SGML as a vehicle for porting hypertext applications between systems

SGML as a vehicle for porting hypertext applications between systems
Eve Wilson
University of Kent at Canterbury
E.Wilson@ukc.ac.uk
Peter D. Shepton
University of Kent at Canterbury (at time of project)
Keywords: SGML, hypertext conversion, Mark-It

Objective
The objective of this project was to investigate SGML as a vehicle for porting documents between diverse hypertext systems. The documents were provided by European Passenger Services (EPS) who issue them in a traditional printed format. They form two major publications:
factsheet giving information about EPS and its services. It consists of text interspersed with images and tables;
an educational package for children at primary school. This contains a higher proportion of images but also questions to test reading comprehension and simple mathematics.
The two types of material were very different and were treated accordingly. The aim with the factsheet was merely to improve the accessibility and browsability of the information through an electronic text while the teaching material demanded more genuinely interactive features that would convert it into a Compiler Assisted Learning Package.
Tagging and Parsing
However, the first stage with both data sets was to define a Document Type Definition (DTD) in terms of the predefined structural components (ie chapters, sections, paragraphs and sentences) and to tag every occurrence of each element in the document. Subsequently, the documents were validated using Mark-It, an SGML parser which reads a DTD and checks that a given document instance conforms with it. Mark-It can also be used to produce a canonical form of the document, in which any markup missing owing to tag minimisation and use of short-refs is replaced. Finally, and most importantly, it has a generator or link mechanism that will facilitate systematic replacement of bugs (and content) in the canonical version. This can be used to replace the SGML tags by formatting instructions to produce a traditional printed text or, more ambitiously, to produce hypertext versions of the document. It relies on the introduction into the DTD of a generator section the application control and link processes that will be used by Mark-It to transform an EPS document instance from SGML into another format.
Target Hypertext Systems
In the project, documents were converted to:
source format for the Unix Guide hypertext system,
HTML format for browsers such as Netscape and Mosaic, and
PC Guide.
Each of these threw up individual problems which demanded independent solutions and threw doubt on the viability of using a single DTD for multiple purposes. For example, the Unix Guide markup produced a concise hypertext document that could be navigated by selecting hypertext buttons and using the scroll bar. For aesthetic reasons, and to improve functionality, the Guide file was split into smaller files and linked using usage buttons, rather than replace buttons. Because Guide naturally uses expansion buttons, the "contents page" of the original document was redundant and the contents element was removed from the DTD.
However, HTML needed a "contents" page to implement hypertext links to link each topic to the relevant section within the document. To maintain DTD consistency, a contents page was generated automatically together with the HTML document file. This, while increasing generator complexity, was inherently pleasing because it meant the contents section did not have to be manually maintained and updated. Again, functionality was enhanced by the use of many small files, rather than a single big file.

PC GUIDE offered the most comprehensive markup possibilities, although it required the use of an intermediary GUIDE Writer in the conversion. Firstly, Mark-It was used to convert SGML into Hypertext Markup Language (HML). HML, not to be confused with HTML, is a subset of SGML containing 27 element tags, that can be used to describe document structure in terms of specific types of GUIDE objects. GUIDE Write creates and links these hypertext objects; automatically generating separate files connected by hyperlinks.

Conversion Comparison
For browsing and navigation, all three hypertext conversions were realisable, and to that extent, demonstrated the portability of the data through an accurate and application independent description of the information. Nevertheless, the project raised issues about the interaction between data portability and system functionality and the level of portability that can be achieved. For example, to optimise functionality for each system required modifications whenever a new system was introduced to match the requirement of that particular system. Also the conversion is ultimately dependent on the mechanisms and power of the Mark-It parser and the ingenuity of the user in writing appropriate link routines.
Multimedia
Multimedia information incorporating sound and video exacerbate these differences. Whilst the initial focus of SGML was textual data, other media can be included by one of two mechanisms. The more obvious is through data entities which provide a detailed and formal description of data. However, if the main aim is portability between hypertext systems, this method is cumbersome as it requires an entity and notation declaration must be modified to conform with the format of the target hypertext system. Consequently data portability is more easily achieved by using empty elements to represent different media. In this method the DTD remains constant (although, of course, commands appropriate to each hypertext format must be included in the generator section). Only one element name is declared for each medium; different files are distinguished using Mark-It's `counter' mechanism.
Comparing the use of sound and video across the three hypertext systems focuses attention on the role of their respective browsers. While the Unix Guide interface itself was primarily designed for text and graphics, multimedia applications can be easily launched by creating usage buttons to run script commands and there are no restrictions on the use of different sound and video formats save those arising from the limitations of the multimedia software on the supporting platform. While web browsers such as Mosaic and Netscape were specifically designed to process media other than text and graphics, the browsers must be configured so that they are linked to the multimedia software via the Multipurpose Internet Mail Extensions protocol (MIME) which enables documents in varying formats to be transmitted over the Internet. Consequently the range of sound and video formats supported by web browsers is more restrictive than with Unix Guide although the linking method is very straightforward involving only a simple external hypertext link; the browser then performs all necessary work automatically. PC Guide has similar restrictions: the sound and video formats that are used are determined by the MS Windows platform. Again inclusion is not complicated as the platform via the MCI performed all necessary tasks.

Conclusion
The markup of EPS information was an evolutionary process ie it was not clear at the start what features should be tagged and continual modification of the DTD was required to ensure that the same definition could be used effectively for all three hypertext formats and the question of the optimum level of mark up for portability is still not wholly resolved. An SGML document ensures that there is an accurate description of the logical structure and components, but to achieve maximum functionality from the target hypertext system, modifications were frequently desirable and much vital and system dependent work had to be done during the parsing stage. Data portability is still greatly constrained by the requirements of the target system.