1. For now, I have stated that using a GPL like EER allows me to compare with an existing tool.

 I a agree that it is easier to find GPL tools to compare than DSL tools. You can then to create a model with the existing MDD-framework and with your new MDD-framework and compare the results. But this will only provide an answer to the question how efficient your framework is in ERR modeling language development.

2. Other advantages are that there has been done a lot of research on EER, that it has a rather simpel metamodel and syntax, which allows me to focus more on the functionalities of the used MDD-framework. 

Compared to the ER the ERR is much more complex  because it is a advancement of the ER adding additional concepts (subclasses, generalization, inheritance.... )

One of the reasonsfor desining a DSL is that an efficent way is needed to express domain problems. The languages is then less complex and more expessive to domain problems. 

When comparing the MDD-frameworks how good they are in DSL development then I recommend to take a domain modeling language. Therefore you have to decide in which area of domains you are interested. This application area then can also be a possiblity to limit you reasearch question.

Best Regards,

Martin Paczona

First, do you mean enhanced entity relationship model or do you refer to something else I am not aware off. Also I guess you are talking about domain-specific modeling (DSM) and maybe also include domain-specific languages (DSLs).

There are several issues with your approach. First, the baseline is implementing something directly in a GPL. Second, the complexity of the tool development (editor, transformations) has a lot to do with the level of abstraction of the DSL in contrast to the GPL. For example, a record description language, such as our instrumentation record language for Kieker (target languages, Java, C, etc.) is much easier to realize, as the source and target language have similar structures. Therefore, you have to differentiate between complexity of the mapping and complexity introduced by the tooling. Third, an important factor influencing the use of a MDD framework is implied by the documentation (or lack of documentation), which limits how the framework is used.

Furthermore, most people (unfortunately) do not use a MDD framework because they have the impression (not experience) that MDD is complicated. This is something you have to assess through a questionnaire or interviews of normal developers. However, there is also research available on this topic.

And one more thing: I would recommend reading "When and how to develop domain-specific languages" by Mernik et al.

Best regards

Reiner

http://dl.acm.org/citation.cfm?id=1118892

https://github.com/kieker-monitoring/instrumentation-languages

Dear,

Thank you for your answers! Yes, I am talking about Enhanced Entity Relationship.

So to make my question somewhat more practical as well:

The EER modelingtools use the UML class diagram notation from Connolly & Begg. I took all the criteria from the book and then looked where the old tool (made in Eclipse Graphiti) came short (by far). Next I made a tool in Sirius and compared it to the old tool by using the same criteria.

My research question is about DSM-tool development, while my assessment is made by making a GPL-tool (EER-tool). We have considered EER as a GPL for this paper.

So my question was, in which extent is making a GPL-tool a limitation for the assessment of current MDD-frameworks that support DSM-tool development?

Kind regards,

Sonny Selhorst