An interesting development is summarized by the notion of ”Ubiquitous Computing”: In this area, miniature systems are integrated into everyday objects making these objects ”smart” and able to communicate. Thereby, everyday objects can gather information about their state and their environment. By embedding this information into a model of the real world, which nowadays can be modeled very realistically using sophisticated 3D modeling techniques, it is possible to generate powerful digital world models. Not only can existing objects of the real world and their state be mapped into these world models, but additional information can be linked to these objects as well. The result is a symbiosis of the real world and digital information spaces. In this thesis, we present a system that allows for an easy access to this information. In contrast to existing solutions our approach is not based on a server-client architecture. Geostry bases on a peer-to-peer system and thus incorporates all the advantages, such as self-organization, fairness (in terms of costs), scalability and many more. Setting up the network is realized through a decentralized bootstrapping protocol based on an existing Internet service to provide robustness and availability. To selectively find geographic-related information Geostry supports spatial queries. They - among other things - enable the user to search for information e.g. in a certain district only. Sometimes, a certain piece of information raises particular interest. To cope with the run on the single computer that provides this specific information, Geostry offers dynamic replication mechanisms. Thereby, the information is replicated for as long as the rush lasts. Thus, Geostry offers all aspects from setting up a network, providing access to geo-related information and replication methods to provide accessibility in times of high loads.