Not opening scientific data is costly. It has been estimated that a significant share of scientific knowledge disappears every year. In a 2014 study less than half of biological datasets from the 1990s have been recovered and when possible the recovery has necessitated significant time and efforts. In comparison, 98% of datasets published on PLOS with unique identifiers (data DOIs) are still available for future research. Open scientific data are fundamental resources for a large variety of scientific activities: meta-analysis, replication of research results or accessibility to primary sources. They also bring a significant economic and social value, as scientific data is commonly used by non-academic professionals as well as public agencies and non-profit organizations. Yet open scientific data is not costless. Ensuring that data is not only downloadable but usable requires significant investment in regards to documentation, data cleaning, licensing and indexation. Not all scientific data can be shared and verifications are frequently necessary to ensure that they do not incorporate copyrighted contents or personal information. To be effective, data sharing has to be anticipated throughout the entire research lifecycle. New principles of scientific data management aims to formalize the preexisting cultures of data in scientific communities and apply common standards. First published in 2016, the FAIR Guiding Principles (findability, accessibility, interoperability, and reusability) is an influential framework for opening scientific data. Policies in support of data sharing have moved from general and broad encouragement to the concrete development of data sharing services. Early initiatives go back to the first computing infrastructures: in 1957 the World Data Center system aimed to make a large range of scientific data readily available. Open data programs were yet severely limited by the lack of technical support and compatibility for data transfer. After 1991, the web created a universal framework for data exchange and entailed a massive expansion of scientific databases. Yet, numerous projects ran into critical issues of long term sustainability. Open science infrastructure have recently become key stakeholders in the diffusion and management of open scientific data. Data repositories ensure the preservation of scientific resources as well as their discoverability. Data hosted on repositories are more frequently used and quoted than data published in a supplementary file.