Many important historic and cultural sources are fragile and scattered in various archives, so that their full knowledge and usage are severely impeded. Project COLLATE develops a distributed Web-based repository with dedicated knowledge management facilities to support users in their work with digitized cultural material. As example domain COLLATE uses historical documents referring to films of the 20ies and 30ies. The repository set up by three major European film archives comprises a large corpus of digitized film censorship documents, related articles, photos, posters and film fragments. In-depth analyses of such documents give, for example, evidence about different film versions and cuts, which can be used for reconstruction of lost or damaged films or identification of persons (actors) and film fragments of unknown origin. As a virtual knowledge and working environment for distributed user groups, COLLATE provides content-based access to the repository and appropriate task-based interfaces for analyzing, comparing, indexing and annotating the material. It supports the users' individual and collaborative work with the sources, continuously integrating the derived user knowledge into the system. This growing body of metadata is exploited by the system using intelligent document processing and advanced XML-based content management and retrieval functionalities. This paper describes the conceptual approach of COLLATE, focusing on the tasks and requirements of the archives and users of the system. The functional system components will be outlined and assessed, describing how various user types and related complex tasks can be supported by comfortable task-based user interfaces in a collaborative working environment.