We are most pleased to have permission to publish this remarkable
research Mr. Downes presented at the University of Milan, November 12,
2002. It is not an easy read. The density of information, learning design
implications and the alternative interpretation of 'economy of education'
are demanding of close attention and thoughtful consideration. We feel
this paper is a milestone contribution to creative models for online learning.
Reusability of Learning Objects
in an Academic Context:
A New Economy of Education?
eLearning: una sfida per l’universita, Milan, November 12, 2002
The purpose of this paper is not to discuss the creation
and use of learning objects per se but rather to look at systems for
locating and distributing learning objects. What will be argued is that this
system is currently poorly constructed, based essentially on what may be called
a silo model of distribution. A series of problems and issues related to this
model will be discussed. In place of the silo model, a distributed model of
learning object repositories is proposed. This model is based on a set of
principles intended to create an open and accessible marketplace for learning
objects, in essence, a learning object economy. To conclude, a model for a
distributed learning object repository network is proposed.
For readers unfamiliar with the concept of learning
objects, the generally accepted definition is that learning objects “any entity,
digital or non-digital, which can be used, re-used or referenced during
technology supported learning.” (IEEE, 2002) Wiley (2000) defines a learning
object as “any digital resource that can be reused to support learning.” Even
so, as Wiley comments, “the definition is broad enough to include the estimated
15 terabytes of information available on the publicly accessible Internet.” In
this paper, a functional definition of learning objects is employed: a
learning object is anything that is exchanged in what may be called the learning
2. The State of the Art
In this section, common methods for locating and retrieving
learning objects will be discussed. In particular, three major systems will be
described: course portals, course packs, and learning object repositories. In
addition, systems for collecting and organizing learning objects, learning
management content systems, will also be described.
A course portal is a website offered wither by a consortium
of educational institutions or a private company working with educational
partners that lists courses from a number of institutions. The purpose of a
course portal is to enable a student to browse through or search course listings
to simplify the student’s selection of an online course. The following are
examples of course portals.
TeleEducation. A New Brunswick, Canada, learning
organization, TeleEducation NB hosts the TeleCampus Online
Course Directory. Courses are submitted by institutions and screened to ensure
that they are fully online. The database contains more than 50,000 courses,
including about 3,000 free courses and 1,200 complete and fully online programs.
TeleCampus provides a subject-based directory and search services.
UNext. Focusing on business education,
UNext collaborates with major business schools such as the Columbia Business
School, Stanford University and the London School of economics to provide
courses in leadership and management, e-commerce, marketing, finance,
accounting, and business communications through the private and for-profit
institution, Cardean University. http://www.unext.com
Hungry Minds. Hungry minds offers more than 17,000 courses
through its online campus, Hungry Minds University, from course providers such
as the University of California at Berkeley, the University
of California at Los Angeles and New York University. Hungry Minds also provides
learning content through publishers such as For Dummies, CliffsNotes, and
Fathom. Created by Columbia University and including
partners such as the University of Chicago, the London School
of Economics and Political Science, Cambridge University Press, The British
Library, The Smithsonian Institution's National Museum of Natural History, and
The New York Public Library, Fathom is a centralized for-profit learning object
repository. While Fathom provides lectures, interviews, articles, performances
and exhibits, its major focus is an offering of online courses from member
http://www.fathom.com (You, 2001)
Course packs are packages of learning
materials collected to support a course. Offered primarily by educational
publishers, course packs are collections of learning materials offered to
instructors for use in traditional or online courses. The course pack may be
pre-defined or custom built by the instructor. The instructor is expected to
supplement the course pack with additional content, educational activities,
testing and other classroom activities.
Some course packs, such as those offered
by XanEdu, are stand-alone. This means that the course pack is distributed as a
separate product and purchased by the student directly through the college or
university bookstore. Supplementary educational materials are offered by the
instructor on his or her course website or are delivered in a classroom setting.
Other course packs are available for use only in a learning management system (LMS).
Course packs delivered through a learning management system are more like
‘default’ online courses. Using tools provided in the LMS, the instructor
selects the course and customizes it for delivery online.
The following are examples of course
WebCT Course Packs.
The learning management system WebCT offers course packs consisting of a course
structure and set of readings offered by publishers with a distribution
agreement with WebCT. Course packs are purchased by the institution on a
seat-license basis and are then customized by the instructor.
Canada’s SchoolNet. In Canada, the leading learning
resources portal is probably Canada’s SchoolNet. A list of resources is
displayed, each with a short description and a link to an external website.
SchoolNet also provides information about each site and provides an “advanced
search” using metadata. Each resource in the “curriculum” area is approved by a
professional “pagemaster”. For the most part, however, SchoolNet links to
institutional home pages, and not to learning resources per se. Teachers
using the SchoolNet service must still search through these sites in order to
locate suitable materials.
MarcoPolo. MarcoPolo is a compilation of teaching resources
from six educational institutions that provide free internet content for K-12
education. What the six partners have in common, and what makes this an
important and interesting development in online learning, is an adherence to
national curriculum and evaluation standards in the subject areas. Material is
categorized by grade level and individual items are matched to individual
learning topics. Despite its strengths, however, MarcoPolo is a closed project;
only the six member institutions contribute content. There is no centralized
search facility and no metadata listings for the resources.
