Linking research & learning technologies through standards

We have already discussed our development activities around the IMS LODE activity for discovery of learning objects. However, what we have described so far presupposes that learning object descriptions are already available to a user, because the user can access those descriptions in their local repository, or through a repository federation they have access to.

But there will not in the foreseeable future be a Super-Federation of all education repositories in the world, nor indeed does there need to be. Rather than unleashing users on all e-learning repositories in the world, it makes more sense for users to discover learning object collections that they don’t already have access to—but which are of direct interest to them. So users should be able to target their searches for content to the collections which will pay off, instead of doing an inefficient, iterative blanket search across Everything.

That means that users should have access to metadata describing learning object collections, and not just learning objects, so that they can do targeted discovery of collections. Such discovery should be based on the attributes users are interested in: not only descriptions of the content, but access rights, language of content, intended audience, and so forth. Moreover, the metadata about collections should include information about how to gain access to the collection through defined service interfaces.

To save time and allow scaling up, that information on service interfaces should be machine-readable: this allows clients to add a new collection to a federation automatically, and access its content with a minimum of configuration. In effect, this allows ad-hoc federations of collections to be built, as needed. With discoverable descriptions of collections available, resource discovery is no longer constrained to the repositories and federations that a user is already set up to access.

To realise this vision of flexible collection discovery, IMS LODE has modelled a machine-readable description of learning object collections, as a profile of the ISO 2146 standard for registry description. We have already posted about the underlying modelling work leading to this. The collection description includes machine-readable configurations for services used to access the collection—such as Harvest or Search. This means that the collection descriptions can support both a repository federation driven by federated search, and a repository federation relying on a registry harvesting metadata centrally.

The second stage of development work undertaken by ADFI was to test this LODE schema for collection description. Because of time constraints, we tested the schema only in the context of a registry harvesting metadata centrally. Our testing used the collection description schema to configure harvest of a collection automatically, once the collection had been discovered. It also updated the harvesting configuration, if the collection schema indicated that the collection location had changed. The automatic harvest configuration and reconfiguration was successful. We did need to make sure that the identifiers for collection object remained constant when the collection was relocated (and not base them on their currently physical URL), so that the new harvested records would not be duplicated over the old.

Testing also went into the scenario of a single collection spanning multiple repositories, and a repository hosting multiple collections. It is important for the LODE model of collections to uncouple collections from the systems that host collections, so that collections can be deployed greater flexibility. (This is why we have systematically written about “collections” rather than “repositories” above.) Our testing confirmed that the schema could support this separation of concerns; we relied on OAI-PMH Harvest as our interaction with collection, so we used harvest set identifiers to differentiate between collections.

Testing thus confirmed the feasibility of the approach to collection description we have sketched, and has given us experience to build on as we help formalise the IMS LODE model. The project wiki for the ADFI development work is publicly accessible, and has further details on the testing undertaken.