Web 2.0 tools are currently evolving to embrace semantic web technologies. Blogs, CMS, Wikis, social networks and real-time notifications, integrate ways to provide semantic annotations and therefore contribute to the linked data and more generally to the semantic web vision. This evolution generates a lot of semantic datasets of different qualities, different trust levels and partially replicated. This raises the issue of managing the consistency among these replicas. This issue is challenging because semantic data-spaces can be very large, they can be managed by autonomous participants and the number of replicas is unknown. A new class of algorithms called Commutative Replicated Data Type are emerging for ensuring eventual consistency of highly dynamic content on P2P networks. In this paper, we define C-Set a CRDT specifically designed to be integrated in Triple-stores. C-Set allows efficient P2P synchronisation of an arbitrary number of autonomous semantic stores.