Link Properties

By Thomas Trickel


Ever since Vanevar Bush introduced the concept [1] and Ted Hoff coined the term, Hypertext [2] and now Hypermedia have held out the promise of making information more accessible and understandable through their ability to link related ideas and concepts. Traditional paper based information is limited by its medium. Traditional text is forced by its medium into a mostly linear form. References are limited to devices such as footnotes, sidebars or parenthetical references which do not provide the speed or convenience of access, that we in this computer age are becoming used to. Hypertext on the other hand allows text to be created and viewed in a non-linear manner. Paragraphs can be linked together in many different orders allowing the information to be presented in other ways. References and annotations can be stored in separate documents, but they can still be connected to the referring document. It is this ability to connect, or link, portions of text that are related but not necessarily contiguous that sets Hypertext apart from traditional text. The promise of Hypertext is then the promise of links.

Initially I thought that links were merely a connection from one node to another. But my early reading [3] introduced the idea of link types and the possibility of an unlimited number of types. This raised the idea that links could be more complex, that a link could be more than just a connector. A link could contain information about itself and the nodes it was connecting. Information contained by a link could be what the link does, what the link is pointing to and how the link relates the two (or more) nodes. An example of this would be a link that when activated caused a pop-up window to display a definition of a word. Three pieces of information that this particular link carries are: it causes a pop-up window to appear, it points to a definition and it has two nodes that are related by a definition. Other links could cause another document to be displayed or a sound to be played. Links could point to other paragraphs within the existing document, to a different document on the same machine or even to a document on a different machine in a different country. Among the relationships that a link could define are: reference, expanded explanation or argues in support of.

This theoretical model proposes three properties of Hypertext Links:

  1. The action the link causes in the user interface.
  2. The resource the link points to.
  3. The way two, or more, nodes are related.
Thus, it seems that links could have the capability of containing more information than my original idea that a link contained merely a pointer to another node or nodes. How does this theoretical model compare with two popular Hypertext implementations?

Links in the World Wide Web

One of the most popular forms of Hypertext today is the World Wide Web(WWW). The WWW uses plain (ASCII) text documents that are annotated with a markup language called HTML (HyperText Markup Language) [4] to convey its information. In the WWW implementation, links are created by using a markup tag called an anchor which takes the form:

<a href ="URL">text used to identify the anchor</a>.

URL [5] stands for Uniform Resource Locators and is the method that the WWW uses to link information. The information contained in the URL defines what type of node a link points to and where the node is located.

The action a link initiates in the user interface is client dependant. Selecting a link causes most web browsers to, either display the resource referenced by the link, or, to start another application in a separate window to display resource types that browser cannot interpret.

Unfortunately, the WWW has no way of representing how the nodes are related, other than what is contained in the text used to identify the link. Link information in the WWW is limited to the current node's URL and the text of the link.

In the WWW the capabilities of the proposed three link properties are:

  1. The links can cause one of two actions in the user interface. Either display another node or start another application to handle the information pointed to by the link.
  2. What the link points to is defined by the URL contained in the anchor, or link.
  3. There is no way, other than what is contained in the text used to identify the link, of representing how the nodes are related.

The Internet Gopher

Calling the Internet Gopher a Hypertext application might be stretching the definition a bit but I believe a look at the implementation of the Internet Gopher's links will prove instructive. Links in the Internet Gopher are either generated on the fly by the Gopher server or stored in a .links file. I will concentrate on the links stored in the .links file because I believe they are the most appropriate for this discussion. A .links [6] file is made up of multiple enteries of several lines of text with keywords to identify the lines. A typical entry looks like:

Name=National Weather Service Forecasts
The items in this list are the information contained in a typical Gopher link. In addition to the information that is contained in a URL, the Internet Gopher adds Name and Numb. The Gopher+ .links definition [7] adds a few more types of information. Of the types added, Abstract is potentially the most useful, allowing the person interested in the link to see more information about the link before it is taken.

The action a link initiates in the user interface is client dependant. However, Internet Gopher clients cause a browser's user interface to display another menu screen, display a file or start another application.

With the addition of the Gopher+ definition a new .links file entry can be defined to contain the relationsip of the two nodes.

Links in the Gopher implementation:

  1. The links either display another menu screen, display a text file or start another application to handle the information pointed to by the link.
  2. What the link points to is defined in .links file entries.
  3. With the addition of the Gopher+ definitions it is possible for the link information to contain the relationship of the two nodes.

The Missing Property

Both the WWW and the Internet Gopher differ from the proposed theoretical model by omitting the relationship of the nodes connected by the link. The WWW and the Internet Gopher implementations could be extended to have a link property that contained the relationship information. In both cases new browsers would have to be created to use the new link property.

The Internet Gopher protocol allows each link to be sent from the server to the browser as a tab delimited text field with the links seperated by a carriage return. Thus, additional fields to be interpreted by a compatible browser could be easily added. This is how the the newer Gopher+ protocol extends the older Gopher protocol. The Gopher+ protocol doesn't explicitly address relationship but it could be added easily.

With the WWW's relience on HTML to add links to a node the HTML code would have to be extended. The browsers would have to be modified to recognize the new extensions. Netscape does a similiar type of extension to keep track of the date that a particular link was visited. A Netscape bookmark file link looks like:

<A HREF="" ADD_DATE="787009776" LAST_VISIT="787816408">Welcome to Netscape! (home of this utility)</A>.

The traditional link of <a href="URL">link text</a> has been extended with the addition of ADD_DATE and LAST_VISIT. LINK_RELATIONSHIP could be just as easily added which existing browsers would just ignore like they ignore the Netscape defined extensions.

The Future of Links

As we rely more and more upon computers to convey our information the information that is being conveyed needs to become smarter. Links in the current implementation of the WWW aren't very smart. Smarter links, links that contain more information than the node type and location being linked would be quite useful to users of the system. An example of this is the classification of links. Currently classifying links requires a person to be involved. However, if links contained their own classification information such as this link points to a document on bicycles, travel and Costa Rica, a computer could easily classify or search for them. In addition to aiding the classification of links smarter links could contain other information like abstracts and sizes of the linked to files that would aid a user in his selection of the link. .

Comments and Criticism

Comments and Criticism directed to:
will be added to this section for future reference.

About the author.


[1] Bush, Vannevar As We May Think Atlantic Monthly, July 1945, pp. 101- 108 or html version

[2] A WWW definition of the word hypertext

[3] Conklin, Jeff Hypertext: An Introduction and Survey IEEE Computer, Sept. 1987

[4] A Beginner's Guide to HTML

[5] A Beginner's Guide to URLs

[6] gopherd on-line man pages

[7] Gopher+ upward compatible enhancements to the Internet Gopher protocol