In the last three decades the management of information technologies has become a major issue in most organizations. In this context, virtual communities (VC), as a means to technologically foster collective social exchanges, are often proposed as viable solution to a wide range of problems. They have been largely used and adapted in the fields of knowledge management, e-commerce, local participation, leisure, innovation, etc... Each one of these domains has developed its own understanding of the concept and its specific analytical tools to evaluate or manage these processes. This thesis takes this heterogeneity as starting point in order to explore the key question of management of virtual communities (VC). Our main hypothesis is that the management of virtual communities does not dependent upon the domain in which it is used. We propose to explore the possibility of considering a generic model of virtual communities management that could be used in any situation. In order to define the principles of such a model, we have observed a large array of VC cases. These cases were analysed in a comparative perspective so as to determine what they have in common in terms of management procedures. This comparison lead us to propose the blueprint of a generic model of VC's management drawing upon, in its core, the socio-technical processes of virtual communities, as most significant dimensions. In our approach, we envisaged the negotiations taking place among the stakeholders involved in the life of virtual communities (managers, intermediary users, end users) as being of central importance. In these negotiations, the managers have to deal with a continuous tension between top down decisions and participative processes. We examine how, in practice, the management of these "many to many" communication systems imply a constant mix of social and technical dimensions. The generic principles we propose stress how managers set limits, define identities and build attractiveness while building their virtual communities. In the last part of this thesis, we discuss how such generic principles could lead to a better understanding of virtual communities management. In this perspective the suggested model could be perceived as an analytical tool to assess virtual communities of all kinds. It should also help to allow cross-fertilisation processes between CV researchers and practitioners of different domains.