XanEdu. Xanedu is a learning resource site that collects
articles from journals, magazines and other resource providers. Instructors may
compile ’course packs’ consisting of collections of these materials; students
who subscribe to XanEdu may access these course packs. The materials are sorted
by category and may also be located using a search mechanism. Like MarcoPolo,
however, XanEdu is a closed project. It draws materials only from selected
publishers. And while it allows subscribed students to browse through its
materials, the vast bulk of resources available on the internet cannot be found
Learning Object Repositories
Learning objects are stored in databases called learning
object repositories. There are two major types of repositories: those containing
both the learning objects and learning object metadata, and those containing
metadata only. In the latter case, the learning objects themselves are located
at a remote location and the repository is used as a tool to locate learning
objects. In the former, the repository may be used to both locate and deliver
the learning object.
Most learning object repositories are stand-alone. That is,
these repositories function a lot like portals in that they contain a web-based
user interface, a search mechanism, and a category listing. Another major class
of learning object repositories functions more like a database attached to
another product. An LCMS, for example, may contain a learning object repository
intended for its exclusive use.
Two major models for learning object repositories exist.
The most common form is a centralized form in which the learning object metadata
is located on a single server or website (the learning objects themselves may be
located somewhere else). An alternative model is the distributed learning
object, in which the learning object metadata is contained in a number of
connected servers or websites. Distributed learning object repositories
typically employ a peer-to-peer architecture to allow any number of servers or
websites to communicate with each other.
The following are examples of some learning object
Merlot. Described above, Merlot is probably the most well
known learning object repository. Merlot is a centralized repository containing
metadata only and pointing to objects located at remote locations. It is
stand-alone, acting like a portal for learning objects. In addition to providing
search and categorization, Merlot provides a peer review service provided
by communities of experts in different subject areas. http://www.merlot.org
Campus Alberta Repository of Educational Objects. CAREO is
a centralized collection of learning objects intended for educators in Alberta,
Canada. A stand-alone repository, CAREO contains metadata and provides access to
learning objects located on remote web servers. http://www.careo.org
Portals for Online Objects in Learning. POOL is a
distributed (peer-to-peer) repository system under development intended to
create a pan-Canadian repository of learning objects. A primary objective of
POOL is to develop and distribute tools for creating connected learning object
repositories. http://www.newmic.com/pool/ (not currently functioning). See
National SMETE Distributed Library. In development for the
(SMETE), NDSL is intended as a “federation” of learning object repositories,
each library using different document formats, different systems of
classification, and different database and repository management schemes. NDSL
is intended to join these libraries using a common search engine called
Emerge and a method for sharing resources called LOVE (Learning
Object Virtual Exchange). (Chen, 2001)
Learning Content Management Systems
Learning objects are typically small, consisting of no more
than the equivalent of an hour or two of instructional time (there is some
debate as to how small a learning object may be and whether educational content
must contain pedagogical features, such as a statement of learning objectives,
in order to qualify as a learning object). Most educational institutions deliver
larger chunks of instruction, called courses. To create a course,
therefore, a set of learning objects must be assembled into a package.
Packages organize learning objects sequentially. In order
to create a course out of, say, a dozen lessons, where each lesson is a separate
learning object, a course author arranges these lessons into a sequence. In some
cases, where the learning objects are smaller units, course designers may need
to create lessons composed of a sequence of individual modules, then the course
as a whole out of the sequence of lessons. However created, the sequence of
objects is used to define course-specific entities as the course outline or
table of contents.
Packages are created using a Learning Content Management
System. While a course author could locate and assemble learning objects by
hand, it would be tedious and unproductive to do so. Courses created using
learning objects are typically created using a development environment called a
Learning Content Management System (LCMS). The LCMS performs two major
functions: it provides authors with a means of locating learning objects, and it
assembles them into standards compliant learning packages (or courses). (Ellis,
Though many types of LCMS are available, the typical LCMS
will contain four essential features: an authoring application similar to the
computer assisted software environment (CASE) described above, a collection of
learning objects (called a repository), a means of sending the completed
course to a delivery system (called a delivery interface), and
Using an LCMS, a course author defines major features of
the course: its topic area, say, or its grade level. The author then instructs
the LCMS to search through the learning object repository for relevant resources
(because the data is in XML, the search can be very precise). From the search
results, the author may review a learning object or select it for inclusion in
the course. The LCMS retrieves the object metadata from the repository and
inserts it into the course package. The LCMS automatically adds
institution-specific formatting and prepares the package for delivery.
3. Problems and Issues
In general, the issues surrounding the location,
distribution and reuse of learning resources online have to do with system
architecture and resource based on what I call the “silo model.” On the silo
model, resources are not designed or intended for wide distribution. Rather,
they are located in a particular location, or a particular format, are intended
for one sort of use only.
The silo model is dysfunctional because it prevents, in
some essential way, the location and sharing of learning resources. In an
important sense, such resources or architectures are broken because they
require some additional step, usually involving manual labor, in order for
developers or learners to make use of the material. The requirement of such a
step adds significantly to the cost of a learning resource and in some case may
prohibit its use altogether. In fairness, this cost or prohibition may be
imposed by design. However, from the point of view of a learning object economy,
the resource or architecture is unusable.
