This paper is a case study based on Steve Anderson's experiences at Rover Group, designing and building, an integrated SGML authoring, management and publication solution.
Rover's TiMS system brings together leading SGML editing tools with sophisticated repository software, to support the process of authoring and managing shared corporate information in a supportive, structured environment, and to enable that information to be published in a variety of formats. Rovers system also provides full support for translated text in all formats, radically reducing the cost of managing multilingual translations.
Rover Group Limited has been working for three years now on the development of TiMS (Technical Information Management System), a project which integrates a fully structured SGML authoring environment with a database containing information pools of SGML elements. TiMS supports the creation and publication of both narrative and graphical materials, and tracks all revisions to individual information units.
Much effort has been committed to devising an object-oriented approach to information management, which penetrates deeper than the document, to address individual information units. These information units, SGML elements which are tracked throughout their lives, are stored as discrete objects and shared many times between output publications.
Rover Group is now examining an evolution of this approach to address a corporate information warehousing strategy, which will enable many users within the enterprise to share a common corporate information pool.
The Rover Group is Britain's largest motor manufacturer, producing approximately half a million vehicles a year. The group designs, manufactures and markets small, medium, and executive cars and specialist four-wheel drive vehicles.
Major market areas are Western Europe, Japan, North America, and Australia, with an increasing presence in Asia and the Pacific Rim.
Rover Group is a major employer with 33,000 people working for the company in the UK and internationally. A further 110,000 people work within companies supplying Rover Group with production material and in the 2,500 dealers and distributors selling the company's vehicles world-wide.
Rover Group's products divide into two distinct vehicle ranges: Land Rover is the name for the company's world-famous four-wheel drive vehicles; the Rover marque has a reputation and tradition of refinement and performance which is continued today with the company's small, medium, and executive car ranges.
Major investment programmes in new models and advanced design and manufacturing techniques have resulted in a new, word-class generation of Rover and Land Rover vehicles for the 1990s. The company holds the BS5750 quality certification and many awards for product and manufacturing excellence.
In most companies that deal with the manufacture of component-based hardware, there is a department which is responsible for the production of technical information, aimed at helping technicians to maintain and repair units and their constituent components. Within Rover Group, the Technical Communications department is dedicated to the production of all the various types of technical information used in dealers workshops, as well as the customer-oriented information provided with a Rover car when it is delivered.
The information provided at a Rover dealership includes not only traditional hard copy, but information delivered electronically by Rover computer systems. The content of the electronic and paper versions of a particular information set may be similar, but the output format is very different, and the means by which the information is accessed and navigated differs radically.
Rovers technical information has historically been authored and managed on a variety of different, incompatible platforms. There were different publishing methods, and different authoring approaches, for substantially the same information: clearly, a lot of the effort expended in such a situation is redundant. Since Rovers technical information is published in many different languages, there are major translation overheads incurred by a revision to any existing publications, as well, obviously, as upon release of a new publication.
Since Rovers cars are designed with re-use of existing components in mind, it makes sense that the information relating to those components, and the major units they make up, should be prime for re-use also. This would help to relieve the authoring workload, at the same time as reducing storage costs for the information, by reducing (or even eliminating) unnecessary duplication; it would also help to reduce translation costs if, in the event of changes to a piece of information, the computer system could identify exactly what changes had taken place, and constrain the translator to re-translate only those elements.
The question at Rover was: How do we ensure that the information we author is re-usable in a style that actually helps the author, rather than adding to the burden of managing a pool of increasingly technical information? The answer was prompted by the imminent SAE J2008 legislation, which affects all automotive manufacturers who sell into the US market. J2008 prescribes the use of SGML as the data tagging protocol for delivery of electronic service-related information, where that information is pertinent to maintenance of emissions-related components of a car. For us, that meant we had to anticipate a need to provide electronic information, with SGML tagging, to any interested parties. Since there were no such facilities within Rover Group at the time, the decision was taken to develop an SGML authoring capability within Rover.
