IRMJ01mcmanus Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 59 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. ABSTRACT In this paper we present the evolution of a collaborative courseware generating system that is featured by XML-based course structure representation, JSP-based dynamic courseware presentation, and WebDAV-based collaborative courseware authoring. While the first system implementation employs a proprietary design using a self-defined XML DTD to represent the course structure, the second and the third system implementation take an open-standard- oriented approach, which are respectively SCORM 1.1 and SCORM 1.2 conformant. In the latter two implementations, all learning resources contained in an existing Java course are re-designed according to the SCORM 1.1 and SCORM 1.2 Content Model and further annotated with corresponding SCORM metadata. In addition, the course structure is re-constructed utilizing SCORM 1.1 Content Structure Format and SCORM 1.2 Content Packaging Specification. The evolution of the collaborative courseware generating system is motivated by our efforts to improve the reusability and interoperability of learning resources. Keywords: eXtensible markup language, Web-based Distributed Authoring and Versioning, Java Server Pages, Sharable Content Object Reference Model Towards Open Standards: The Evolution of a Collaborative Courseware Generating System Changtao Qu and Wolfgang Nejdl, University of Hannover, Germany INTRODUCTION Since the summer semester 1999, the joint CS1 course “Introduction to Java Pro- gramming” (Info1 for short) has been shared among three German universities and one university in Italy. During the past three years, we have been successively working on three system implementations of Info1 with the purpose of exploring effi- cient approaches to improving the reusabil- ity and interoperability of learning re- sources. While the first system implemen- tation employs a proprietary design using a self-defined XML (eXtensible Markup Language) DTD (Document Type Defini- tion) to represent the course structure, the second and the third system implementa- tion take an open-standard-oriented ap- proach, which are respectively SCORM (Sharable Content Object Reference Model) 1.1 (ADL Technical Team, 2001) 701 E. Chocolate Avenue, Hershey PA 17033, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com ������� INFORMATION SCIENCE PUBLISHING This chapter appears in the journal, International Journal of Distance Education Technology, 1(3), edited by Qing Li and Weijia Jia. Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. 60 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. and SCORM 1.2 (ADL Technical Team, 2001a) conformant. In the latter two imple- mentations, all learning resources contained in Info1 are re-designed according to the SCORM 1.1 and SCORM 1.2 Content Model and further annotated with corre- sponding SCORM metadata. Also the course structure is re-constructed utilizing SCORM 1.1 CSF (Content Structure For- mat) and SCORM 1.2 CP (Content Pack- aging) Specification. In the following we will present these three system implemen- tations of Info1, showing its evolution to- wards open standards. GENERAL DESIGN In Figure 1 we illustrate the general infrastructure of the collaborative courseware generating system. In general, the system is constructed from a WebDAV (Web-based Distributed Authoring and Versioning) based courseware authoring module and a JSP (Java Server Pages) based courseware publishing engine. The standard data inter- face between both is XML. Although the general infrastructure is commonly shared by all three system imple- mentations, there are several essential dif- ferences between them. First of all, the three system implementations are differ- ent in how they represent the course struc- ture using XML. This essential difference clearly marks the system’s evolution to- wards open standards. Moreover, the dif- ferent representations of the course struc- ture also determine the reusability of the JSP-based courseware publishing engine that is responsible for dynamically present- ing the XML-based course structure on the Web. In Figure 2 we firstly illustrate a com- mon module of all three system implemen- tations: the WebDAV-based courseware authoring module. It is used to support col- laborative courseware authoring in three system implementations. The courseware authoring