There are numerous ways a learning resource or architecture
may follow the silo model. In this section, a number of these are listed. Few
products embody all of these problems. But most contain instances of at least
one of these problems. And even a single instance of the silo model is enough to
prevent a learning resource or architecture from being used as part of a
A standard is proprietary when it is secret or when
patents, copyrights or other restrictions prohibit its use. The standard is
created by a commercial entity and specifies “equipment, practices, or
operations unique to that commercial entity.” (National Communications System,
1996) With the advent of the internet, proprietary standards are much less of an
issue than in years past. Nonetheless, proprietary standards continue to abound,
especially in the realm of multimedia formats.
The use of a proprietary standard divides a distribution
network into those people or systems able to use the standard, and those people
or systems unable to use the standard. For example, a document created using DXF
for Autocad may not display properly in Cadkey, which uses CADL, or ACIS, which
uses SAT. Another example is XrML, a digital rights management language
developed by ContentGuard. Developers have been reluctant to use the standard
because of Microsoft’s control over the standard. (DRM Watch, 2002)
Proprietary standards pose numerous risks to developers.
One risk is that the standard will cease to be supported in new software.
Documents encoded in older MS Word formats, for example, need to be converted
before they can be used. There is the risk that licensing terms may change, and
as a consequence, require that user pay unexpected licensing fees. If the
standard is not widely shared or distributed, as is the case, for example, with
Microsoft Windows, it is difficult to develop new applications, and the holder
of the standard enjoys an advantage over competing products. Additionally, the
choice of viewing software may be limited. Because of these risks, it is
difficult to encourage wide adoption of proprietary standards.
Several of the systems listed in the previous section
depend in whole or in part on proprietary standards. Course packs designed for
Web CT, for example, cannot easily be used in competing learning management
systems. It is necessary to use a content migration utility (some versions of
which are no longer supported) to obtain interoperability.
Overly Strict Standards
Even when a standard is non-proprietary, it may be the case
that the standard is too limiting for widespread use. If, for example, a
standard requires that only a limited type of data will be transported by a data
transmission system, then novel applications using different types of data will
be impossible to develop.
Much of the criticism around the Sharable Content Object
Reference Model (SCORM) was focused on this sort of objection. SCORM was
developed to support self-study modules designed for use by the U.S. Military.
Learning objects defined using SCORM are mutually independent, meaning that only
the most basic sort of sequencing is enabled. This has led critics to suggest
that SCORM is not flexible enough to allow for a variety of pedagogies. (Welsch,
In a similar manner, transport protocols may also be too
strict. Just as, for example, a road is much less strict (and therefore much
more widely used) than a railroad, so also a distribution network that delivers
only learning objects (and not, say, journal articles) is less likely to be used
than a network that delivers both.
Some of the systems described in the previous section
adhere to standards that are too strict. Any system requiring SCORM compliance,
for example, will be viewed in this way. So also will repositories that list
learning objects only, such as Merlot.
Standards may be unreasonably strict in other ways. The GNU
General Public License (GPL), for example, requires that any product developed
using GPL software must also be GPL. Since the GPL is intended “to make sure the
software is free,” all modifications of GPL software must also be free. (GPL.
1991) While the purpose of this condition is to ensure that developers cannot
convert a GPL application into a proprietary application, the interpretation is
that GPL prohibits the development of any proprietary applications within a
given application environment. (Microsoft, 2002)
Another issue related to the strictness of standards in the
complexity of the standard in question. If the standard is too complex, use of
the standard requires an involved process or development tool. Legacy content,
which might have met a laxer standard, must be converted to the new standard.
XrML has been criticized because of its complexity (DRM Watch, 2002) as has
SCORM (Welsch, 2002).
Under the name of “enterprise
solutions,” learning content management systems have become tightly integrated
monolithic software bundles. Such integration is even touted as a benefit by
many software companies. Saba Software, for example, promises to “replaces
today’s ad hoc processes and disparate systems with a single system and a
unified view of everything your organization needs…” (Saba Software, 2002)
Purchasers of such systems are as a consequence committed to a single solution
for all aspects of learning management. If, for example, you do not like the
discussion board or quiz generation tool in WebCT, perhaps finding it too
complicated to manage (Shelangoske, 2002), there are no alternatives;
third-party products cannot be simply ‘plugged-in’ to replace the WebCT default
The purchase of such a system
additionally requires paying for much more than may be desired. Because an
essential component of learning content management systems is a database of
learning objects (Nichani, 2001) a purchaser is
committed to buying hardware and software support (for example, a database
system such as Oracle) that may be well beyond their needs. In a tightly
integrated system there is no means to deploy third-party or hosted services to
manage part or the entire database; it must be located in-house.
closed marketplace exists when an owner of a learning content management system
has only a limited selection of content to choose from. This limitation occurs
when the LCMS vendor reaches an exclusive agreement with a content publisher to
distribute materials. Such agreements formed the bulk of press announcements
through 2001 and 2002.
of the major distributors establishing priority in learning management systems,
XanEdu has reached distribution agreements with a number of vendors, including
Blackboard, Fathom, Microsoft, America Online, and Gallileus.
agreements make it more difficult for purchasers of competing systems to obtain
access to XanEdu’s exclusive library. In such cases, each student must obtain a
separate XanEdu account, providing credit information and paying XanEdu
directly. Similar restrictions prohibit direct access to a wide variety of
published content produced by other vendors.
such agreements make it more difficult for content publishers to sell to users.
