Previous work on the user experience in e-commerce has focused on issues as diverse as privacy, trust, interface design, bargaining strategies, and different cultural perceptions (Ackerman et al., 1999; Basso et al., 2001; Callahan & Koenemann, 2000; Liang & Doong, 2000; Spiekermann et al., 2001). While, with the exception of interface design, there is no reason to believe that conclusions from these studies would not be applicable in the m-commerce arena, all of this work has had the underlying assumption that the underlying communication system is capable of delivering an optimum QoS. As has been highlighted in the previous section, this will be difficult in an m-commerce setting due to two main competing factors ‚ firstly, the increasing pressure to add multimedia content to the shopping experience in order to attract and retain customers; secondly, the bandwidth on offer, even at the 3G and 4G levels, might not be sufficient for such applications, bearing in mind the potential increase in the number of users and the number and diversity of devices accessing m-commerce applications.
Whereas server response time will still continue to be a main factor in determining user take-up of m-commerce, this does not preclude enriching the user m-commerce experience with multimedia content. This is primarily because, as has been pointed out in the preceding section, multimedia need not necessarily be sent with full-quality parameters (thus increasing the server burden ) for users can be quite tolerant to low-quality multimedia presentations.
Scenario: During a financial meeting, a company director needs some figures on incoming orders and the business climate in the chemical industry. She voices her query into her mobile phone. Like a meeting companion, the repository interface passes her query to a remote retrieval interface together with further keywords from the recent meeting. This helps the retrieval interface to eventually expand the query in order to achieve better results. The repository interface knows the colleague's actual working environment and presents the required figures and a news report on her handheld device. If she is out of the office, results are sent to her mobile or e-mail account. While viewing the news she prefers to have the figures being transferred to her semitransparent eyeglasses so as to have this important data displayed in front of her eyes. Based on the information obtained, the director decides to buy 100,000 shares of the chemical company and voices her request to her shopping agent.
This scenario is not far- fetched and can be facilitated by a suitable multimedia-enabled B2C m-commerce architecture integrating user requirements, an example of which is presented in Figure 2. Here, user profiles are created offline and store not only user shopping preferences but also multimedia quality preferences (such as "don't mind black-and-white versions of images", start transmitting video at 5fps, and do not transmit video if the available bandwidth is not enough for a 3fps display). Such profiles are stored together with device profiles in an intelligent client interface, which may also, depending on the capabilities (processing power, memory size ) of the client devices, include agents geared to enhance the shopping experience.
The user and device profiles get sent, together with the initial client request to initiate the m-commerce session, to a QoS broker, a virtual server which performs the call admission functions based on knowledge of workloads of servers comprising the server pool. If a session is admitted, the QoS broker passes on the client details both to an application server (server i in Figure 2) in charge of commercial transaction management, as well as to a designated multimedia server (server j in Figure 2). The latter accesses multimedia data-bases which store different quality versions of the same material and dynamically retrieve into their caches the version corresponding to the requested quality. As mobile devices are unlikely to possess significant computational capabilities, any such multimedia material is decompressed on the server side and sent uncompressed to the requesting device. While dedicated multimedia (and application) servers ensure that the processing burden does not fall on one server, the fact that full- sized multimedia data would have to be transmitted to the client implies that, in the initial stages at least, transmission might have to start off with a lower quality and then, depending on resource availability, build on to the requested quality. Moreover, such a scheme of quality adaptation has been shown to be perceptually preferable (Kawalek, 1995).
A QoS monitor on the multimedia server periodically updates the QoS broker with QoS information relating to the transmitted multimedia content. Moreover, each such server adapts the WebGraph framework (Mohapatra & Chen, 2002) to provide scaleable, dynamic presentations of multimedia, with each weblet potentially representing a different quality version of a multimedia clip (a 5fps clip would have a weblet with a refresh rate of 0.2s, while one transmitted at 2fps would have an associated weblet with a refresh rate of 0.5s).
Future m-commerce activities will place considerable processing pressure on servers. While our proposed architecture tries to relieve some of these constraints by having designated application and multimedia servers, other considerations need to be addressed as well. Thus, the variety of devices that might engage in m-commerce activities means that servers must provide suitably tailored content for each of these devices. The XML-FO (eXtensible Mark-up Language Formatting Objects) standard (W3C, 2002), a display formatting language employed to ensure exact and detailed layout control, irrespective of devices and scripts, can be used in such scenarios. Moreover, since current wireless devices can handle presentation aspects when specified with XML-FO, this need not necessarily be done on the server side.
From a different perspective, current network conditions (such as static versus moving, battery powered versus live power) will further increase the pressure for servers to provide scaleable multimedia content but will also furnish the impetus to provide dynamic switching capabilities dependent on contextual characteristics.
Kant and Mohapatra (2000) identify the need for novel ways of interaction between mobile device users and servers in m-commerce scenarios, particularly a voice-recognition-based natural language query mechanism which would simplify transactions, particularly from a user perspective, if one bears in mind the relatively limited interfaces characteristic of wireless devices. In our architecture, such an interaction mechanism can be integrated in the intelligent client interface. Again, the XML family of languages can be used here, particularly in the synthesis of the elaborated query responses, through VoiceXML (Sharma & Kunins, 2002), an XML-based markup language for voice applications, which abstracts the complexity of speech processing from developers.
Last but certainly not least, another important issue which must be taken into account when integrating multimedia perceptual considerations in m-commerce applications is the fact that they are task dependent, and more work needs to be done in order to explore the perceptual impact of multimedia clips presented with differing spatial and temporal parameters, when done so in this specific context. Indeed, if previous results stemming from work exploring the perceptual impact of varying multimedia quality, which have shown that user tolerance increases when tasks are involved (as opposed to task-independent settings; Kawalek, 1995), are an indication of the influence of the task at hand on perceived quality, then chances are that resource requirements needed to maintain an acceptable quality of user experience will not increase.