Chris Haddad has been interviewed on MDD (the generic form of the MDA drug). He discusses the holy grail of MDD, how it fits into SOA (if at all), whether you can gain code reuse, and finally some advice.

What advice do you have for developers who are starting to look at MDD?

Focus on finding tools that have an open architecture, especially in the generation and implementation of code artifacts. [For example,] you should be able to generate unit tests and you should be able to generate not only source code but also deployment descriptors and documentation. The generation aspect is a huge win. You can use MDA to basically remove the monkey-coding that you're otherwise sending offshore. And I think a lot of people who are looking at offshore strategies as a cost savings reduction should also be looking at MDA and MDD and seeing the big cost and efficiency savings that can be made there.

Read The Holy Grail of model-driven development

In related news: New Novell SOA suite supports Linux

Novell Inc. announced this week at LinuxWorld Conference and Expo the availability of Novell exteNd 5.2, a service-oriented architecture suite that supports Linux from the desktop to the server enabling design and deployment of Web services applications.

Hello,

here my 2Cents.

From the interview:

They aren't following the MDA specification.

I'm not quite shure what Chris Haddad means by "specification". The fact that the OMG prescribes that model must be expressed in UML, stored in MOF, and interchanged using XMI? The really important questions of MDD are not (yet?)answered by MDA, that is: Transformation, Model refactoring, Debugging on model level, etc). Even code generation (at the end of the day I want to have code, not models) is only a very, very small part within that "specification".

From the interview:

Also, you need to realize that UML itself doesn't go far enough. You really need to tailor it with your own domain concepts and models. A lot of people really use UML out of the box when they really should be customizing it first to provide a framework that targets the type of applications that they're building.

In fact, you should not be stuck to UML at all. If you have concepts in your domain that can not be modelled with UML, create your own DSL. Further, MDD tools should not rely on UML (or MOF, or XMI, whatever), rather they should be open to accept any textual language.


Best regards,
    Dirk

Talking about the Holy Grail is a good metaphor. Lot’s of people were looking for and nobody did find it. Looking back with nowadays experience we smile about it. The Holy Grail simply is a myth. Is building a complex application jointly with “modeling” analysts, architects and programmers. It sounds great. I don’t think I am pessimistic. I only have the chance to look back and think of many fashions that came up and after a little while people acted like politicians, forgetting easily what they said yesterday.
- What is different with MDD? What is better than all the other attempts like business languages like cobol or sql, case, 4GL, ...
- How do You want to teach big teams interoperating language?
- Business analysts often tend to talk imprecisely. Why should they change?
- Abstract approaches require a big team of people thinking abstractly. What will You do with the good old programmers and the easy times spent coding?
Sometimes code is the only truth. Abstractions help to get an overview, but certain details easily get lost. For me MDD is a new escape-strategy. Find a interesting technology that’s complex and sophisticated. Sell it as the Holy Grail. But what, if real people in real projects can’t handle it? Kick off the people.
Don’t You think it could be better to invest some energy in finding the requirements – functional and especially the non-functional ones, which very often are the keys to success or failure – and to concentrate on good code that fit these requirements.

Martin Prischmann, codecentric - http://www.codecentric.de

has anyone seen any research or whitepapers comparing a semantic web technology (for example, an owl or rdf based ontology using jena) and a uml centric mda approach? there are arguably better inferencing tools and a wider audience support for the w3c's semantic web approach.

has anyone seen any research or whitepapers comparing a semantic web technology (for example, an owl or rdf based ontology using jena) and a uml centric mda approach? there are arguably better inferencing tools and a wider audience support for the w3c's semantic web approach.

That's my impression, too.

I think the beauty of model-driven development shines brightest when combined with agile development.

The whole point to me of MDD is to get things done quickly, responding to changes in scope, requirements etc. This is the whole focus of agile development - to encourage change and go with what you know so you can see tangible software quickly. I think the bulk of "time" spent on projects (outside of debugging Weblogic or JBoss or your third-party library of the week) is in understanding what you're building.

Think of the value you get out of POCs and prototypes... I think this is whole point?

Some good info on all this at http://www.agilemodeling.com with links to books and the like.

mike, logicalapps, http://www.logicalapps.com

Haddad: The Holy Grail is to allow business analysts to directly contribute to a model, and have architects and developers contribute to the model. For instance, you would have platform and infrastructure developers create models of the framework APIs that those business developers [could then use] to create business applications. That really is the Holy Grail, especially when you talk about Business Process Modeling and the ability for business people to do workflow-based process coordination using tools.

There are already products that do this. I used to work at this one: http://dharbor.com/products/pfs_feat.html Since then I have done a few projects using their stuff. I can't imagine ever writing all the code, descriptors, etc anymore. UML modelling with wizards, including biz process workflows, importation of legacy schemas. Start the server that runs inside Weblogic or JBoss, browse to and deploy your model, everything gets generated and deployed. They also have an (architecturally discrete) RIA front end, web served xml that defines Java components and their composition into applications, plus drag-drop IDE for it. All with security, administration, SDKs, on and on. It's still complex to use if you try to do complex stuff with it.(Especially the front end - isn't it always that way?) Kind of a heavyweight solution, like EJB in general. There must be others.

Paul, there is not much info about dharbor technology on their site. To get to the white papers with the most promissing titles one has to register and wait. (I have been waiting for ower a week now, but no dice). You mention that you have done projects using their solution - would you mind contacting me off this thread for a chat (sano at strings dot net)?

Sorry for this off-topic post. Would be nice if there was a member-to-member messaging mechanism.