Since the 1950s and 1960s, systematic approaches to the diffusion of various flows (information, people and goods) and, especially, innovations across urban networks (or systems of cities) have emerged in geography and regional science. Both US and European (Sweden and France) scholars, through direct or indirect collaboration, developed various conceptual frameworks and empirical case studies of urban networks in such respect (see Peris 2016 for a review of this vast literature). However, these schools of thought evolved in different ways in later decades, resulting in a wider diversity of urban network studies and definitions. Yet those schools have in common that they remain relatively abstract in scope, proponing their own theories and concepts, with limited transversal dialogues among them; when operated and applied empirically, the most common geographical scale to study urban networks remained at the national level (Ducruet and Lugo 2013). It is, perhaps, the Pour l’Avancement des Recherches sur l’Interaction Spatiale (PARIS) team in France which innovated the most in that its members gradually expanded the geographic scope of their initial French-based studies to the whole European continent, through analyses of cities connected by multinational firms (Rozenblat and Pumain 1993) or by airline and railway networks (Cattan 1995). These large-scale investigations questioned the influence of city size and specialization on the attractiveness of specific activities through a relational perspective, thereby reaching beyond the traditional analysis of technical networks and planar graphs in this discipline (Ducruet and Beauguitte 2014). It is quite common nowadays, with the explosion of computer capacity and the adoption of new methodological tools related by simulation models and big data, to find global-level studies in the literature on urban networks of all kinds. Yet, this chapter would like to insist on a remaining dark corner within urban network research, namely, cities connected by water or maritime transport.