The basic record-oriented data and metadata model employed by most digital (and traditional) libraries has a limited ability to fully model this multi-dimensional information context.

First, metadata records, and metadata repositories, primarily represent individual item properties. They often fail to completely model contextual relationships [43] that surround resources and do not distinguish among the multiple entities – resources, metadata, agents, ontologies – that are part of that relationship structure. Furthermore, because they are frequently based on fixed schema or models, they are difficult to adapt to evolving information needs. The NSDL metadata repository, for example, only distinguishes between collections and items and represents only the membership relationship between them. Because the MR is stored in a relational database, each new relationship requires schema redefinition. This lack of flexibility has proven problematic due to the changing requirements over the span of the NSDL activity.

Second, the static nature of metadata records, which are generally created once by resource creators or catalogers, is also problematic. Resource context is dynamic, expressing changing patterns of use, preference, and the shifting cultural environment. Recker and Wiley write "a learning object is part of a complex web of social relations and values regarding learning and practice. We thus question whether such contextual and fluid notions can be represented and bundled up within one, unchanging metadata record" [48].

Third, an information model that is metadata-centric inevitably runs against the problematic fuzziness of the "data or metadata" distinction14 [19]. For example, we have noted above that one of the useful forms of contextual information is annotations. Are these metadata (about something) or data in their own right? There is no one answer, but an architecture that imprints the distinction between data and metadata makes it difficult to deal with such ambiguities.

Finally, we have also noted the importance of information reuse – the ability to take primary resources and combine them into aggregate learning objects or lesson plans [46] and recursively reuse and re-factor new objects. Because the physically-bound information units in the traditional library were not amenable to such reuse, a metadata – centric approach – managing descriptive records – was possible. However, a digital library needs to be resource-centric, providing the framework for managing, manipulating, and processing content and metadata and the seamless line between them.