Relational databases scattered over the web are generally opaque to regular web crawling tools. To address this concern, many RDB-to-RDF approaches have been proposed over the last years. In this paper, we propose a detailed review of seventeen RDB-to-RDF initiatives, considering end-to-end projects that delivered operational tools. The different tools are classified along three major axes: mapping description language, mapping implementation and data retrieval method. We analyse the motivations, commonalities and differences between existing approaches. The expressiveness of existing mapping languages is not always sufficient to produce semantically rich data and make it usable, interoperable and linkable. We therefore briefly present various strategies investigated in the literature to produce additional knowledge. Finally, we show that R2RML, the W3C recommendation for describing RDB to RDF mappings, may not apply to all needs in the wide scope of RDB to RDF translation applications, leaving space for future extensions.