(cross-posted from the Modeling Languages Portal )
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 specially 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 benefitial for the research group.
Obviously, in order to make sense for the technology provider to invest on 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 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 be solved) ensures the return of 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 it 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 soon (we are exploring now a collaboration with ProDevelop , the company behind the open source Eclipse Modeling tool MOSKitt ).