Until now, information has been interchanged by rendering it on a visual display, e.g., a sheet of paper. This requires that one has to be able to ``see'' the visual layout in order to infer and understand the underlying structure present in the information.
The advent of electronic information presents an opportunity to improve on this situation. The computer, as an intermediary in the process of information interchange, can take over the task of displaying the information according to the individual user's needs and abilities. But in order to achieve this, the electronic encoding needs to capture high-level structural and semantic information. Here, we outline some of these needs, and point out pitfalls to avoid.
The advent of screen-access programs, -computer programs that allow a visually impaired person to ``hear'' the contents of a visual display, screen enlargers and Braille translation programs- has provided a degree of access. As a consequence, online availability of some of the latest information has made these immediately accessible. However, there are certain pitfalls to be avoided when designing document encodings in order to ensure universal accessibility. The encodings used by the information servers to disseminate information need to capture information in a display-independent manner.
Information that is stored in a display-independent manner, i.e., without assuming any particular mode of display, is universally accessible. Such universally accessible information can be viewed using a visual client, displayed using a larger font by a person having limited vision, translated to Braille, or listened to using an audio-enabled client.
On the other hand, storage formats such as Postscript or the Portable Document Format (PDF) -PDF is a document encoding format used by the Adobe Acrobat to enable document exchange between disparate computing platforms- assume only a visual mode of presentation and are inaccessible. This is because page-description languages like the above merely capture the visual appearance of a document and it requires a human looking at the physical rendering to reconstruct the underlying information structure. Thus, though excellently suited to produce good visual reproductions of a -a purpose for which they were explicitly designed- they are also unsuited for producing non-visual renderings of the information.
To make this more concrete, here is a specific example. A server that maintains a large collection of documents can provide the information either as graphical images of the printed page, or as high-level markup documents, e.g., marked up using HTMLTeX. If the former approach is used, the information will only be accessible through a visual client. On the other hand, a server that provides the information in a display-independent form is more flexible, since the choice of how the information is displayed is left to the client. Such display-independent information servers are crucial for providing universal accessibility.
Markup languages like the Standardized General Markup Language (SGML) and the Hypertext Markup Language (HTML) capture the structure present in a document, leaving it to a client program to produce an appropriate display. It is this separation of information capture and display that make these markup languages the best choice.