A common complaint against people like me is that we just create our own localized digital libraries. A digital library, at least for those working within the confines (I did not say coffins) of a library, is an addendum to a normal library.
Rarely is software, however small, that was built by one group re-used by others. If, as the past few years have shown, it is almost impossible to cooperate on developing any piece of digital library software, we at least would need a framework that can guide software development to make it possible that the software is re-used. Very few attempts have been made to define such frameworks. There are a few however. One seems to be the DELOS framework, which I haven't studied yet, another is the 5S model, one that I am studying now.
The 5S model decomposes the problem of making a digital library into 5 components: Streams, Structures, Spaces, Scenarios, and Societies, all words starting with an S, in case you failed to notice.
Streams are basically information resources on the Internet. Text files, streaming video, you name it, they are all streams. Structures are structures within the streams. Perhaps the prime example is a XML document, where the XML explicates the structure of the stream. Other structuring principles are possible of course. Spaces are the operations one can perform on the (structured) streams. Operations can vary from indexing to defining an ontology. Scenarios is where the user enters the scene. Scenarios are a set of operations (state-transitions) that a user performs, or can perform, while using a digital library. Societies, finally, are groups of users with differing information needs. Societies and scenarios can be seen as an explication of user centered design (I surmise).
What is really good about 5S is that it does not stop at inherently vague descriptions such as given in the previous paragraph. Streams, structures, spaces, scenarios and societies are defined in terms of set theory and relational algebra. All definitions are formal. The advantage of this is that such definitions can guide implementation. For example, descriptive metadata is defined in graph theoretical terms, clearly suggesting ways to formalize metadata. If these basic structures are defined formally, exchange of both data and implementations should become easier. They could even hint, well more than hint, at how to define protocols for information exchange.
I wonder if there are environments in which 5S is already used. I am still exploring the model. It seems hard to get information on the practicalities of the 5S model. Nevertheless, theoretically it seems not only sound, but very attractive too. I like pretty... uh... work.
