One of the earliest discoveries in hypertext research was the importance of swift link traversal. The rule of thumb, well established by the time of the first Hypertext Conference in 1987 and since known as "Akscyn's law" in honor of Rob Akscyn (the hypertext pioneer who was one of its first discoverers) states that:
Hypertext systems should take about 1/4 second to move from one place to another. If the delay is longer, people may be distracted; if the delay is much longer, people will stop using the system. If the delay is much shorter, people may not realize that the display has changed.
This rule was established in the course of the ZOG program which installed a pioneering hypertext help system aboard an aircraft carrier, back in 1983 when hypertexts (and personal computers) were a laboratory curiousity.
Robert Akscyn, Donald McCracken, Elise Yoder, KMS: A Distributed Hypermedia Systems for Managing Knowledge in Organizations, Hypertext '87 Proceedings, ACM, New York, 1988
Six years later, Jakob Nielsen's survey of "The Matters that Really Matter for Hypertext Usability" (Proceedings Hypertext '89, ACM, New York, 1989 pp. 239-248) identified a response-time study by Patterson and Egido as one of the key usability results then known. In short, we've known for more than a decade that predictably swift link traversal is one of the most important factors in making hypertext systems effective, pleasant, and usable.
Links on the web are often quite slow -- particularly for the vast majority of users who connect by modem [This was written in the late 90's -- ed]. Network traffic delays and server load can also make Web links slower and less reliable; these delays are unpredictable and can affect any network connection. Delays are magnified, moreover, by elaborate formatting and by graphic richness; sophisticated browsers, knowing that users will experience significant delays in downloading a page, often impose additional delays to set up and format the page. Far from moving toward conformity with Akscyn's law, current Web page and browser design violate it more severely each week.
What can we do?
One solution is to abandon hypertext links. Since web links are slow, they will be used only where unavoidable -- for movement to and from our site. Within our domain, we may avoid links by presenting everything in one scrolling view. This strategy may be useful for simple messages, but is unwieldy for communicating complex ideas to a diver audience. (It also eliminates the possibility of meaningful transitions, and reduces the Web to a feeble simulation of paper)
Another strategy is to exchange one large delay for many small ones. We can sometimes arrange a Web site to to bundle large parts of the data in a package that is pre-loaded at the entrance. Within the site, link response is quick because time-consuming sound, graphic, video, and applets have been pre-loaded onto the user's computer. Some readers, however, will not wait through the initial lengthy delay; they will leave the site before it can offer them a reason to stay. The pre-loading strategy also demands either a coercive hypertext structure, where the designer knows in advance what the reader will see -- or it will waste the reader's time in downloading information that may never be used.
Despite reckless violation of Akscyns law the Web is hugely popular. Does this mean that no longer in force? Or people will tolerate unpleasant delays if reward sufficiently compelling?
Finally, we may try to mask slow link speed by providing some information immediately. Interlaced GIF files (and low-resolution graphic variants) are familiar examples. Even more important, though, is the simple expedient of placing some compact, fast-loading information near the top of a page. This information will usually be displayed quickly, giving the user something to read while her computer fetches the rest of the page. (HTML tags that help browsers with layout, such as HEIGHT and WIDTH information for images and tables, make this trick effective.)
A better solution, for sophisticated hypertexts that must provide crisp performance, may be to use the Web as a way to provide access to, and information about, hypertexts that can be downloaded (or purchased) as a unit and then performed on the user's machine. Unlike the pre-load strategy, which assumes that we're simply fetching some ephemeral data, we may view a hypertext as an artifact, something that we'll want to store and use. Automatic link strategies (discussed in an earlier HypertextNow) can easily keep the downloaded version connected to a Web repository, permitting seemless additions and updates. With the hypertext stored locally, unpredictable network delays cease to trouble us. With data readily available on our local computer, we can easily obey Akscyn's law.