Data spaces receive considerable attention nowadays since they are at the heart of numerous large-scale European research initiatives shaping the data economy. Their goal is to establish secure environments that enable cross-organizational data management and thereby collect, integrate and make available heterogeneous data from various sources. Although we can observe a great interest in establishing new data spaces, questions of what exactly makes a data space and what it takes to design one remain open. To clarify that, we extracted and organized data space characteristics based on the analysis of 53 papers, as well as an empirical analysis of 47 real-world data spaces. We formalize the findings in a taxonomy to provide an intuitive tool that captures important data space design options. Our paper contributes to the understanding of an emerging artifact with significant implications for business, namely data spaces.