Data-centric storage, which supports efficient in-network data query and processing, is an important concept for sensor networks. Previous approaches mostly use hash functions to store data, where data with the same key valueare stored in sensors at or near the same geographic location.We propose a new data-centric storage method based on sorting. Our method is robust for different network models and works for unlocalized homogeneous sensor networks, i.e., it requires no location information. The idea is to sort the data in the network based on their key values, so that queries -- including range queries -- can be easily answered. The sorting method balances the storage load well. We present a sorting algorithm that is both decentralized and efficient.