As well as the ability to create information in an authoring environment, Rovers Technical Communications department were looking for a number of very specific features in any new systems environment. Paramount among these were the ability to manage information units throughout a full life-cycle, the ability to search for and retrieve information by a variety of meaningful criteria, and the ability to share information among an extended workgroup in a controlled manner. Furthermore, the sources of technical information were to be various, and to include engineering data from manufacturing systems, and graphics imported directly from CAD systems. These aims were compatible with the use of SGML to provide a fully controlled technical information management system, incorporating end-to-end processing within the SGML structure, without compromise.
TiMS (Technical Information Management System) is a development project aimed at providing a coherent systems solution to authoring, management, and distribution of Rover's technical service information, and its publication on a variety of media, in multiple languages.
The approach we favoured in developing TiMS was to provide an integrated solution, combining COTS (Commercial Off-The Shelf) products and custom-developed applications in a client-server environment. The key ingredients of this integration are:
An SGML editor
A Repository Manager
A Database Management System
A Workflow Management System
We elected to produce the integration software and various low-level system functionality in-house, since this gave Rover the maximum of control over the application development.
The following sections give a description of the basic functionality of TiMS, with our reasons for choosing the various methods and tools.
It was necessary to provide some means of standardizing on publication structures. To facilitate this with a minimum of human effort, TiMS allows the creation, management and storage of SGML templates. Each template represents the 'ideal' contents of a publication of a particular type, based upon a DTD which represents the structure of that class of publication. Since there are many classes of publication in Rovers information set, there is tremendous benefit in allowing the assembly of instances 'automatically', based upon templates.
It was clear that the TiMS users needed to interact with each other in a collaborative groupworking environment. To enable this, we developed a set of bespoke groupworking and workflow management applications which enable the end user to import a project plan created in Microsoft Project, whereupon the system will automatically create a publication or publications in the TiMS database, based upon SGML document templates.
The TiMS user has the ability to edit the structure of the automatically created document instance, adding to, or modifying the instance according to requirements. The addition of new elements, or modification of existing, will give rise to Tasks in the workflow system, and these tasks are allocated by the system to the appropriate workgroups or individuals.
Individual users of TiMS log into their workflow application, which is the basic interface they work within while in the TiMS system. The Task Workbench is the launching pad for all the functionality required to carry out the authoring processes available in TiMS.
One of the primary requirements of TiMS users was to be able to carry out information authoring processes in a clean, user-friendly environment. It was felt that Rovers authors should be first and foremost engineers, and should not be required to understand SGML, particularly not to the extent that they needed to insert tags into the document instances. This decided us to go for a WYSIWYG style of graphical user interface (GUI). Using a standard GUI like MS Windows meant that, with a minimum of re-training, TiMS users could also benefit from a variety of other office applications on the same PC platform.
After some deliberation, we opted to use the GriF SGML editor in our integration. This was essentially due to the GriF products ability to run under MS Windows, its customisable application programming interface, and the WYSIWYN (What You See Is What You Need) presentation models it was possible to produce, which allow us to represent the same information onscreen in multiple ways, simultaneously if desired.
Another major consideration for us was that the editor had to be a native SGML product, to reduce the overheads (and potential cost) of carrying out a translation from SGML into some proprietary format upon loading an instance into the editor, and the reverse upon export. In a corporate networking environment, performance is at a premium, and the additional delays incurred by a non-SGML editing package were unacceptable.
One of our key business drivers is the reduction of duplication, both of effort in the creation of information, and of information storage. There is a need also to manage a changing information base though time. To satisfy these version control requirements, we clearly needed some kind of control mechanism which recognized the evolution of information units, and allowed us to track the changes to the information during its life in the system. Further than this, we needed the ability to reproduce any historically published information, in exactly the form it was published in when originally released.
Since we were providing the TiMS users with a collaborative authoring environment which allowed shared data to be retrieved, modified and published, there was also very clearly a need for the version control mechanism to act as a data repository manager. This would allow information units to be checked out for editing, and checked back into the database, all under full control of the repository manager.
The repository management tool we selected was Life*CDM, a dedicated SGML repository which operates on an Oracle 7 database. Life*CDM is produced by CRI (Computer Resources International), one of Rovers prime suppliers on the TiMS project. We have integrated Life*CDM into TiMS in such a way as to manage SGML elements as discrete information objects. Since Life*CDM is primarily a compound document management database, we have the flexibility to use its repository management capabilities at all levels, from the publication down to its component objects.
