The standard rendering style used in Openstreetmap today produces hardly readable maps in countries where the usage of latin script is not the norm, at least from an average westerners point of view. Our map style uses a renderer independent approach to solve this. We use localization (l10n) functions that create readable names. They are implemented as stored procedures in the PostgresSQL database which contains the Openstreetmap data. The targeted latin langage (german, english, …) can be easily selected. The talk will show how these functions currently work and will give an outlook on potential future extensions. In contrast to almost all legacy geographic data Openstreetmap does already contain a lot of localized data acquired by mappers from all around the world, which should be used whenever possible (Example: japan instead of 日本). Automatic transliteration can then be used as an alternative if no latin names are available in the database. Especially when using transliteration there are many pifalls which have to be addressed depending on language and country. Some of them have already been dealt with by the current implementation and are presented in the talk. Others, which appear difficult or impossible to solve are also shown. Another challenge which exists in localization of maps are political problems. I will briefly describe some of these issues at the end of my talk. Sven Geggus (Fraunhofer IOSB)