Unless affiliated with a publisher (and consequently willing to accept
publishers’ terms and conditions), content providers are unable to make their
material available for selection by LCMS users. Because LCMS content selections
are offered as a bundle, often from LCMS vendors, content providers not selected
to become part of this bundle are excluded from selection.
consequence of such a Byzantine marketplace is that established publishers with
large content libraries are favored. Because of the overhead involved, and
because established publishers are wary of the competition, free content is
discouraged and generally unavailable. This has the consequence of increased
prices for content consumers.
The combination of monolithic
systems and closed marketplaces tends to favor large educational institutions
over smaller colleges and independent study. If it is necessary to purchase a
large LCMS and pay premium prices for educational content, a smaller institution
with fewer students cannot compete with institutions with enough students to
distribute the cost. Independent study in such an environment is increasingly
difficult, with most choices for potential students difficult to find or simply
A system is disintermediated when there is no form of
assessment or review guiding the selection of learning resources. The
purchaser’s only guide to the quality of learning material, in such a system, is
obtained directly from the vendor. In a disintermediated system, there is no
independent third party available to filter selection, assess or certify
materials, or to comment on their potential use.
The contrary to disintermediation is intermediation. Some
systems, such as merlot, attempt to provide a rudimentary for of intermediation
through the provision of peer reviews of educational materials. Merlot’s system,
however, is closed in the sense that only a select group of people may provide
reviews. And it is limited in the sense that reviewers evaluate only materials
found in Merlot.
The need for some form of intermediation is evident from
the numerous ad hoc mechanisms already in place. Such systems are
typically institution-specific and involve the use of proprietary forms and
assessment criteria. The system provided by dlnet, for example, provides a
specific set of criteria and a review form. It is used only by reviewers rating
material for inclusion in the Digital Library Network for Engineering and
Technology. (dlnet, 2002)
Similar systems are employed by the Peer Review of
Instructional Technology Innovation (PRTI) program in the Broadband Enabled
Lifelong Learning Environment (BELLE) project and the Development of a European
Service for Information on Research and Education (DESIRE) project. (Place,
2000) In both cases, the purpose of the review is to establish a scope and
selection criteria for the repository.
Systems where a review process is intended to select
materials for inclusion in a specific repository may be described as
“gate-keeping” services. Such services are undesirable for several reasons.
First, they create significant overhead by requiring that each item be reviewed
manually, causing a backlog in the addition of materials to the repository.
Moreover, the results of the review are unavailable to third parties; the
reviews are available only to users of a specific repository. Moreover, there is
no means in such a system for third party or dissenting reviews.
In the case of many other systems, there is no review
mechanism available at all. A purchaser of online articles or journal
publications from a subscription service has only the article abstract available
to guide selection. The reader must pay the access cost in order to determine
that the abstract is misleading or that the content is not relevant.
Though progress has been made recently (with, for example,
the IMS Re-useable Definition of Competency or Educational Objective (RDCEO) (Kraan,
2002)), there is a tendency to view the network of learning objects and
repositories as a stand-alone service on the world wide web, not integrated with
or compatible with many other resources and services available.
This is an issue mostly of perception rather than
implementation. It results from the presumption that an application profile,
such as SCORM, is a standard, and thereby the presumption that SCORM sets out
the one and only way to describe learning objects. This has been the basis for
much discussion, including heated exchanges surrounding the idea that “SCORM is
for everyone.” (Rehak, 2002) In fact, many application profiles, even in the
educational arena, exist. (Friesen, 2002)
In fact, SCORM is application profiles, which in turn are
“schemas which consist of data elements drawn from one or more namespaces,
combined together by implementers, and optimized for a particular local
application.” (Heery and Patel, 2000) Understood as such, it is therefore
unreasonable to expect that any given application profile, even SCORM, would be
widely used in multiple contexts.
The issue of selective semantics arises when a network
application, such as a network of learning object repositories, standardizes on
a given application profile. Such specialization restricts the usefulness of
such a network to the application envisioned by the designers of the application
profile, and thus precludes different (even closely related) applications. A
repository network, for example, that standardized on SCORM would preclude from
consideration resources that are useful to course designers, such as journal
articles, but which may not be described as learning objects per se.
Though it is not possible to find a network designed along
such principles, there is no shortage of learning content systems proclaiming
themselves to be “SCORM compliant.” Viewed in this light, unless such systems
are designed to manipulate RDF data, rather than only SCORM data, such systems
are announcing merely that they are not suitable for a wide array of
applications (though they may be ideal for environments envisioned by the
designers of SCORM).
Digital Rights Mismanagement
The issues related to digital rights management (DRM) are
legion and need not be reviewed at length here. That said, since DRM will be an
essential component of any network of learning object repositories, it is
necessary to survey some of the major issues.
The first and probably the most significant concern is that
no simple DRM solution has been widely implemented. This is because in many
implementations, digital rights management has been conflated with the idea of
digital rights enforcement. Thus, for example, the first widespread of
proprietary electronic content required the use of specialized devices, known
generically as eBooks.
