Running context-based systems with a fixed infrastructure involves substantial investments. There have been efforts to replace those systems with self-organizing ones. Therefore, recent systems use peer-to-peer (P2P) technology as a basis. Context-information is bound to a specific location and thus should be stored on a nearby node. Common P2P algorithms use one-dimensional ID spaces. However, locations have at least two coordinates, namely x and y. We use space-filling curves to map the two-dimensional area onto the one-dimensional ID space. In this paper we discuss the suitability of different space-filling curves for the average case and for stochastic scenarios.