New classes of services for digital libraries have been explored in the research community with many now being tested and deployed in real-world settings. Unfortunately, there have been problems, some predicted and some unforeseen, in the development of these services. Moreover, a number of problems have been identified as becoming critical in the future, especially the difficulties associated with preservation of archives of digital content when data formats and media both have limited lifespans. We have analyzed a number of these services in a research library (personal collection management, annotation, institutional repositories, and learning object repositories) in terms of their user and task requirements for three prototypical users: a researcher, a teacher and a student. Given these, we develop a unified conceptual model of how these services should work together in collaborative situations and some of the requirements for systems and service organizations that might implement them. This model centers around a vision of the digital library as the hub of a personalized, integrated information management environment. In particular we note the importance of: a flexible collection model that encompasses traditional collections managed by librarians, community collections managed by either professionals or users, and private, individual collections; both synchronous and asynchronous interactions between users and data; integration between library services and tools for using the information managed by the library; user annotations as metadata associated with documents and document fragments and the integration of this facility with metadata search; a data model that provides universal, granular access to the contents of documents and other resources whether they are structured, semi-structured or unstructured-, a data security model that provides full support for a range of private to semi-public to fully public data and metadata for users and user communities, and simple, flexible and integrated control of privacy, security, and integrity of all such data and metadata. Beyond these, one significant observation that deserves greater emphasis is the need to provide direct and sophisticated support for a variety of epistemic communities in any such digital library. In particular, both users and communities need new tools to organize and use the information they are collecting in the context of their personal and collective knowledge. The claim is that for these patterns of use, it is less useful to provide universal ontological models or category hierarchies than it is to provide sophisticated tools to allow both automatic and user-directed semantic categories and relations to be defined, refined and adapted for information management in context. Finally, this analysis reveals a set of technical requirements that highlight a number of critical gaps in the conceptual integration of the services and in the infrastructure underlying their current implementations. One model for building such set of facilities takes the traditional three-tier enterprise application model and adapts it for use as a layered application model for integrated data-driven applications[2]. We then consider current and potential technologies for implementing this model and suggest that the best hope for a solution that meets both user and technical needs lies in a combination of integrated services built on a document modelling and collaboration infrastructure that we call NODAL, the Network-Oriented Document Abstraction Language[1]. We illustrate how this model may enable even the most ambitious visions of the potential of the collaborative digital library.