module comprises a WebDAV-based courseware repository used to store course script files, and an XML file used to represent the course structure. The latter also serves as the standard data interface between the courseware authoring module and the courseware publishing engine in order to cleanly separate course content from the courseware presentation. The WebDAV- based courseware authoring module is shared by all three system implementations of Info1, which can enable geographically dispersed authors to collaboratively accom- plish the courseware authoring process. WebDAV (Goland, Whitehead, Faizi, Carter, & Jensen, 1999) is an IETF speci- Figure 1: The General Infrastructure of the Collaborative Courseware Generating System WebDAV- based courseware authoring module JSP-based courseware publishing engine XML-based course structure representation WebDAV-based courseware authoring module Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 61 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. fication originally designed to add interoperability and collaborative capabili- ties to the Internet. It provides sets of ex- tensions to the HTTP protocol that allows geographically dispersed users to collaboratively edit and manage documents directly on the remote server. The current functionalities of WebDAV include: (1) lock- ing mechanism, used to prevent the “over- writing” of changes in a distributed, multi- user authoring environment; (2) namespace manipulation, used to manage document repository on the remote server; (3) prop- erty manipulation, used to handle XML- based metadata of documents; and (4) col- lections, used to create sets of related docu- ments and to retrieve listing of their mem- bers. Utilizing WebDAV, the courseware authors can “in-place” (directly on the re- mote server) implement most activities needed for collaborative courseware authoring, e.g., editing course script files stored in the courseware repository, ma- nipulating repository’s namespace, utilizing locking mechanism to prevent the “over- writing,” or manipulating properties of a specific course script file in order to ex- change ideas and opinions among authors. In fact, according to our practical experi- ence, the WebDAV-based courseware authoring module has greatly improved the efficiency of the courseware authoring pro- cess (Qu, Gamper & Nejdl, 2001). THE FIRST SYSTEM IMPLEMENTATION: PROPRIETARY DESIGN The first system implementation of Info1 adopted a self-defined XML DTD to represent the course structure. In Fig- ure 3 we illustrate this XML DTD. In the DTD definition, several self- defined XML elements, e.g., “CourseUnit,” “CourseElement” are adopted to describe the course structure. Also the metadata of the course scripts (e.g., URIs or URLs) are described in these elements in the form of “attributes.” Although principally this is a proprietary approach to representing the course structure, we can still achieve a certain reusability of the courseware pub- lishing engine based on this DTD. Actu- ally, all courseware represented using the above XML DTD can be directly rendered by the JSP-based courseware publishing engine without the need of any re-configu- ration process. Here we refer readers to our previous publication (Qu, Gamper & Figure 2 : The WebDAV-Based Courseware Authoring Module To publishing engine XML-based course structure WebDAV-based course script repository 62 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Nejdl, 2001) for a more detailed descrip- tion of the courseware publishing engine designed in the first system implementa- tion. THE SECOND SYSTEM IMPLEMENTATION: SCORM 1.1 CONFORMANT DESIGN Although we have achieved certain reusability of the courseware publishing engine thanks to the inherent flexibility of XML, the first system implementation has two notable drawbacks. First, it is propri- etary. On the one hand, the course struc- ture represented using the self-defined XML DTD cannot be directly rendered by other courseware publishing engines. On the other hand, the courseware publishing engine bound to the self-defined XML DTD cannot be re-used to generate other courseware represented using other XML formats. Second, the metadata of learning resources contained in Info1 are not anno- tated and managed in the first system implementation, which makes it very diffi- cult to reuse and exchange learning re- sources between our partner universities. Therefore, in