A recent survey about data practices covering 25 ESFRI Research Infrastructures in the realm of GOFAIR and a recent analysis of the practices of 55 research infrastructure initiatives showed that almost all claim to be FAIR compliant and recommend Open Science principles, but that the actual practices are different. The conclusions from these studies widely confirmed an earlier broad survey within the Research Data Alliance in 2014 which led to the start of the RDA Data Fabric Group. While the final step of data publishing has been subject of many improvements, the situation in the data labs where the actual work of data scientists is occurring did not change so much despite all increase in awareness. Combining data from different silos and/or disciplines is still highly inefficient and time-consuming. About 80% of the time in such data projects is wasted with data wrangling.
Assumptions that the formulation of the FAIR principles and the development of the Digital Object approach (recently extended to FAIR Digital Objects (FDO)) may lead to fast steps towards convergence and thus reduce inefficiencies soon, turned out to be too optimistic. Three major reasons can be mentioned that indicate that more time will be necessary: (1) The FAIR Principles allow many interpretations and are not blueprints to build data infrastructures. As a consequence IT experts are making different suggestions of how to overcome the inefficiencies and are all claiming to support the FAIR Principles. It is too early to expect fast acceptance of the FDO approach which would be one way of implementing FAIR. (2) The statement from G. Strawn, one of the Internet pioneers, that "standards are good for science, but bad for the scientists" is correct in so far as researchers prefer to use known methods to be able to continue their research. Currently, there is no interest from researchers to change their practices except for small adaptive steps. This is by the way a re-occurring pattern knowing that it took decades to broadly adopt the Internet. (3) Funders do not yet dare to fund sufficiently large reference implementations and testbeds to work out important researcher directed components that could convince researchers to adopt new ways.
In my talk I would describe the observations from the studies about data practices, especially mention a few reoccurring patterns across disciplines and also discuss with the help of a workflow example, how a FDO approach could facilitate data driven research across disciplines.