Though eBooks satisfied the need to enforce digital rights,
they were generally considered a failure because they required the purchase of
specialized hardware and could not interoperate with anything else. As Hillesund
(2001) notes, “Today there are two factors working against e-books and hindering
diffusion. These factors include the overall poor quality and high prices of
reading devices and the lack of proper and interoperable digital rights
management (DRM) systems.” Insisting on physical control of digital materials
stymies the exchange of these materials. (Lyon, 2001)
The state of digital rights management for web-based
resources is not much better. In order to access content, it is typically
necessary to negotiate access with each separate supplier. A person dedicated to
purchasing online content, for example, may have to obtain separate accounts
with Corbis (an image service;
http://www.corbis.com/ ), Lexis-Nexus (a clipping service;
http://www.lexis-nexis.com/ ) , Salon (a magazine;
http://www.salon.com ), and so on and on. In many cases – the most notable
being the online distribution of music – there is no means to obtain access to a
full catalog of material.
The use of clearing houses that characterized first
generation digital rights management is insufficient for the wide variety of
materials and business models desired in online content exchanges. No trusted
fiduciary agent, as described by Lyons (2001), exists to facilitate the exchange
of learning resources. Consequently, a fractured and distrusting system of
credit-card deposits, proxy servers and disabled file formats has emerged. This
has resulted in content that is difficult and expensive to obtain and
impractical to use.
4. Design Principles
These design principles are intended to govern the
development of an architecture for a distributed learning object repository
network (DLORN). The purpose of the principles is to guide the description of
the components employed, the standards followed, and the principles governing
the operation of the network.
These principles are in one sense descriptive and in
another sense prescriptive. They are descriptive in the sense that they attempt
to capture the essential elements of what is likely to be the most successful
system for the distribution and use of learning materials on the internet. They
are prescriptive in the sense that they are intended to inform the development
of such a network.
The protocols used by components of the components of DLORN
to communicate with each other and with external systems are described,
documented, and freely available to the public at large. The purpose of this
principle is to encourage the development of complimentary systems that may
interact with and support the functionality of DLORN.
For example, a DLORN should embody interoperability with
other networks and systems that are being developed by libraries and museums
worldwide. In other words, the DLORN is not a network with own proprietary
communication protocols open to only repositories within system but can operate
with others outside systems such as the Open Knowledge Initiative (OKI; http://web.mit.edu/oki/)
and to be aware of other communications protocols, such as Z 3950 (Miller,
1999), to augment its own information objects with those from other collections.
The standards developed or used by DLORN shall be
royalty-free. The purpose of this standard is to ensure that there is no a
priori overhead cost incurred by agencies wishing to offer services
compatible with DLORN. Imposing an a priori cost immediately poses a
barrier to small and medium sized enterprises that may wish to participate and
it biases the network toward the provision of commercial content only.
Enable, Don’t Require
Where possible, DLORN will not require adherence to a
particular constraint, but rather, will allow users of the system to exercise
options among various models. The design of the system will be to allow systems
that exercise different options to interoperate and to work within the same
This principle is essentially based on the idea of defining
different levels of compliance required for interoperability within the network
as a whole than would be required by specific instances of the system. At the
network level, a minimal standard is desired in order to achieve the widest
functionality possible. One way of stating this is to require interoperability
at the syntactical level only, without stipulating as to the content being
This need must be balanced against the need for a more
robust interoperability, one that requires a common understanding of meaning as
well as sentence structure. Although interoperability is possible, if the
agreement consists of syntactic structures only, such interactions are
functionally meaningless. Greater agreement is desired, and the greater the
level of semantic agreement within two systems, the greater the
In practice, what this means is that although the network
as a whole imposes no prior semantic restrictions, in order to use the network
it is necessary that some semantical agreement is required for two
instances to interoperate within this framework. In other words, though the
network imposes no restrictions on how something is described, evaluated,
valued, or transacted, entities within the network must define how these are to
be described. 
Open-Source Infrastructure Layer
The infrastructure layer is the set of components that
provides end-to-end functionality for DLORN. It is described in the paper
Distributed Learning Object Repository Network Infrastructure Layer
(forthcoming). The set of components in the infrastructure layer will be
developed and distributed as royalty-free open source software. The purpose of
this principle is to demonstrate functionality without requiring financial
advances, and to provide a base of functional components on which other services
and applications may be developed.
Open or Proprietary Service Layer
Over and above the infrastructure layer, it is hoped and
anticipated that third parties will develop components with increased
functionality, offering an improvement in design or services over and above the
functionality provided by the infrastructure layer. Such components may be
developed as free and open applications, or they may embody commercial and
proprietary components. The purpose of this principle is to enable the
development of commercial applications that generate a revenue stream for
software developers and service providers.
Component Based Architecture
The DLORN is to be designed not as a single software
application, but rather, as a set of related components, each of which fulfills
a specific function in the network as a whole. This enables users of the DLORN
to employ only those parts of DLORN that suit their need, without requiring that
they invest in the entire system. It also allows for distributed functionality;
a user of DLORN may rely on a third party to provide services to users. The
purpose of this principle is to allow for specialization. Additionally, it
allows users of DLORN to exercise choice in any of a variety of models and
Any given component of DLORN may be replicated and offered
as an independent service. Thus, it is anticipated that there will be multiple
instances of each component of the DLORN infrastructure. The purpose of this
principle is to provide robustness. Additionally, it is to ensure that no single
service provider or software developer may exercise control over the network by
creating a bottleneck through which all activities must pass.