order to achieve more interoperability, especially in order to find an efficient way to reuse and exchange learning resources, we decided to shift to an open standard: SCORM 1.1 in the sec- ond system implementation. The SCORM 1.1 was released by ADL (Advanced Distributed Learning) in January 2001. One of the most important features of SCORM is its good compatibil- ity with other learning resource specifica- tions. The SCORM 1.1 smartly references IMS Learning Resource Metadata Speci- fication (IMS, 2001a) (in SCORM 1.2, also IMS Content Packaging Specification (IMS, 2001)) and IEEE LOM (Learning Object Metadata) (IEEE LTSC, 2001) as well as other specifications and further in- tegrates these specifications with one an- other to form a more complete and easier- to-implement model. With regard to metadata sets, the SCORM 1.1 is down- Figure 3: Self-Defined XML DTD Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 63 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. wards compatible with IEEE LOM 3.5 and IMS Metadata Specification 1.1. Regard- ing Content Structure representation, it de- fines SCORM 1.1 CSF, which itself is de- rived from AICC CMI CSF (AICC, 2001). The SCORM 1.1 also defines a Content Model consisting of three components: Raw Materials, SCO (Block), and Course. Together with the metadata specification and CSF, the Content Model can enable the reuse and exchange of learning re- sources at different aggregation levels. More importantly, the SCORM 1.1 also provides a RTE (Run-Time Environment) that offers a standardized way for SCO (Sharable Content Object)-based learning resources to communicate with an LMS (Learning Management System) through the use of common API. During the devel- opment process, the RTE can provide us with the beneficial guidance to the system implementation. In general, the SCORM 1.1 conformant design of the second system implementation consists of four tasks: (1) Adapting existing learning re- sources into the SCORM 1.1 Content Model. The learning resources contained in Info1 include not only some self-made “internal” materials, but also lots of “exter- nal” learning resources that directly exist on the Web. According to the SCORM 1.1 Content Model, these “internal” and “ex- ternal” learning resources are reasonably designed as Raw Materials, SCO (Block), and Course in the second system imple- mentation, as depicted in Figure 4. During the system design, we’ve given a special consideration to the differ- entiation between Raw Materials and SCOs. While each course unit of Info1 can be naturally designed as an SCO and all its underlying raw materials (e.g., figures, tables, etc.) can be naturally designed as Raw Materials, the “external” resources have to receive more attention while being adapted into the SCORM 1.1 Content Model. Because the SCO represents the lowest level of granularity of learning re- sources that can be tracked by an LMS using the SCORM RTE, and also the SCO itself must be independent of learning con- text, we intentionally designed all “exter- nal” learning resources as Raw Materials in order to retain some reasonable learning context between “external” resources and SCOs (course units) ( Qu & Nejdl, 2002). Additionally, we have also organized sev- eral SCOs into higher aggregations (Blocks), which can further facilitate the Figure 4: The SCORM 1.1 Conformant Info1 Block Raw Materials Raw Materials Course SCO 64 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. reuse and exchange of learning resources at different aggregation levels. (2) Representing the course struc- ture using SCORM 1.1 CSF. The SCORM 1.1 employs CSF to aggregate learning resources into a cohesive unit of instruction, e.g., course, lesson, and mod- ule, etc. In comparison to the use of self- defined XML DTD in the first system implementation, representing the course structure using SCORM 1.1 CSF consti- tutes the key to our shift from proprietary design to open-standard-oriented develop- ment. On the one hand, the CSF-based course structure can be now directly ren- dered by any other SCORM 1.1 conformant courseware publishing engines; on the other hand, our courseware publishing engine implemented in the second system imple- mentation can be now re-used to generate other SCORM 1.1 conformant courseware. In Figure 5 we illustrate the SCORM 1.1 CSF-based course structure representation of Info1. It could be directly rendered