Sean Kavanagh's Discussions

Is UML a good analysis formalism?
2 Replies

Started this discussion. Last reply by Sean Kavanagh Mar. 23, 2009.

View All

Sean Kavanagh's Page

Latest Activity

Sean Kavanagh commented on Angelo Hulshout's blog post 'MDSD tooling index'

How about adding OOA Tool to the list. It is the only tool offering Shlaer-Mellor OOA/RD support at present (as well as Executable UML support).

June 21, 2009

Mike Finn left a comment for Sean Kavanagh

April 23, 2009

Sean Kavanagh left a comment for Mike Finn

April 23, 2009

Sean Kavanagh replied to Michael Osmond's discussion 'Competitive Advantage'

A very interesting observation. It highlights the fact that using MDD simply to make a specific project more efficient or better in some way is not an end in itself. MDD only becomes a strategic advantage when a whole software department (or organis…

April 19, 2009

Sean Kavanagh replied to Sean Kavanagh's discussion 'Is UML a good analysis formalism?'

Well in my case, I use Shlaer-Mellor OOA to my analysis models, not UML. The problem with developers doing analysis while coding is that any analysis done is intertwined with design and implementation logic which hides any analysis decisions from t…

March 23, 2009

Rui Curado replied to Sean Kavanagh's discussion 'Is UML a good analysis formalism?'

People use all sorts of shortcuts to jump directly to coding. It's human (developer) nature. Many developers just like to do analysis while coding. If the currently accepted analysis methods are fantastically complex as you said, why bother? I'm p…

March 23, 2009

Sean Kavanagh added a discussion

Is UML a good analysis formalism?

Before UML, we had OMT, Booch, Use Cases, Shlaer-Mellor etc. for capturing formal analysis models which could be shared with a team and the customers for review purposes. All of the different methods/notations were small but well defined. They could…

March 23, 2009

Sean Kavanagh replied to Rui Curado's discussion 'Is this community dead already, and what does that mean to MDSD?'

I agree it is far to early to condemn this site. I suspect there is too much diversity in skills/interests for the number of members at the moment and I agree with Ed Merks that the benefits of posting here are currently a lot less than posting a bl…

March 17, 2009

Profile Information

My interest in Model Driven Software Development:

I'm interested in Shlaer-Mellor OOA/RD and Executable UML based development.

My experience of Model Driven Software Development:

I'm an expert on Shlaer-Mellor OOA/RD and am currently building a Shlaer-Mellor and Executable UML CASE tool. I'm also creating a new 'standard' metamodel and interchange format for Shlaer-Mellor OOA/RD along with a two-way mapping to Executable UML.

My website

http://www.ooatool.com

My blog (or equivalent)

http://ooatool.blogspot.com

Comment Wall (3 comments)

You need to be a member of The Model Driven Software Network to add comments!

Join The Model Driven Software Network

At 10:18pm on April 23, 2009, Mike Finn said…: I lurk in the shadows like an assassin in the dark! I never lost faith but nowadays it goes under a different name. Thanks for the welcome.

At 10:33am on January 22, 2009, Vlad VARNICA said…: Hi Sean,

I like your website this is really cool :-)

It seems that my example is not well explained.
Actually, this is not the same tool because this Eclipse platform was used by two different plugins. My example is using an XMI 2.1 exported from the RSA 7 emx model which was developed with Eclipse 3.2. This xmi was then pasted and copied inside Eclipse 3.4 and reopen using EclipseUML.
I fully agree that Omondo and RSA design and model are very similar, to be honest they are the same. IBM has donated the EclipseUML2 metamodel to the community and all RSA development should now only use this model. We have a native integration to this model (no transformation needed. This is an Omondo patented technology and not open source) therefore we have the same model. For Omondo this is a native model and for RSA this is an export. Except this little difference the model is the same.
Our business strategy is to enter as many as possible new strategic account. IBM RSA is already used by over 95% of the fortune 100 companies. We want therefore to expand RSA models use among developers and JEE architect without loosing any model information. I mean that many models has been developped but not really used or badly used because there was a gap between modeler and developers. Now this gap doesn't exist anymore and it is time to use these models again and give them to the integration team which prefer to use agile approach and not anymore RUP.
Our strategy is to promote agile uses of models among developers and explain to architect that they can still work as they used to do because what ever is the technology the modeling job is still the same.
It is for example possible to start a project with IBM RSA for modeler, then to provide a model to the integration team and keeping modeling in an iterative process using EclipseUML for the end of the project. This means that a MDA or MDD approach using RSA is compatible with agile and iterativty by using EclipseUML at PSM implementation stage.
My personal recommendation for large projects is to use RSA at requirements and MDA levels, and then switch to Omondo EclipseUML for Java/jee project at PSM level. For smaller and Java/Jee oriented projects (project team smaller than 20 developers) then it is better to immediately start with EclipseUML.

Concerning now our market, we don't need executable UML because we use native JDT integration and there is no need for developer to get the code generated. They need to be free and code as they like :-)
EclipseUML allows these PSM modelers/developers to do their job and just use UML when they need it by merging all their existing code with the model without losing any model information. See more at...

It means that users are not anymore locked to a tool vendor.
Concerning the "I can get off-the-shelf modelled packages" this is long talk and to be honest I don't know. The current problem we have is that EclipseUML2 is an EMF transformation and the new standard has been technically implemented in the last 18 months. It is today not easy to just interchange business models between tools for code generation purposes because RSA, Papyrus and other use an GMF model which is transformed into EclipseUML2 models while EclipseUML2008 use the XMI model directly. It means that modelers should provide GMF models and use EMF transformation for some tools and the xmi models for Omondo.
Ok, it is still possible to get an xmi export from GMF using EMF but this is a difficult job and the granularity of the EclipseUML2 model information will not be perfect after such a transformation.
It is also impossible to directly change the XMI model and then recreate an GMF model because it is only working on a top down approach.
My point of view and recommendation is to use EMF at first iteration in order to create a stable and well structured architecture. Then to only use the EclipseUML2 model in a iterative approach. I mean that even if manipulating models with no code, we need iterativity in our daily job and not just an xmi export. Many UML 2.2 information can only be customized at xmi level and not before.

To keep it simple, I think that native integration (JDT/model/editor native integration using incremental triple synchronization) with no intermediate transformation is the bast way to simplify modeling and add productivity.

At 10:35pm on January 21, 2009, Vlad VARNICA said…: Sean,

I don't agree because with the EclipseUML 2 model we finally have a common metamodel between different tools. I mean that RSA, Togethersoft, Omondo, Papyrus and other tools are suing the same metamodel and can interchange models.
For example you paste and copy a RSA model into a Java project. then create a class diagram with EclipseUML and open the RSA model with the Omondo xmi editor. You can now drag and drop any element inside EclipseUML and have a common project composed by two different models. Note that this flexibility is only possible with EclipseUML because of our native integration on the top of the xmi and because Omondo is not using any transformation engine to create a model.
We can therefore manipulate the model and any model from any compatible UML 2.2 tool vendor. We can also merge models etc...
I have written an example of this model interchange, see more at..
The current problem is somewhere else but I don't know where and why ? Is UML too complex and just using a waterfall approach ? I have written a critic of the Waterfall approach, see more at...
btw, I don't know what is going on but it is really not cool for our UML ISV market because we can't anymore make a leaving of our software and only consulting is possible.