Any provider of learning materials may prepare and
distribute learning materials through DLORN. Though DLORN will support the
registration and indexing of various providers, this registration will be free
and optional. The purpose of this principle is to ensure that providers are not
faced with a priori ‘membership fees’ or similar tariffs in order to gain
access to potential purchasers. This does not preclude restrictions, tariffs or
controls on specific instances of a DLORN component. However, in any case where
a restricted component, such as a for-profit metadata repository, exists, an
equivalent unrestricted component, such as a public metadata repository, will
There will be no prior restraint imposed on the
distribution model selected by participants in DLORN. Specifically, DLORN will
accommodate free content distribution, co-op or shared content distribution, and
commercial fee-based content distribution. The purpose of this principle is to
ensure fair and open competition between different types of business models, to
ensure that users of DLORN are not ‘locked in’ to the offerings provided by
certain vendors, to provide the widest possible range of content options, and to
ensure that prices charged for learning content most accurately reflect the true
market value of that content.
DLORN imposes no prior restraint on the metadata standards
used by participants to describe given resources or services. Metadata
repositories are tolerant of different standards employed by different providers
of learning materials. Metadata repositories also (attempt to) provide output in
the standard requested by users of the system. This means, for example, that a
vendor may elect to employ IEEE-LOM to describe its learning materials, while a
consumer may request information in the form of the CanCore profile. Standards
tolerance extends to the description of digital rights, classification and
taxonomies, and evaluation and annotation. The purpose of this principle is to
enable an inclusive marketplace, to reduce risk by vendors when metadata
standards are selected, and to enable the development of vendor-specific or
custom metadata for particular uses.
The description of DLORN will include descriptions for
communication using multiple channels or multiple modes of communication. For
example, DLORN will enable requests using web services such as XML-RPC or SOAP,
gateway interfaces such as HTTP-POST, and harvesting protocols such as OAI. The
purpose of this provision is to enable redundancy in the system. It is also to
reduce the liability of the network should any given standard become a
royalty-based standard. It is also to provide software developers the greatest
range of options for the creation of new services.
Multiple parties may provide metadata describing a given
learning resource. There is no prior restraint exercised by providers of
learning materials on evaluations, appraisals, comments and other fair
descriptions of their learning material. The purpose of third party metadata may
be to provide alternative classification schemes, to indicate certification
compliance, or to provide independent assessments and evaluations of learning
resources. The purpose of this principle is to ensure that potential users of
learning resources to obtain neutral descriptions of that material. It is also
to create an environment for the creation of optional but value-added third
party services for which fees or other costs may be charged.
Integration with the Semantic Web
DLORN should be considered as an implementation of and an
extension of the semantic web. This means that DLORN metadata data and services
would be available to the semantic web as a whole. It also means that DLORN can
and should incorporate elements of the semantic web, such as sector-specific
ontologies, into its own design. The purpose of this principle is to ensure that
DLORN is capable of the widest reach possible. It is also to reduce the
duplication of effort between developers working in specific domains and
educators working in the same domain.
Multiple Data Types
No prior restrictions are imposed on the data types to be
transported through DLORN. This includes but is not restricted to various
content formats, proprietary or otherwise, such as HTML, and the like. This
provision is also intended to allow learning resources that are not learning
objects, as variously defined, to circulate through the system. For example,
academic papers distributed through the Open Archives Initiative, news articles
distributed through various vendors, conference, class or seminar registration
and information may also be distributed through DLORN. The purpose of this
proposal is to enable any learning resource, including in-person learning
services, to be accessed, and not merely a specific subset of learning
Simple Digital Rights Management (DRM)
The principle behind fee-based and subscription-based
transactions is that it should be easier to buy material than to steal it. Thus
where possible, the acquisition of rights and the exchange of funds will be
automated. The purpose of this principle is to reduce transaction and clearance
costs for purchasers of learning materials.
Transactions within DLORN are brokered. That is, typically,
a given provider of learning materials will work with a single agent who sells
to multiple purchasers, and a given user will work with a single agent who
conducts transactions with multiple vendors. Vendors and users may select from
any number of brokering services, so that no single transaction agent controls
the network. Vendors and purchasers may act as their own agents. A vendor or
purchaser may elect to employ multiple agents. Agencies acting on behalf of,
say, a provincial department of education, may act as agents for a given
populations, such as the students of that province. The purpose of this
provision is to eliminate the need for the creation of multiple accounts, to
allow users to user resources from multiple vendors, and to provide a choice of
agents, and therefore a greater likelihood of trust.
This principle states in effect that users of the system
own their own personal data. The user’s agent operates on behalf of the user,
and releases information or money only with the agent’s explicit consent. The
purpose of this principle is to engender trust in the system and to ensure
privacy when dealing with multiple agencies.
5. The Distributed Network
A network rather than a (single) system
What we are proposing is a set of inter-related
applications distributed over the internet and communicating with each other.
This seems to me to be the single major factor distinguishing our approach from
other approaches as defined in IEEE/P1484.1/D9 of IMS Repositories. This is
accomplished in three major steps:
Separating the functionality of an LCMS / LMS architecture into distinct,
stand-alone components that communicate over TCP/IP
Allowing (encouraging) the development of multiple instances of these
Providing indexing or registries of these instances
Thus, for example, instead of
envisioning a single metadata repository that indexes all learning objects (or,
as we see within common practice, all learning objects within a specific domain,
such as a geographic region or company), we envision multiple learning object
repositories that may or may not focus on a particular domain.
