Identifiers, migrating metadata and orphaned objects
Sometimes you want an identifier to unambiguously refer to a resource: punch in a loved one's phone number, and you expect their phone to ring, and no other. Much the same identifier - resource functionality would be desirable for learning objects, but that has not been sorted yet. CETIS organised a meeting with a broad swathe of stakeholders to address the issue.
Take the following scenario: a teacher makes a new learning object for her course. The object is stored in the college's repository, where it gets an identifier of some sort- a string of letters and/or numbers that is sufficiently unique to distinguish it from all the others. Then the collection of the college is rolled into a larger repository or network of repositories, and there happens to be another object with the same ID; now what? Just changing the ID of one will break any references to it made earlier...
What's needed, then, is a practical, simple way of making sure that the ID assigned to an object really is globally unique. This apperently simple job, however, is just the start of a whole list of other issues.
The main outcome of the CETIS identifiers meeting was a tentative inventarisation of such issues, and an initial plan for resolving them. The identifier issue is not new, and members of organisations such as vendors Intrallect, Autodesk, FDLearning and Simulacra, as well as standards groups IMS Global and Australia, The Norwegian E-standard project and MELCOE and Universities ranging from Carnegie Mellon to Staffordshire Uni all brought their expertise to bear.
The problem itself has been addressed already by groups that include CanCore, IMS and TSO, but the immediate spur for this meeting was the 'Identifiers for learning objects' discussion document by UKOLN's Andy Powell.
The essence of the identifier problem is the reconciliation of the need for identifiers to be the same forever, for everyone, independent of where the resource is, who owns it and in what kind of system it resides in. Not an easy task when the resources -the learning objects- are themselves composed of multiple objects, can be changed, have alternative versions, copied and aggregated into, and disaggregated out of, larger learning objects. Add to that the necessity of maintaining different sets of metadata (and thus metadata identifiers) about the same object for different groups, and you have quite a puzzle.
Fortunately, the problem has already been solved by a good many other sectors; as CETIS' Scott Wilson pointed out, identifiers for learning objects are not necessarily that different from identifiers for tins of beans. There are, however, some issues specific to e-learning.
The publishing industry, for example, may require a very robust system with a central authority that guarantees the one and only identifier for a specific object. Such functionality is necessary for the kind of rights management systems they'd like to see. For an educator, such an overhead may not be necessary for an object that started life as a little extra to support an existing lecture.
Another such e-learning specific requirement was pointed out by Carnegie Mellon's Dan Rehak: the ability to choose an appropriate version of an auxiliary object, based on the preferences of the learner who is requesting it from within another learning resource. It is, for example, quite normal to have a manual, helpfile or glossary handy when going through the learning tasks set by a piece of course content. Such 'scaffolding' material may have versions that are essentially the same in content, but different in branding, language or access method (e.g. screenreaders for the visually impaired, etc.). It would be nice to pick the appropriate version from a combination of the auxiliary resource's general identifier (i.e. the ID of the content) and a learner's profile.
Other identifier issues are not so much e-learning specific, as they are about implementing systems in practice. There are, for example, at least six different technologies for attaching and maintaining unique identifiers for digital objects in general. Each has its own qualities, but the important thing is that the e-learning sector as whole settles for one; or at least makes sure that the solutions used by different communities are interoperable.
As the meeting has already addressed some of the more philosophical aspects of identifiers, the next steps involve gathering use cases from everyone with a stake in the issue. The participants themselves have already come up with a respectable list of cases, but the "global" and "forever" nature of the issue means that this really is one problem that needs thorough input from all concerned.
Once a sufficient number of use-cases have been assembled, they need to be mapped against a list of core qualities of e-learning identifiers that the meeting has already drawn up. Add some agreed definitions of concepts, and some synchronisation of effort between appropriate spec bodies like IMS and the Dublin Core Metadata Initiative (DCMI) and no object need ever be lost again...
For more information, download the full meeting report (1.7 Mb, MS Word).