Though much discussion has centered around the nature and use of learning objects, less attention has been paid to the problem of their distribution. Part of this is attributable to their nature: depicted as stand-alone content entities, learning objects do not appear to require any special effort to deliver: just bundle them up and ship them (by zip file, by CD, by aircraft carrier) to their designated recipient.
Closely tied to the question of distribution, and the subject of most discussion on the subject, is the question of finding learning objects. Thus, for example, we see the major paper on the subject of distribution, IMS’s Digital Repositories Interoperability specification (IMS, 2003) spending a great deal of effort defining ‘search’ and ‘retrieve’ parameters. Other major content distribution initiatives, such as the Open Archives Initiative (OAI 2002) are additionally defined in terms of search and retrieval functions.
In content networks there are three major approaches to search, and therefore, three major types of distribution networks. Which of these networks is selected for a learning object distribution system will have a significant impact on the type of search and distribution possible.
Federated Search Systems are the oldest and (arguably) the most popular search systems for content networks. The idea of a federated search system is that a single search is distributed to a number of stand-alone repositories. Each repository conducts the search internally, then returns the results, which are then displayed and sorted, usually according to their source. Some examples:
EduSource (demo) http://184.108.40.206:8080/axis/search/index.html
Because it conducts a separate search through each collection (or ‘repository’) of content objects, a federated search is more suited for networks consisting of a small number of very large collections. The federated search, for example, is most suited to a search of library collections.
Harvesting Search Systems are newer than federated searches and were developed in large, interconnected networks such as the world wide web. In a harvested search, a centralized search facility, called a ‘harvester’ or ‘aggregator’, contacts each of a large number of smaller resource collections and obtains information about that collection’s contents. The searcher submit a request to the aggregator, which conducts the search internally in its collected data, and then refers the searcher to the location where the original object is stored.
Most web searches are conducted using a harvested system. Note that a federated search would be nearly impossible in a large and distributed environment such as the web, since a search request would need to be repeated through hundreds, thousands or even millions of individual web sites. A federated search also means that individual content providers do not need to support searches on their own sites: they merely need to provide access to their data to a harvester. However, a harvesting system is not suitable for small networks of large collections, since the traffic created by the harvesting would far outweigh the traffic created by individual searches. Some examples:
DLORN (demo) http://www.downes.ca/cgi-bin/dlorn/dlorn.cgi
Peer-to-Peer Search Systems are composed of very large numbers of self-supporting repositories that share search information among each other. In one sense, a peer-to-peer system is similar to a harvesting system, since members continually poll each other for new content, but there is no centralized aggregator or harvester. A peer-to-peer system is also similar to a federated search since a single search request will be distributed to a number of other connected systems (or ‘peers’), but the search is distributed (or ‘propogates’) over an ever increasing number of collections until a result is found.
The decentralized nature of a peer-to-peer search is what makes it attractive to many providers. The most notable use of the peer-to-peer system has been (to date) the underground file-sharing networks such as Napster or Kazaa. Because content, and therefore searches, are distributed over a large number of systems, no single system needs to cope with a large traffic load or numerous search requests, and therefore a node may be run on an individual’s desktop with an ordinary internet connection. Some examples:
Many more examples of learning object repositories may be found at the University of Milwaukee’s Center for International Education: http://www4.uwm.edu/cie/learning_objects.cfm?gid=37
The three major models may be differentiated along a number of axes, however from the point of view of educational content one of the most useful axes is the degree of control over learning resources offered by different systems. In a federated system, because both the search and access to the objects are handled by individual collections, the greatest degree of control is allowed. By contrast, since a peer-to-peer system distributes both the search and the objects, the least degree of control exists. A harvesting system, which freely distributes search results but allows repositories to control their own objects, offers a middle approach.
The emphasis in learning object distribution thus far has been toward federated search systems. A federated search system can rigidly control access to search results, requiring authorization before these results are released. This option is preferred by owners of commercial educational content, since even search results are marketable content. A federated search system also promotes branding and, because the number of repositories searched is limited, can be used to reduce competition from wider networks of less expensive or free content.
But though content producers have many reasons for supporting a federated search system, it is not clear that the needs of a global network of online learning repositories will be best served in this way. Much content will by necessity remain outside the network, thus limiting the choices of participants. Moreover, such large systems require considerable overhead, and therefore cannot be supported by providers of inexpensive or free educational content. Though many providers are not ready for the wide-open environment of the peer-to-peer world, they are often willing to surrender some control in order to reach a wider market or to provide lower cost or free content.
Thus, just as the world wide web has adopted the harvesting approach for the location and distribution of online content, it is likely that the educational world will find itself moving away from large library-based federated systems and toward a harvesting model. This explains the increasing popularity of the Open Archives Initiative, which supports harvesting above all, and the even more rapid and widespread adoption of RSS, the major XML language used to support harvested searches today.
I have tried to demonstrated harvested searches using RSS in several of my projects. DLORN, mentioned above, is a system that harvests information about learning objects from several sources, including EdNA, and provides search results to web browsers or other applications. Edu_RSS is a service that shows how the harvesting of a subset of all available content repositories can provide a highly focused and yet broad-based search. Edu_RSS is located at http://www.downes.ca/cgi-bin/xml/edu_rss.cgi
IMS (2003). IMS Digital Repositories Specification. http://www.imsglobal.org/digitalrepositories/index.cfm
OAI (2002) The Open Archives Initiative Protocol For Metadata Harvesting. http://www.openarchives.org/OAI/openarchivesprotocol.html
All articles in this series by Stephen Downes
What do we know about knowledge?
So What is Knowledge Management Anyway?
From Knowledge Management to Learning on Demand
Learning: More Than Just Knowledge
Designing Learning Objects
Learning Objects Standards
Distributing Learning Objects
Using Syndicated Learning Content
How Learning Communities use learning
Beyond Learning Objects
Learning in Communities
Creating and Capturing New Knowledge