In other words, the model we
are envisioning resembles much more the model employed by the World Wide Web
than it does the model envisioned by a content management system. In my opinion,
this is a key turning point.
Core Components of the Network
Learning Object Repository - hosted by vendors on vendor sites, provides
vendor metadata and learning object servers
Metadata Repository - hosted elsewhere, harvests metadata from vendors
and amalgamates, allows queries from eLearning systems. Norm Friesen has written
a useful backgrounder on harvesting:
eLearning system - queries metadata repository, user selects resource,
retrieves resource from learning object repository, displays
This core functionality is relatively simple and is already
established in other domains, for example, in news syndication. Consider the
following combination of components:
News Object Repository - Original articles are posted on news site and
RSS metadata is available for harvesting
Metadata Aggregator - such as NewsIsFree collects metadata, indexes
(maybe) and provides (topic-specific, sometimes) search
News Viewer - such as Amphetadesk - accesses the aggregator for an index,
then retrieves the selected item from the news repository
Contrast to Library Model
Most other implementations, including IEEE/P1484.1/D9,
employ a model whereby learning materials are akin to books in a library (or, in
some other way, 'content' to be managed). Consequently, they envision that
implementations of the architecture will access collections of this content,
typically (but not always) stored on location. The process they envision is
Acquire the content
Index or classify the content
Deploy the content
In a network model, there is
no need to manage collections of content. So instead of working with learning
objects specifically (as defined by
IEEE/P1484.12.1 or pick your
specification) the network works more generally with what may be called learning
resources, or even more precisely, learning opportunities. This includes, but is
not limited to:
Learning objects, properly so-called
Other academic works, such as journal articles
In-person classes or seminars
Instructors, coaches and tutors
While it is permissible to
search for a specific category of learning opportunities, such as a learning
object, the design does not require that all resource fit that particular
category. This is enabled by tolerating the use of different schemas in learning
Learning opportunities in this
model should therefore more accurately be thought of as akin to 'processes'
rather than 'things'. The desired result of, say, a learning object search
system is not so much to acquire a resource as it is to locate it
and, when appropriate, display it or run it.
Part or all of the learning
resource may or may not be cached on location, but this is left to the
discretion of the particular instance and is not a defining feature of the
Component Registry Services
In the network proposed, there
are multiple instances of each component. Of course, there are multiple learning
objects. But there is in addition multiple learning object repositories
(typically, one for each learning object vendor) and multiple metadata
In order to provide access to these resources, it is
necessary to provide indexing or registry services. The purpose of these
services is multifold:
To provide a list of the available instances
To establish and verify ownership of these resources, for the purpose of
maintaining or updating information about them in the system
For example, consider the list
of learning object repositories. A vendor wishing to offer learning objects
through the network will need to declare that the repository exists and where to
find the list of available resources. By registering the repository, the vendor
is able to make its presence known and to ensure that important information –
such as its URI – will not be changed by third parties.
The registry system envisioned
is consistent with existing approaches to the provision of services on the
internet. It is anticipated that the repository indexing service would resemble
the UDDI and WSDL protocols.
Norm Friesen has written a
useful discussion regarding the registration and indexing of resources.
Functionality of the System versus Functionality of the Learning Resource
Many models of learning object
architecture presuppose that the system being deployed contains a great deal of
functionality. For example, IEEE/P1484.1/D9 includes as two (of the four)
essential components the ‘coach’ function and the ‘evaluation’ function. A wide
variety of other functions are embedded in LMS and LCMS design, for example,
class registration, discussion and chat.
The weakness of this approach is that the purchaser of an
LMS or LCMS is restricted to only one choice in the delivery of these functions,
that is for example, restricted to only one discussion board or one class
registration system. This makes LMS and LCMS systems needlessly complex, and
needlessly restricts the range of options available to the purchaser. Thus, such
functionality is envisioned in this model to reside in the learning opportunity,
thus greatly increasing the range of choice available to developers.
This functionality of the system is therefore defined in
the learning opportunity, rather than in the system itself. This is the most
immediate and obvious different between this approach and IEEE/P1484.1/D9. In
the IEEE draft standard, elements such as 'coach' and 'evaluation' are defined
as components of the architecture. On this model, they are resources that may be
deployed within the architecture.
In addition to the three core elements, a number of
secondary elements are also required in order to meet a number of the objectives
of learners, learning institutions and content providers. As in the case of the
core components, there may be multiple instances of any secondary component.
This allows users of any core component to exercise choice in the selection of
secondary components. These components include:
A system of third-party metadata
A digital rights system
A learner (user) information system
A reporting or tracking system
What is significantly
different about this model and models envisioned in IEEE/P1484.1/D9 and IMS
Repositories is that:
The components are optional: you develop (or buy) them and use them only
if you need them
For any given component, you may select one of many instances
These components may reside outside your own system
As in the case of the primary components, a registry
service is developed for each type of secondary component.
Third Party Metadata
Third-party metadata is a
crucial component of the network that is not really envisioned by IEEE-LOM or
IMS (though to be fair they do permit reference to third party ontologies, as in
IEEE/P1484.12.1 9.21 and 9.22. See http://www2002.org/CDROM/alternate/744/index.html
for some description and uses of third party metadata.
