|
 Technology
& Market Needs
TELCERT was borne out of longstanding
market needs: from learning technologists, standards communities
and researchers and providers of services to support a elusive
goal – interoperability
of content, services and learning systems that could
be assured from the start
The use of technology to support learning has advanced
significantly in recent years, both in terms of the increasing
sophistication of online learning services and the availability
of open specifications which have brought about convergence
in the techniques used by vendors to achieve systems integration
and in particular content portability. There are many products
which offer interpretations of the specifications, but
there are currently no independent, third-party mechanisms
for assessing conformance to a common interpretation of
the specifications, or a widely acknowledged reference
implementation against which people can compare their efforts.
Implementations often implement differing subsets of the
specification, add extensions, use alternate bindings (e.g.
Resource Description Framework, versus XML schema) or are
implemented upon differing underlying technologies (e.g.
JAVA versus .Net, SOAP, …).
Top of Page 
Barriers
to Interoperability
There are costly drawbacks for vendors and customers in
the current situation:
- Vendors have to design and implement their
own test suites, usually concentrating on the specific
ways in which they interpret and use the open specifications.
- Customers setting up
learning management systems are faced with the need that
minor incompatibilities between systems from different
vendors add up, have to be analysed and must be overcome
by developing case-specific mediation software.
The Advanced Distributed
Learning Initiative has piloted testing with their SCORM
test system, but this addresses a single profile of the
specifications, aimed at meeting the training needs of
the US military. As is evident by the growing number
of implementations around the world, online learning
is not a ‘one size fits all’ undertaking.
Both vendors and users seek reassurance that the desired
interoperability between implementations can be achieved,
such as that provided through the scrutiny of a standard
conformance test suite. Conformance testing needs to
be allied with the capability to adequately service localisation
issues required to meet the full requirements of user
communities. This is of special importance in Europe
with its large diversity of culture and learning settings.
These issues are compounded by the fact that new technologies
and methodologies are gaining popularity amongst implementers,
and these in and of themselves raise challenges for any
conformance testing which might be put in place in this
domain.
Top of Page
TELCERT
Objectives
Thus the Scientific and Technical objectives of the TELCERT
project are:
- Communicate the novel techniques and technologies
(e.g. JAVA, UML, XML and new code/API stub generation
tools) being adopted for implementation of technology
enhanced learning products and services
- Raise awareness of how to define community requirements
in a consistent manner using Application Profiling techniques
- Demonstrate how state-of-the art research into
testing can be exploited to define flexible conformance
testing strategies, able to accommodate application and
data localisation
- Assist the content industry and tool vendors in
supporting Application Profiles in their commercial offerings
which will successfully pass the conformance tests
- Execute a coordinated plan of dissemination activities
to aid European and International adoption of the project
results
- Achieve post-project adoption of the conformance
testing developed and the Application Profiles this represents
Top of Page
XML,
UML and Standards Development
It is established practice by Learning Technology Specification
and standardization organizations to develop a specification/standard
by:
- Defining an appropriate ‘Information Model’ that
describes the core data objects and their properties;
- Defining the binding of the information model onto
an appropriate data representation. This typically
takes the form of XML (Extensible Mark-up Language) and, to
a lesser extent, RDF (Resource Description Format);
- Describing the recommended
practice and implementation when using the binding to
support a particular application.
The state-of-the-art
followed by organizations such as IMS is to describe
the Information Model using the Unified Modelling Language
(UML); UML is the internationally recognized object-oriented
software system design methodology. The usage of UML
allows a specification to be developed that includes
a behavioural description as well as just the data model
as normally expressed in an XML-based specification.
UML support tools such as Rational Rose, TogetherJ and
Poseidon allow a UML description to be saved in various
forms including JAVA. This means that an appropriately
constructed UML-based specification can be readily exported
as an Application Programming Interface (API). Most UML
tools use the XML Metadata Interchange (XMI) as their
interchange format, thereby enabling different tools
to edit a specification. There is a growing base of knowledge
on how best to represent the details of a UML description
in XML and XML-based formats and XML is at the core of
most Learning Technology Specifications.
The adoption of behavioural descriptions is particularly
significant because it enables the creation of robust
Conformance Specifications. This in turn allows the
development of a clear test strategy and the enforcement
of compliance requirements. The conversion of a UML description
to XML requires the mapping of the class operations onto
the corresponding response/request messages. These
messages are themselves XML documents that are exchanged
using an appropriate messaging infrastructure e.g. SOAP+HTTP/HTTPS.
The state-of-the-art is now focused on using WSDL (Web
Services Description Language) as the binding of a
specification’s
information model to a web services application.
In most learning systems, several specifications/standards
have to be combined to support a particular application.