by any SCORM 1.1 conformant courseware publishing engines. (3) Annotating and managing learning resource metadata. In order to facilitate the reuse of learning resources, all learning resources contained in Info1 are annotated with SCORM 1.1 metadata on the basis of three aggregation levels (Raw Materials, SCO (Block), and Course). Dur- ing the metadata annotation process, we’ve paid special attention to the metadata’s compatibility with other popular specifica- tions while still retaining 100% compatibil- ity with the SCORM. The SCORM 1.1 Metadata Information Model is broken up into nine categories: General, Lifecycle, Meta-Metadata, Technical, Educational, Rights, Relation, Annotation, and Classifi- cation. Besides complying with all guide- lines provided by the SCORM “best prac- tice” for each category, we applied the ACM Computing Classification System (http://www.acm.org/class/1998) in the “Classification” category, which fits very well to describe learning resources at the “ontology” or “terminology” level. Also in the “Relation” category, the relationships between three aggregation levels are de- scribed using “HasPart,” “IsPartOf,” etc., which nicely reflects the course structure at the “metadata” level. In addition, in order to effectively manage the learning resource metadata, we choose a native XML database: Apache Xindice (http://xml.apache.org/xindice) to store SCORM metadata Application Pro- Figure 5: The SCORM 1.1 CSF-based course structure representation of Info1 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 65 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. files. As a so-called native XML database, Apache Xindice provides a natural way to store, retrieve, update, search, and discover SCORM metadata. In Xindice, all metadata Application Profiles are stored in their origi- nal XML format without the need for any pre-processing. The search and update of metadata can be easily accomplished, tak- ing advantage of W3C XPath language and XUpdate language from XML:DB Initia- tive. (4) Constructing the SCORM 1.1 RTE. The SCORM 1.1 RTE actually serves as the new courseware publishing engine in the second system implementa- tion. It takes the SCORM 1.1 CSF as the input and then dynamically generates the courseware presentation on the Web. In Figure 6 we illustrate the infrastructure of the SCORM 1.1 RTE. The SCORM 1.1 RTE is constructed on a JSP&Servlet-enabled Web server: Apache Tomcat 3.2.3. On the server side, a JSP component is used to dynamically render the SCORM CSF-based course structure into the navigation menu depicted in the left frame of Figure 4. Additionally, on the server side there are also several Java Servlet components responsible for controlling actual sequencing of SCOs, han- dling the communication between RTE and SCOs (e.g., getting and setting Data Model), and managing the persistence of Data Model. On the client side, a non-face Java Applet is implemented as the SCORM RTE API Adapter and embedded in the left frame of Figure 4. This API Adapter Applet provides the communication to the RTE server-side Servlet components for Data Model persistence management. Note that on the client side, the SCOs cannot make direct communication with the RTE server to call API functions. All calls from SCOs must take the API Adapter as a broker and use client-side JavaScript. Moreover, all learning context existing within an SCO must be managed by the SCO itself using embedded client-side JavaScript. THE THIRD SYSTEM IMPLEMENTATION: SCORM 1.2 CONFORMANT DESIGN At the beginning of October 2001, we began to develop the third system imple- mentation inspired by our desire of pursu- ing more openness and interoperability of the collaborative courseware generating system. The third system implementation is based on the SCORM 1.2, released by ADL in October 2001. In comparison to the SCORM 1.1, the SCORM 1.2 has sev- eral important improvements. Regarding metadata specification, the SCORM 1.2 sits on a higher level than SCORM 1.1, offer- ing downwards compatibility with IMS 1.2.1 (instead of IMS 1.1 in SCORM 1.1) and IEEE LOM 6.1 (instead of LOM 3.5 in SCORM 1.1). With regard to the Con- tent Structure representation, the SCORM 1.2 deprecates SCORM 1.1 CSF and pro- vides a CP specification which is derived Figure 6: The Infrastructure of the SCORM 1. 