Interaction over the Web

A number of factors contribute to delays over the Web − network problems due to high network traffic, slow bandwidth or remote site failures, low processing power of host and client machines and poorly designed Web site interfaces. Although there are many high- speed computers appearing on the market and the network capacity is increasing daily, the problems of delays still remain. This tends to affect users’ interaction in Web applications in general, but even more so in collaborative Web-based applications.

The length of delays experienced on the Web varies and unpredictable delays have a worse effect on interaction than consistent response times (Section 2.2.2). According to a pilot study (Byrne and Picking, 1997), Web users are critical of delays and regard time-related factors to be of importance to Web usability. The Web is therefore an ideal environment to

study the effect of response time and the role that time plays on the interactive process. Note that, issues directly related to Web site design will not be considered here as the discussion centres on the Web infrastructure.

2.3.1 Problem areas

Traditional interface design tends to hide the underlying system architecture and computations from the users (Section 2.2.3). Web interface design seems to adopt the same policy, thus reducing its usability. For example, when a user visits a Web site, it is impossible to tell how long the browser will take to load the whole page. This obviously depends on the size of the page, any associated graphics or computations that must be carried out on the local machine and the network speed. However, users are not usually provided with such a level of detailed information, which will allow them to make informed decisions.

2.3.1.1 Response time

Like traditional human factors research into response times (Section 2.2.1), research on Web systems has also shown that users need response times of less than 1.0 second when moving from one page to another if they are to navigate freely through an information space (Nielsen, 1995). Web users are not currently getting sub-second response times; hence they get frustrated (Nielsen, 1997). A response time of no more than 10 seconds has been recommended as the limit for keeping users attention focussed while waiting for Web pages to download.

Often, the Web browser affords users an understanding of the progress in the computer activity whilst they are waiting, through a progress indicator. Although this enable users to tolerate delays, browsers should provide useful progress bars that communicate what percentage of the entire download for a page has been completed (Nielsen, 1997).

2.3.1.2 Network latency

On the Web, the latency of the network is the most obvious cause for delays. A faster network connection only tend to increase the Web performance by a small factor. For instance, upgrading from a dialup modem to an ISDN line only doubles the performance (Nielsen, 1997). Even if a high bandwidth connection to the Internet is used at both user- end and server-end, the response time will increase during network bottlenecks, especially for cross-continent connections and for use at peak hours. In extreme cases, the communication may be broken for longer periods if not completely, for instance during mobile work (Dix, 1995a).

2.3.1.3 Collaborative interaction

The nature of collaborative work itself introduces further delays. Group users do not only have to wait for feedback of their own actions but they must also wait for the effects of others actions − feedthrough. User feedback and feedthrough may improve by using a

faster network connection, but if the cooperative task requires many short network interchanges, additional delays will occur due to buffering and processing at remote and local sites.

Irrespective of the source, delays disrupt user interaction in general, and in particular they affect the nature of the work process during collaborative activities over the Web.

2.3.2 Coping Strategies

Users also adopt coping strategies on the Web to deal with delays when they are seeking for information, but the strategies tend to vary depending on the users knowledge.

A study (McManus, 1997) showed that when users were aware of the location of the Web page they were trying to reach and they had some knowledge of the hardware and/or browser being used, they tried to reach the desired information as efficiently as possible by performing some of the following actions:

• multi-threading

• download pages to the local machines for browsing at a later stage

• expand the cache to allow quicker access to pages viewed earlier in the session

• deactivate automatic image loading

However, when users had little knowledge of the information they were looking for, they minimised the time spent locating that information by carrying out some of the following actions:

• use a site or author they trust and follow their links

• avoid sites that contain a large number of graphics or frames

• use search engines

• use personal information feedback or agent

• use FTP

Another interesting observation from the study (McManus, 1997) was the fact that users actions varied depending on the granularity of the interaction. For instance, when users’ interaction were over a long time scale they would adopt measures such as, download pages to the local machine, expand the cache or use personal agents. But in short time scale interaction, users would rather deactivate automatic image loading or avoid sites that contained lots of graphics and frames.

The underlying Web infrastructure does not assist users when they adopt coping strategies. Users could be provided with some help to overcome the problems of delays in some cases. However, it is not always desirable to support all the strategies that users undertake, as this adds an extra burden on the task at hand and may interfere with the different tasks that the users are performing (Section 2.2.5).

2.3.3 Potential solutions

A number of suggestions have been made in the research literature to improve usability on the Web despite its intrinsic delays. Because it is not always possible to control the occurrence of delays, the impact of delays can be reduced by providing users with a greater control over temporal issues, such as Web page loading times. For example, the browser may open some form of dialogue that queries the course of action the user wishes to take, thus creating a sense of rhythm during the interaction (Kutar, 2001).

Other research have suggested that users should be able to decide beforehand whether the value of the information they are trying to download outweighs the cost in retrieving it. For example, small chunks of ‘meta-data’ about a link can be downloaded and the size and type of information at the link location can be provided to the users via a pop-up right button menu (Bentley, 1997). Similarly, the browser could render images as thumbnails to allow users to evaluate the cost and benefit of viewing those images.

A slow but consistent interface with a regular response time may be preferable to an inconsistent interface with occasional fast responses (Section 2.2.2). However, the notion of slowing down our ever fast computing power is seen as an outrage in a world where speed is increasingly more important (Gleick, 2000).