Even in the case of behavioural-oriented specifications/standards
this requires the tailoring of the specifications to
the particular application domain; the result is called
an ‘Application
Profile’ (AP). An AP enables the intrinsic
flexibility of a specification/standard to be removed
to fit a particular domain; Learning Technology Specifications/standards
are designed to support any practice, which make
it impossible to construct a suitably robust Conformance
Specification for compliance. Therefore, an AP is
crucial for the creation of a useful and realizable
Conformance Specification.
Top of Page
Technology
Challenges for TELCERT
- How should a Learning
Technology Specification best be represented in UML to
facilitate the derivation of the corresponding Conformance
Specification? This assumes specifications that are based
upon behavioural descriptions and/or data models;
- What
is the best mechanism for the tool-based translation
of a UML description into its equivalent XML, WSDL and
JAVA bindings? This enables the support of multiple APIs
using a single Information Model;
- How should a Learning Technology Specification be
constructed to facilitate the preparation of an
Application Profile and the associated Conformance Specification?
This requires the development of a core specification from which
the Application Profiles are derived;
- What is the best approach to minimize the number
of Application Profiles created, for example, due
to geographic localisations? It may be possible to identify and agree
a core Application Profile or at least maximize reuse of elements
across APs;
- What is the best way to construct content for an
Application Profile that can be used to test the
degree of conformance of a system to a particular Conformance Specification
and how easy is it to re-author that content to
support a different Application Profile.
Top of Page
Technology
Options & Directions
Prior experimentation in implementing
Content Packaging Tools has demonstrated that any such
tool, in the context of the TELCERT development, would
have to be capable of being configured on-the-fly to (1)
a given user-community specification AP, (2) user interface
language and (3) backwards support of previous release
versions of the specifications included in a given profile.
Significant flexibility and ease of modification by both
content creators and test administrators is therefore required.
A candidate approach offering the capability to support
this diversity is to provide a Content Re-Engineering Tool
with a common core engine that supports a DOM, with XML
import and export, local persistence and remote communications,
with a defined API for interfacing with a plug-in editor.
The plug-in editors are driven by the relevant XML Schema,
together with a set of additional XML ‘helper’ files
that support the editor UI and map the UI elements to the
appropriate DOM elements. However, there is a real need
for robust testing against diverse profiles (supported
specifications, language, vocabulary substitution, element
sub-sets) in order to validate this approach.
Providing a reference implementation of such a tool,
mirroring the test system, will significantly accelerate
the process of getting tool developers up to speed with
developing products that will successfully complete conformance
testing.
Existing test methodologies, such as are described in
ISO 9646 (OSI Conformance Testing Methodology and Framework)
and IEEE 1003.3 (POSIX test methods) address testing of
protocols and APIs. ISO 9646 also includes a formal, processable
test notation (TTCN3) which has been extended to address
object testing. These standards underlie the tests that
industry has developed for UNIX, Bluetooth, GSM, and other
ETSI standards, WAP and many others. But very little has
been done in testing XML content and processors beyond
the core W3C specifications.
The test suites developed for WAP Forum and the recent
Open GIS test engine both developed by The Open Group include
elements that may be applicable to the learning technology
test problem as they are dealing with XML based specifications
as well as providing the test execution infrastructure
necessary to support conformance testing.
Test tools must of course support all of the operational
requirements for conformance testing, and must also deal
with the requirement for testing to Application Profiles
that may be unique to a given community. Rather than expect
those communities to retain the skills needed to implement
tests specific to their needs, the test system comprised
of the test tools developed in this project must be able
automatically to adapt and extend their test coverage by
processing the Application Profile descriptions.
The already mentioned UML is becoming the de facto industrial
standard for software system modelling and documentation,
while the recent massive initiative of the OMG to Model
Driven Architecture pursues a complete separation between
the abstract specification of system functionality and
that of its implementation on a specific technology platform.
This separation between the abstract system specification
and an implementation for a given platform is central to
the approach defined in TELCERT for supporting multiple
Application Profiles against a given set of Learning Technology
Specifications.
Top of Page
Innovations
in Software Testing
This project will also provide a significant twist to research
in software testing and validation. Nowadays huge attention
is attracted by the use of explicit models in software
development, as opposed to implicit or undefined models
using more traditional software design approaches. After
having been almost ignored for decades by practitioners,
rigorous specification and design methods finally enter
the industrial laboratories as a strategic means for developing
higher quality and more reliable systems. Considering the
adopted modelling notations, key features for exploration
are:
- formality, so that vital system properties
can be automatically verified, using, e.g., model-checking
techniques
- provisioning for adequate abstraction and composition
mechanisms, so that complex, open and evolving systems
can be effectively described
- standardisation, to counteract the proliferation
of dialects and to favour interoperable interchangeable
models.
|