1 RTE API CSF-based course structure Learning resource repository JSP and Java Servlet Apache Tomcat 3.2.3 Web server API Adapter (Java Applet) SCO JavaScript Web browser Data persistence 66 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. from the IMS CP specification 1.1.2. As a matter of fact, the use of SCORM 1.2 CP enables a new functionality of the collabo- rative courseware generating system. That is, on the basis of the SCORM 1.2 CP, learn- ing resources contained in Info1 can be physically packaged and unpackaged. This will greatly facilitate the exchange of learn- ing resources between different LMSs. In general, in order to shift the sec- ond system implementation to the third one, we have to fulfill four tasks: (1) Transferring learning re- sources from SCORM 1.1 Content Model to SCORM 1.2 Content Model. Since the SCORM Content Model remains almost untouched from version 1.1 to 1.2, except for several nomenclature changes, the transferring process is relatively straight- forward. (2) Representing the course struc- ture using SCORM 1.2 CP. The SCORM 1.2 CP extends the latest IMS CP specifi- cation with several additional SCORM-spe- cific elements, particularly in the “organi- zation” section where SCORM 1.2 Con- tent Structure is located. By means of such sort of extension, the SCORM 1.2 CP can effectively define the structure and the in- tended behaviour of a collection of learn- ing resources along with the 100% down- wards compatibility with the IMS CP. In comparison to the second system imple- mentation, in which the course structure is represented using SCROM 1.1 CSF, rep- resenting the course structure using SCORM 1.2 CP in the third system imple- mentation can achieve more interoperability thanks to the higher popularity of IMS CP. More importantly, because the course structure is now self-contained described in a SCORM 1.2 CP Application Profile, including all descriptions of dependency and relationships existing between learning re- sources, not only those “internal” resources existing physically in a package and de- scribed by URI, but also those “external” resources existing on the Web and described by URL, all learning resources in Info1 can now be exchanged between different LMSs based on the SCORM 1.2 CP, either par- tially or as a whole. Such sort of exchange, namely, importing, exporting, aggregating, or disaggregating packages of learning re- sources, makes it feasible to reuse the learning content at various aggregation lev- els. Figure 7: The SCORM 1.2 CP Application Profile of Info1 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 67 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. As an example, in Figure 7 we illus- trate the SCORM 1.2 CP Application Pro- file of Info1. Based on this CP Application Profile, Info1 can be not only physically packaged and unpackaged, but can also be dynamically presented on the Web by any SCORM 1.2 CP (also IMS CP) conformant courseware publishing engines. (3) Annotating and managing learning resource metadata. Because IEEE LOM, the cornerstone of SCORM 1.2 metadata specification, has experienced considerable changes from version 3.5 to version 6.1, all SCORM 1.1 conformant metadata generated in the second system implementation have to be modified accord- ing to the SCORM 1.2 metadata specifi- cation in the third system implementation. Fortunately, since the SCORM Content Model remains almost unchanged from version 1.1 to 1.2, we only need to concen- trate on the syntax changes during the transferring process. In addition, regarding the metadata management, the architec- ture of Apache Xindice-based metadata repository need only several slight modifi- cations. (4) Constructing the SCORM 1.2 RTE. Since the RTE definition remains unchanged from SCORM 1.1 to SCORM 1.2, all implemented functionalities in the SCORM 1.1 RTE can be directly trans- ferred to the SCORM 1.2 RTE implemen- tation. However, since the course struc- ture is now represented using SCORM 1.2 CP, we need to modify the “parser” JSP in order to handle the SCORM CP Applica- tion Profile. Also the new functionality of SCORM 1.2, namely, physically packag- ing and unpackaging learning resources based on the SCORM 1.2 CP Application Profile, needs to be developed. CONCLUSIONS In Table 1 we list a brief comparison between three system implementations. The evolution of the collaborative courseware generating system is actually motivated by our efforts to improve the reusability and interoperability of learning resources. From