This section discusses how the current reference model is constructed, its current state and how it should be evolved.
Reference Model
The analysis of the current state and the related work indicates that the existing architectural artefacts are too few, fragmented and inconsistent, which does not justify the creation of a reference architecture. Discussion within the various communities showed a lack of a common understanding of research infrastructures, which obstructs the development of standards and services to provide the desired interoperability. As such there appears to be a bigger need for a reference model, which will provide an abstract framework for understanding significant relationships among the entities of research infrastructures. A reference model will improve the discussion, description and comparison of the current infrastructures, as well as the functionality and quality of the desired infrastructure.
An architectural framework was needed to deal with the complexity and different stakeholders of any system, and in particular for distributed infrastructures of the ESFRIs. Initial attempts looked at e.g. the Zachman Framework, which did not provide enough guidance and TOGAF, which appeared too heavy and focused on change. ENVRI recommended the use of RM-ODP. RM-ODPs viewpoints allow the model to be used by the different stakeholders of research infrastructures and its focus on distributed processing and its transparencies provide a useful mechanism to describe both the core entities as well as the -archetypical- distributions of research infrastructures.
The reference model is created from the analysis of the existing documentation. As this documentation appeared fragmented (with the lack of a reference model), much of the input comes from the authors' experience in CESSDA, CLARIN and DARIAH and their collaboration with researchers and service providers throughout the data lifecycle.
It must be clear that this is the first version of the reference model. This version should now be applied to existing infrastructures to validate and improve both the reference model as well as the resulting infrastructure. Next activities include alignment with existing models like OAIS and DSA, validation with the ESFRIs, addition of example cases and, if needed, further detailing of the viewpoints.
A shortcoming of RM-ODP is that it may be difficult to understand the scope of the different viewpoints [VALLECILLO]. This reference model will soon be provided with examples that describe its application.
[VALLECILLO]