Research groups develop plenty of tools aimed at solving real industrial problems. Unfortunately, most of these tools remain as simple proof-of-concept tools that companies consider too risky to use due to their lack of proper user interface, documentation, completeness, usability, support, etc. Therefore, most of the tools are only used to convince reviewers to accept a research paper and then are completely forgotten due to lack of resources to invest in tool development (funding for non-core research activities is very difficult to get/justify).
This is very bad for research groups that risk missing the opportunity of having a large user base for their tools along with the benefits that this brings to the table (e.g. empirical validation of their research, feedback, visibility, collaboration opportunities and so on). This is especially true in emerging software engineering areas with growing industrial interest but without a dominant tool/s monopolizing the market and where some of our research tools could make a difference.
We believe that Model Driven Engineering (MDE) is one of these areas so it is even more urgent that we come up with a strategy to produce better tools that can have a real impact on how companies develop software.
The solution we have adopted in the AtlanMod team is to pursue the industrialization of our research tools by developing a partnership with a technology provider that ensures the existence of an open source but commercial-quality version of the tool. As part of the agreement, the technology provider commits resources on the non-core aspects of the tool and takes over traditional software development and maintenance tasks (including performance and usability improvements, bug fixing and user support) in exchange of visibility and the possibility of offering specialized services around it (e.g. trainings or customizations to specific clients). In our experience, this is a sustainable business model for the technology provider and very beneficial for the research group.
Obviously, in order to make sense for the technology provider to invest in the tool, the tool has to be valuable to a big community of users (or big companies). That’s why in fact our strategy involves three different actors:
Research groups solve research challenges posed by the community and develop a proof of concept to show that the solution is feasible. If a technology provider thinks that the techniques proposed by the research team are good and that the community is big enough, the provider and the team develop a partnership to create a commercial-quality version of the tool.
The fact that this is an application-driven research (i.e. the starting point is a real problem that a big company wants to have solved) ensures the return on investment for the technology provider. Adopting open source as the common denominator in all these activities is not absolutely mandatory but facilitates a lot the communication between the different actors and maximizes the benefits of the relationship (e.g. for the research group is easier to publish papers about it and the technology provider could commercialize services and adaptations on top of the tool for other big companies sharing the same problem).
We first successfully applied this strategy to industrialize our ATL model transformation tool thanks to our partnership with Obeo and later we have replicated this strategy on our MoDisco tool in collaboration with MIA-Software . Hopefully, other similar experiences will follow.
If you want to know more details about our industrialization strategy you can read this paper that explains our ATL experience . We will present this paper AT the Third International Workshop on Academic Software Development Tools and Techniques .
FNR Pearl Chair. Head of the Software Engineering RDI Unit at LIST. Affiliate Professor at University of Luxembourg. More about me.
I feel very happy for read about successful cases in MDE approaches. Specially the ATL, a tool that I consider very good, and well implemented. So much so that I adopted it as the infrastructure for my master theses.
But I have doubts if the tool developed in my master theses can cross the edge between a research tool and a industrial tool. I guess that the worst hurdle is tool evaluation, for the scientific community you need to prove, using a rigorous process, that your approach is correct (and I spent almost the same time to do it as to develop the tool). But for the industrial community, they dont care about it, your tool can have errors, but must have the characteristics very well pointed in your article.
to conclude, I just know some advices to cross this gap. Specially, How a researcher can deal with scientific as so as industrial community?
Thanks
I agree that the scientific community asks to prove that your approach is correct but this “proof” depends a lot on the kind of method you are proposing. Other people wanting to use your method must be sure that what you propose works
But, IMHO, what must be correct is the method you propose, if the tool implementing the method has some bugs then well, it just means that you are like everybody else in the world!
Great article.
I appreciate the method and the real experience. I think this is the only viable way for industrialization of software research.
We are having a similar experience with WebML (www.webml.org), with a few differences:
– for us, the substrate is the cross-fertilization coming from European Research projects (W3I3, WebSI, Cooper, …)
– in our case, the technology provider has been built from scratch internally. We opened a startup (WebModels Srl), as a Spinoff of Politecnico di Milano, with the mission of commercializing and instrumenting the WebML language. The company ended up producing the WebRatio tool(www.webratio.com), and is continuously interacting with our research lab and with the big customers that keep challenging the approach, the language and the tool.
These two aspects allowed more than 10 years of evolution of the language and the tool. If you want a flavour of the experience, check out this paper, published in John Mylopoulos Festschrift by Springer:
S. Ceri, M. Brambilla, P. Fraternali: “The History of WebML Lessons Learned from 10 Years of Model-Driven Development of Web Applications“. In book: Conceptual Modeling: Foundations and Applications, Essays in honor of John Mylopoulos, Springer LNCS, Festschrift series, vol. 5600, 2009, pp. 273-292
http://dbgroup.como.polimi.it/brambilla/10yearswebml
Just a quick update: I now described our path to innovation and industrialization more precisely here:
http://www.modeldrivenstar.org/2011/10/10-years-of-webratio-thinking-about-our.html