diagram mapped back to the final physical database design.

I've done a lot of database design and administration, and I've done enough of the development side of things to be dangerous. If you're working in the object oriented world, learning UML is probably not a bad idea. A lot of corporations use Rational Rose and will follow a Rational Rose development methodology. Just don't believe that it solves all of the modeling needs on a project, because it doesn't. That's not just my opinion. There are a lot of people with a lot of experience in the field who would agree with me.

Now stepping off soap box.

06-22-2002, 07:08 PM	#10 (permalink)
technobard Centurion Nova Prime Join Date: May 2002 Location: Oak Park, IL (USA) Posts: 287	You're right. Not all apps have to use databases. Go beyond a certain size, however, and most do. The value of using any kind of modeling increases with the size of the project. It's a lot of overhead for smaller projects. And whether you intended to or not, you just proved my point. People who are trained to model data see a problem with UML. Most developers usually don't. That's not a criticism, just an observation. The whole point of UML was to model software not data. It was designed by three computer scientists, object modelists specifically, who tried to take the best of their respective object oriented methodologies and come up with a "unified" solution...as in Unified Modeling Language. (i.e. Booch, Jacobson, and ?) I've worked on a couple of decent size projects as a data architect that used UML. In one case, the software architect didn't get it. It was extremely difficult trying to reconcile the data modeling world with the programming-centric world of UML. Before I could make any real headway, I was pulled onto a higher priority project for 12 weeks. When I went back to the original project, they had ignored all of the things I told them not to do. None of it was rocket science, but things were screwed up. I was pissed, because I had to fix them. The second project, the 12 week design gig, was exactly the opposite. We used UML, but the software architect had an open mind. We managed to work out a collaboration that made for a great design process. In short, we found a way to address the data needs outside of the UML, while making sure that the class diagram mapped back to the final physical database design. I've done a lot of database design and administration, and I've done enough of the development side of things to be dangerous. If you're working in the object oriented world, learning UML is probably not a bad idea. A lot of corporations use Rational Rose and will follow a Rational Rose development methodology. Just don't believe that it solves all of the modeling needs on a project, because it doesn't. That's not just my opinion. There are a lot of people with a lot of experience in the field who would agree with me. Now stepping off soap box.