What Is Hypermedia?
Journal Article http://www.dsto.defence.gov.au/publications/2817/
SIGLINK Special Interest Group on Hypermedia newsletter, ACM
There is no standard definition of hypermedia (hypertext) in the current literature. Many definitions make reference to the 'nonlinearity' of the access to the information. Strictly speaking, nonlinearity implies an order of presentation which is not governed by a linear total order. Searching for a word in a text document, or querying a relational database are also means of accessing data nonlinearly, yet a word search or a database query are not labelled 'hypermedia'. It seems that hypermedia implies more than just the usual definition of nonlinearity.
The common thread that runs through most definitions is the idea that hypermedia permits the user to customise the presentation of the information in some way. At least some part of the information presentation is in the user's control. In the simplest cases, only the order of traversal is decided by the user. But the more complex and powerful hypermedia systems let the user add to the hypermedia link structures as well. In every case, the user creates a personalised environment tailored to meet their needs.
It is this personalisation and customisability of the information presentation that characterises hypermedia. This incorporates the idea of nonlinearity, since nonlinearity gives control of the order of traversal to the user.
Does this mean that the "best" hypermedia systems are the ones that permit the greatest flexibility of retrieval and presentation to individual users? Since there is a multitude of different requirements, there can be no fixed set of link operational specifications which are sufficient for all needs, However, most hypermedia systems
have real difficulty in supporting the creation of new link operational specifications. At the moment, hypermedia link functionality is mostly restricted to the link operational specifications encoded by the system developers. The users at a site can sometimes modify the scope of existing link specifications through the use of scripts, but there is generally no facility for creating completely new retrieval mechanisms.
Not all aspects of customisability are important to all users. In general, the customisability is most important to the multi-user systems, especially those with very large collections of information. In such circumstances, users need to be screened from the vast mass of irrelevant information. As the collection grows, more refined ways are needed to retrieve the information. Increasing numbers of users implies a need for robust security and filtering mechanisms.
Customisability can be implemented at many levels. The read-only hypertexts only permit the user to control the traversal of links. However, the majority of hypermedia systems permit the user to annotate the information. This is a basic requirement, after all, for most of us have at one time or another, pencilled in comments in the margin of a book (footnote) or paper, or attached those little post-it notes.
Other systems allow the user to create new instances of links, giving the user more control over the retrieval as well as the order of presentation. This is also a basic requirement in hypertext systems, however it is not always supported in "published" documents.
Some improvements in the customisability of hypermedia systems have already occurred. One good example is the 'open system' hypermedia model where, in principle, the user can designate what application is used to display the information they wish to see. This approach overcomes the problem of built-in displayers in other
hypermedia systems which force the user to adjust to the system's interface, not the other way around.
Improvements in Customisability
One of the most important improvements in hypermedia is the dynamic link. The dynamic link has the potential to adjust to the underlying information. This ability to adjust means that the user can link in information from a new source and is no longer restricted to information already in the system's storage. This capability depends on the implementation but the principle does work and is demonstrable.
One approach to dynamic linking is to attach scripts to objects or groups of objects. Script-based systems are becoming more widespread, especially those in an object-oriented environment. The purpose of these scripts can include the modification of the scope of existing link operation specifications. The usefulness of the scripting approach is somewhat compromised by the need to previously create the data structures to which the scripts are attached. In practice, this is not always possible, especially with read-only material or information on remote sources.
Remote and read-only information sources are not the only problem areas for hypermedia. Public-domain or commercially-available information often comes in proprietary formats, so we cannot depend on information being provided to us in the format we require. Importation of such information into our own data structures may need more time and space than is justifiable. One important point to keep in mind is that very often, the material is copyrighted so that duplication via conversion may be legally prohibited.
So we have to assume that some part of the information we want to use must to be accessed in its proprietary form. This means that markup strategies cannot be used, even when the manual effort required is not prohibitively high. It also means that we must be able to access the information without importing it into our own storage area and data structures. Finally, we want to be able to create links from the information as well as into it. But in the current crop of hypermedia systems, these requirements can be very difficult to meet.
From Customisability to Distributed Control
It is the customisability of information that characterises hypermedia and makes it so attractive as an information retrieval and presentation tool. Beyond the usual annotations and browsing control, just how far can we take customisability?
The word 'customisability' implies that the control and ownership of the hypermedia system functionality belongs to the system producers. As we put more and more aspects of the information retrieval and presentation into the hands of the users, any modifications become less like mere customisation and begin to look like full-scale structural changes. After some point, the users, not the producers, become the owners of the hypermedia system functionality, and the changes they can make represent fundamental changes to the retrieval capacity of the system. The ownership and control of the retrieval mechanisms have changed from the corporate to the distributed.
So how can a distributed control system be implemented? We need to specify links so that they can operate over any sort of information. The answer lies in extending the notion of dynamic links. Almost all dynamic link specifications in current hypermedia systems are of the form "Given this starting point, compute what the destination point is". This sort of specification gives us links whose destinations are computed, usually without any need to alter or import the destination information. Now all we have to do is to make the source set of the link dynamically-computed also. This is done by expressing the source set not as a static object or group of objects, but as a rule which expresses exactly which objects belong to the source set of the link. This can be called a fully dynamic link.
Essentially, a fully dynamic link consists of two parts. The destination computation (we call this the resolution) decides what the retrieval operation is, and the source rule determines where we call the retrieval operation from.
Apart from the obvious advantages of gaining access to non-importable information sources, there are a number of other benefits arising from the use of fully dynamic links. If the link specifications are represented in the appropriate way, all of the following features become available to the hypermedia system.
The first is that run-time creation of new link operation specifications becomes possible. A hypermedia system can be purchased with only a very basic set of link operation specifications, and the linking functionality of the systems can subsequently be upgraded by the local users to suit their own needs. This feature also means that is it possible to deal with fundamental changes in the underlying information or in the organisation structure by changing the link operation specifications, instead of by changing the hypermedia system's application code.
Another very useful feature is interoperability, which includes the ability to delegate dynamic link computation to specialised applications. This gives real flexibility to the users who can create new link operation specifications to be sent to their local relational database management system, a full-text retrieval system, or to any specialist application which has a language permitting information manipulation or querying.
The fully dynamic link also permits the user to include their own link usage history and preferences. These user preferences form part of the source rule and can include any aspect of the user environment or link ranking. Users can filter out irrelevant links, also private links can be created that other users do not see.
The nonlinearity of information access is not in itself the key feature of hypermedia. Many recent improvements on hypermedia systems involve either making some aspect of the information retrieval or presentation into a user-configurable element, or permitting the user to create personal workspaces within a larger information system. Many of these developments in hypermedia systems reflect the transfer of control from the corporate bodies to the individual users.
Hypermedia is characterised by the ability to change the retrieval and presentation of information. This characteristic is maximised by the concept of distributed control of information presentation and retrieval.
I wish to thank Eric Youle of DSTO for his valuable comments, and for long discussions on this work. Janet Verbyla has also provided useful comments and is my valued collaborator in this work.
Footnote : Just suppose Pierre Fermat had a hypermedia system instead of paper books. Then the famous Fermat's Last Theorem would not have been the mystery that it has for the last few centuries. Fermat would have had plenty of room to scribble in the "marvellously simple" proof he said he had, and there would have been no excuse for not providing it!