The core principle of third
party metadata is that there may be multiple metadata files, perhaps even
located on different hosts, written by different authors (some for-profit), that
describe a single learning resource.
For example, a single learning
resource may have associated with it:
A description, in IEEE/P1484.12.1, created by the author or owner of the
An indication of certification, using a specialized metadata schema,
provided by a professional association
Metadata containing a review (or a reference to a review), provided by a
public service agency
Digital rights information, authored by and hosted by a DRM handling
Classification of the object, authored and hosted by a library authority
Digital Rights Management
A principle objective of the digital rights system (DRM)
enabled by the network is to create a system where multiple suppliers work
through a common interface. To enable this, it is important to provide a choice
of business models. A business model consists of two essential components:
The definition of the business rules, and
The application of the rules in software functionality
In traditional DRM, the
definition of business rules is represented in specific DRM metadata. Two major
approaches exist, ODRL and XrML, though numerous sub-variants exist. These
approaches are XML schemas defining the allowable documentation of specific
rights for a specific (group of) learning object(s). See
In order to establish DRM for
a given learning resource, the metadata associated with this resource identifies
the metadata, usually managed by a third party (see below), defining the DRM
associated with the learning resource.
Beyond Digital Rights Management: Employee/Consumer Rules
Most examinations of DRM deal in general with the
application of business rules to learning object transactions. For the most
part, these are rules established by the content owner or vendor. But it is
important to look beyond the traditional formulations whereby all the rules are
established by the vendor. Classes of employee/consumer rules will also be
identified and handled by different parts of the system.
An approach similar to DRM is
taken for the definition of employee/consumer rules. Using a (n as yet
undefined) XML schema, the various employee/consumer rules, such as the ones you
have listed above, are defined in an XML file owned by the employee/consumer.
This file may be maintained by a personal information service or buyer's agent
(several such files may exist to handle different aspects of employee/consumer
rules - for example, pricing, personal information, financial information and
presentation will likely be located in different files, handled by different
These rules are applied by
various subsystems: the metadata repository, the learning object retrieval
system, and the viewer itself.
Employer rules are established using the same system as
employee/consumer rules. By ‘employer’ in this document we could also include
entities such as school boards, colleges or universities, professional
associations, and indeed, any third party given permission by the employee or
consumer to apply rules.
During processing, if employer
rules apply (a consumer/employee uses the same system for job training as, say,
hobby learning), then the employer rules are merged with the consumer/employee
rules. They are then applied in the subsystem as appropriate.
Learner / User Information System
An additional secondary
component is a set of learner information systems. The concept is similar to the
‘resume’ or ‘portfolio’ system described by Chuck Hamilton of IBM at NAWeb.
Details to follow.
 This paragraph was significantly informed by Norm
Friesen and Toni Roberts.
Chen, Si-Shing. 2001. NBDL
(National Biology Digital Library). Slide Presentation.
Dlnet. 2002. Guidelines for the Reviewer. National Science
DRM Watch. 2002. XrML 2.0 Review. Giant Steps Media
GPL. 1991. The GNU General Public License.
Ellis, Ryann K. 2001. LCMS Roundup.
Friesen, Norm. 2002. Survey of Learning
Object Metadata Implementations. CanCore.
Heery, Rachel and Patel, Manjula. 2000. Application
profiles: mixing and matching metadata schemas. Ariadne. 25.
Hillesund, Terje. 2001. Will E-Books Change the World?
First Monday, Volume 6, Number 10.
Institute of Electrical and Electronics Engineers, Inc.
2002. Draft Standard for Learning Object Metadata (IEEE
Jack, Hugh. 2001. DG: 18.2.1 Proprietary “Standard”
Formats. Design Engineer on a Disk.
Kraan, Wilbert. 2002. Objective Re-Usable Competency. CETIS.
Lyon, Gordon. 2001. The Internet Marketplace and Digital
Merlot. 2002. Merlot Peer Review.
Microsoft Corporation. 2002. Microsoft Shared Source
Philosophy: Frequently Asked Questions.
Miller, Paul. 1999. Z39.50 for All. Ariadne. 21.
National Communications System. 1996. Telecommunications:
Glossary of Telecommunications Terms.
Nichani, Maish. 2001. LCMS = LMS + CMS (RLOs).
Oregon State University. 2002. Baccalaureate Core Course
Saba Software. 2002. Saba Learning, Enterprise Edition.
Place, Emma. 2000. Quality selection: ensuring the quality
of your collection. DESIRE information gateways handbook.
Rehak, Dan. 2002. SCORM is not for everyone. CETIS.
Shelangoske, Susan. 2002. Beginning WebCT Instruction:
Lesson 4, Quizzes and Grading. Cleveland State University.
Welsch, Edward. 2002. SCORM: Clarity or Calamity. Online
XanEdu. 2002. Press Releases.
You, Jee Young. 2001. Click and Learn: Fathom. Silicon
About the Author
Downes is a senior research officer, Institute for Information Technology,
National Research Council, Moncton, New Brunswick, Canada. Mr. Downes
specializes in research in online learning, online communities, and knowledge
management. He is USDLA Journal's Editor-at-Large. He may be reached at:
firstname.lastname@example.org. . Some of his work and research may be accessed as listed
Author/publisher, OLDaily, a daily newsletter:
Section Editor, The Technology Source: