In this paper INDIGO, an approach to infrastructures for digital libraries is presented. It fulfills two crucial requirements to digital libraries: scalability and the ability to handle newly evolving document types. Based on a classification of digital library architectures, the main reasons for limited scalability and extensibility of digital libraries are identified. To overcome the identified problems the concept of mobile structure knowledge, on which INDIGO is based, is developed. The architecture of INDIGO is outlined and examples for the application of the concept are given.