We are using an Oracle 7 relational database, which is the RDBMS selected by Rover Groups Systems Strategy group. Although we use the repository manager as an object oriented information management layer, the relational capabilities of the database are used fully, since many of the access methods used by TiMS and related systems depend upon data driven searches.
Much of the information published by TiMS is highly graphical in content. TiMS authors have the ability to include complex graphic figures, including call-outs, which are stored as SGML elements in the database. A single base graphic may be re-used any times throughout the information space, each instance with its own particular set of overlays.
The illustration package is integrated, like the SGML text editor, into the TiMS Workflow Manager. Authors can place a requirement for an illustrator to produce a particular graphic, or set of graphics, simply by inserting an element into the document instance, which 'reserves' a space for the graphic, which is filled by the graphics content automatically upon production of the graphic by an illustrator.
TiMS also enables creation and management of CALS tables, which the GriF editor allows to be edited in the WYSIWYG display. Data created in these tables is saved into the database, both for publication and for re-use by other systems.
The TiMS system maintains links between the original, English-language versions of information units, and the translated versions of those information units, in many different languages. The system also recognizes when the PCDATA content of an information element has changed, which gives rise to a translation task for each of the required languages for that element.
The translation task is directed toward the relevant translator (using an extension of the TiMS Workflow Manager functionality), who uses a TiMS client application to provide the newly translated information unit as an SGML instance which is then checked back into the TiMS repository under full configuration control.
Because the TiMS translation toolkit is developed as an extension of the TiMS system, and includes similar functionality (including the GriF editor), the translators benefit from the same controlled, supportive environment as other TiMS users, and work with a WYSIWYG representation of the information, rather than the SGML tags. This means that Rover has much more flexibility in terms of selection of translation resources, since learning overheads are low.
TiMS makes full use of SGMLs ability to separate content from format, in order to allow the re-use of information units in a variety of output formats, on a variety of media. This means that the same information can be re-used in different publications, some of which are dedicated to technicians, others of which are aimed at the car-owning public, some of which are on paper, and some of which are delivered electronically.
All the information formatting is carried out upon extraction from the database by the publication process, which features fully automatic rules-based application of publication styles.
TiMS captures information from engineering sources, and some of the information entered into TiMS is fundamentally engineering data. This data is passed from TiMS onward, to other systems in the Rover dealership, and to systems within Rover itself. The approach we have taken in building our DTDs allows us to capture far more information into an SGML element than ever is published in a single output, which maximizes the benefit gained from the 'author once, use many' approach.
A full approval cycle is supported by the TiMS applications, which feature functionality designed to ensure completion and support quality checking, and quality assurance, by an editing authority within the workgroup. A finite state machine tracks the status of every object on the TiMS database, and only upon achievement of full approval can a publication move to the point where it can be published fro the system.
We are confident that the TiMS system represents a major achievement, in that it provides a very complete set of functionality to the end user. With TiMS, we have provided a fully controlled and supportive working environment, in which users of the system can work unimpeded by the need for a deep understanding of the underlying SGML, and are allowed to pursue their primary business, which is the creation of high-quality technical information.
Steve is the Project Manager on Rover's TiMS project. He was recruited by Rover Group Systems for his specific skills in computer systems design, and the management of systems development projects. He took over the management of TiMS at the end of 1994.
Prior to joining Rover Group in 1993, Steve was Principal Analyst with a Birmingham-based house, CFC, specialising in the development of bespoke and packaged PC database sofware in a networked environment.
Steve has produced computer systems for customers including Save the Children, Korps Rijkspolitie (Dutch State Police), Nationwide Building Society, Ernst & Young, ACP Rent-a-Car (Lisbon), and the Maxwell group of companies. He has also worked in the IT departments of the GKN Group of Companies, Walker Group of Companies (WSS Information Services) and British Steel Distribution.
Steve is a participant in Rover discussion groups examining the potential benefits behind electronic management and distribution of information in a commercial environment.