the proprietary design in the first system implementation to the SCORM-based development in the second and the third system implementation, our system always evolves towards open stan- dards and has become increasingly open and interoperable. Currently the exchange of learning resources based on the second and third system implementation has been Table 1: A Comparison Between Three System Implementations 1st Version 2nd Version 3rd Version WebDAV-based collaborative courseware authoring Yes Yes Yes Course structure representation & courseware publishing engine Non- interoperable Interoperable with SCORM 1.1 CSF and AICC CMI CSF Interoperable with SCORM 1.2 CP and IMS CP 1.1.2 Metadata annotation and management N/A Conformant with SCORM 1.1, IMS 1.1, and LOM 3.5 Conformant with SCORM 1.2, IMS 1.2.1, and LOM 6.1 Exchange learning resources N/A Yes, based on SCORM 1.1 Content Model Yes, based on SCORM 1.2 Content Model Physically package & unpackage learning resources N/A N/A Yes, based on SCORM 1.2 CP or IMS 1.1.2 CP Courseware interactivity N/A Yes, based on SCORM 1.1 RTE Data Model (AICC CMI Data Model) Yes, based on SCORM 1.2 RTE Data Model (AICC CMI Data Model) 68 Journal of Distance Education Technologies, 1(3), 59-68, July-Sept 2003 Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. already underway between several Ger- man universities and institutions. Also the SCORM 1.1 and SCORM 1.2 conformant metadata repositories are now being inte- grated into an e-learning peer-to-peer net- work, Edutella (http://edutella.jxta.org), with the purpose of further improving the reusability and interoperability of learning resources. REFERENCES ADL Technical Team (2001). SCORM V1.1. http://www.adlnet.org ADL Technical Team (2001a). SCORM V1.2. http://www.adlnet.org AICC (2001). CMI 001-AICC/CMI Guidelines for Interoperability V3.5. http://www.aicc.org/ Goland, Y. Y., Whitehead, E. J., Faizi, A., Carter, S., & Jensen, D.(1999). HTTP Extensions for Distributed Authoring- Changtao Qu is currently a Ph.D. candidate in Computer Science at the Learning Lab Lower Saxony, University of Hannover, Germany. He received his B.S. and M.S. degrees in Electrical Engineering respectively from Tianjin University, P. R. China in 1990 and Harbin Institute of Technology, P. R. China in 1995. His current research interests include learning management system, distributed computing, and the Semantic Web. Contact Mr. Qu at: Learning Lab Lower Saxony, University of Hannover, Expo Plaza 1,D-30539, Hannover, Germany. Email: qu@learninglab.de. Wolfgang Nejdl has been Full Professor of Computer Science at the University of Hannover since 1995, after being Associate Professor of Computer Science at the RWTH Aachen (1992-1995). In 1988, was Visiting Researcher at Stanford University, in 1992 to Xerox PARC, 2001 and 2002 again at Stanford University. The Institute for Information Systems/ Knowledge Based Systems headed by Dr. Nejdl does research in the areas of artificial intelligence, adaptive hypermedia systems, as well as metadata-based learning repositories and peer-to-peer systems. Dr. Nejdl has published more than 140 conference and journal publications in these areas and has been a member of numerous program committees and editorial boards in these areas. He is member of two ESPRIT Networks of Excellence in the areas of artificial intelligence and the Semantic Web. He is Director and Founding Member of the Learning Lab Lower Saxony (L3S), a unique German competence center focussing on learning technology research, with members from five universities. Contact Dr. Nejdl at: Learning Lab Lower Saxony, University of Hannover, Expo Plaza 1,D-30539, Hannover, Germany. Email: nejdl@learninglab.de. WEBDAV. RFC 2518 IEEE LTSC (2001). IEEE LOM Working Draft 6.1. http://ltsc.ieee.org IMS (2001). IMS Content Packag- ing Spec. V1.1.2. http://www.imsproject. org IMS (2001a). IMS Learning Re- source Metadata Spec. V1.2.1. http:// www.imsproject.org Qu, C., Gamper, J., & Nejdl, W.(2001). A Collaborative Courseware Generating System based on WebDAV, XML, and JSP. in Proc. of the 1st IEEE International Conference on Advanced Learning Technologies, Madison, USA. Qu, C., & Nejdl, W.(2002). Towards Interoperability and Reusability of Learn- ing Resource: a SCORM-conformant Courseware for Computer Science Edu- cation. in Proc. of the 2nd IEEE Interna- tional Conference on Advanced Learn- ing Technologies, Kazan, Russia.