This paper describes compact storage models for gazetteers using state-of-the-art finite-state technology. In particular, we compare the standard method based on numbered indexing automata associated with an auxiliary storage device, against a pure finite-state representation, the latter being superior in terms of space and time complexity, when applied to real-world test data. Further, we pinpoint some pros and cons for both approaches and provide results of empirical experiments, which form handy guidelines for selecting a suitable data structure for implementing a gazetteer.