Roughly a decade has passed since the FAIR principles were coined, and in that decade, we have seen a growing interest in publishing research data and a rise in the number of datasets being published. In that same decade, the minimal retention period for research data has often been set at 10 years in guidelines and policies (e.g. Data Management Regulations Leiden University 2021). While this gave a minimum period, the unwritten assumption has often been that the aim was –with some exceptions- to keep data indefinitely. Moreover, FAIR, in essence, promotes the permanent keeping of published research data. However, ten years on, we are now in a situation where we realize that maybe not everything needs to be kept permanently, and that we should instead consider a more selective approach, because of environmental consequences, information overload, as well as financial considerations. Therefore, we now face the challenge of making informed decisions about retaining or discarding datasets. So how do we decide what to keep, and what the following steps are, and who gets to make these decisions? This workshop will start with an overview of the state-of-the-art guidelines and policies around preservation and retention. After that we focus, in moderated groups and using several concrete examples of published datasets, on four key aspects of future-proof data publishing in the interactive part: Meaningful selection: what to store and in what formats Appraisal/re-appraisal: criteria, how to make the decision, who decides what How to formalize this in a sustainable way: who will still be there after 10 years? How to move towards an “Internet of data” where data is only a query away while keeping it manageable in terms of size (and up to date)?