"Aggregative Digital Library Systems (ADLSs) provide end users with webportals to operate over an information space of descriptive metadatarecords, collected and aggregated from a pool of possibly heterogeneousrepositories. Due to the costs of software realization and systemmaintenance, existing "traditional" ADLS solutions are not easilysustainable over time for the supporting organizations. Recently, theDRIVER EC project proposed a new approach to ADLS construction, basedon Service-Oriented Infrastructures. The resulting D-NET software toolkitenables a running, distributed system in which one or multipleorganizations can collaboratively build and maintain theirservice-oriented ADLSs in a sustainable way. Aggregative Digital LibrarySystems (ADLSs) typically address two main challenges: (1) populating aninformation space of metadata records by harvesting and normalizingrecords from several OAI-PMH compatible repositories; and (2) providingportals to deliver the functionalities required by the user communityto operate over the aggregated information space, for example, search,annotations, recommendations, collections, user profiling, etc.
Repositories are defined here as software systems that typically offerfunctionalities for storing and accessing research publications andrelative metadata information. Access usually has the twofold form ofsearch through a web portal and bulk metadata retrieval through OAI-PMHinterfaces. In recent years, research institutions, university libraries,and other organizations have been increasingly setting up repositoryinstallations (based on technologies such as Fedora, ePrints, DSpace,Greenstone, OpenDlib, etc) to improve the impact and visibility of theiruser communities' research outcomes.
In this paper, we advocate that D-NET's 'infrastructural' approach toADLS realization and maintenance proves to be generally more sustainablethan 'traditional' ones. To demonstrate our thesis, we report on thesustainability of the 'traditional' OAIster System ADLS, based on DLXSsoftware (University of Michigan), and those of the 'infrastructural'DRIVER ADLS, based on D-NET.
As an exemplar of traditional solutions we rely on the well-known OAIsterSystem, whose technology was realized at the University of Michigan.The analysis will show that constructing static or evolving ADLSs usingD-NET can notably reduce software realization costs and that, forevolving requirements, refinement costs for maintenance can be mademore sustainable over time..."