Tuesday, March 14, 2006

Where are we going?

I'd like to return to the questions raised by Joe Winchester in his article to see if we can provide some answers as they relate to the scope and vision of the proposed Eclipse Modeling Project.

Regarding the value-add perspective of high-cost commercial tool vendors, I'd agree this is a reason we've not see such tooling contributed in whole to open source. This is reasonable, considering the large effort it takes to produce such tooling (or at least, as it has in the past). Projects such as GMF promise to significantly reduce the cost of producing high-quality graphical editors for practically any domain. This will pressure commercial vendors to provide another level of value-add functionality in their products, or see themselves be commoditized much like the IDE market has been.

A significant part of this next level of value-add will come with the continued realization of model-driven development technologies, as I see it. This relates to Joe's second point, regarding the perceived usefulness of "modeling" tools by many. I suspect most would agree with the Agile value of "working software over comprehensive documentation," where the latter is typically seen as the primary value-add that modeling tools bring to development (like it or not). With model-driven development, the production of working software from models is the focus, and what if anything will grow popular interest in "modeling" tools.

To this end, the Modeling project aims to unify and extend the current capabilities found among Eclipse projects (both standards-compliant and alternative). Recent contributions to GMF and the EMFT project (model query, validation, transaction, and OCL), not to mention improvements coming to JET, brings the overall set of capabilities within Modeling much closer to what will be needed for model-driven tooling to continue evolving at Eclipse. The GMF project takes a model-driven approach, and leverages several of these capabilities to generate domain-specific modeling tools, which in turn can leverage the same capabilities to allow for the production of more model-based generative tooling. The potential is there, although significant work remains, most notably in the area of model transformation.

In conclusion, I would agree with Joe that the reason that we have not yet seen good, high-level tools in open source is that the (relatively small) community hasn't gotten around to creating them. While a goal of the Modeling project is to provide a home at Eclipse for high quality frameworks and exemplary tools of this nature, it is just getting started. The easy part was in its declaration, while the hard work to realize it through additional contributions and improved interoperability is at hand.


Blogger Bill Quinn said...

Hello Mr Gronback, I am a Borland Share-Holder and I recall that back in May '05, Our Company announced a Java Roadmap and that Peloton would be here in the first half of '06 and combine the best of JBuilder and Eclipse. Peloton would be the first commercial JBuilder product built atop the Eclipse Platform. Has Borland achieved this promise and will it all be sold off with the tools? Can't we make a percentage of each Borland Tool "Open Source" and keep the rest? Thank you Bill

1:56 PM  

Post a Comment

<< Home