The wireless Internet scenario exhibits several peculiar characteristics that need to be considered in service provisioning. Mobility of users and access devices is pushed to the extreme. Users can connect to the network from ubiquitous points of attachment and wireless portable devices can roam by maintaining continuous connectivity. Frequent disconnections of users/devices are rather common operating modes that can occur either voluntarily to reduce connection costs and to save battery or accidentally due to the loss of wireless connectivity.

Moreover, the wireless Internet exhibits a high degree of heterogeneity of both access devices (in terms of screen size and resolution, computing power, memory, storage, operating system, and supported software) and networking technologies (IEEE 802.11a/ b/g, Bluetooth, IrDA, GPRS, and UMTS). In addition, this heterogeneity seems not only a temporary aspect due to the novelty and immaturity of the technology, but is expected to last in the open and global wireless Internet.

These distinctive features of mobility and heterogeneity pose new challenging issues and undermine several assumptions of traditional distributed services. Traditional service provisioning relies on a relatively static characterization of the context. For instance, resource availability is typically independent of both the user current location and the access device properties (location and heterogeneity transparency). Changes in the set of accessible resources are relatively small, rare, or predictable.

On the contrary, in the wireless Internet, it is crucial to consider rapidly changing contexts and to frequently reorganize service provisioning in response to context modifications. Client mobility requires solutions that properly and promptly handle changes of client location, modifications in locally accessible resources, temporary disconnection, and changing network topology. In addition, users can change their portable access devices, with different wireless technologies, even at runtime. All the above elements require context-aware service management at provision time.

Service provisioning in the wireless Internet requires the full visibility of location information. For instance, middleware/service components should be aware of the location of both users and involved resources to forward stock trading transaction requests to the server, instances that minimize the current client/server distance. Middleware/service components should also have visibility of different kinds of system level data, such as the access device characteristics and the currently available wireless bandwidth, respectively, to customize service provisioning and to guarantee effective resource usage. These aspects are particularly crucial in wireless provisioning environments because of the scarcity and the high cost of resources. System-level data should be propagated up to the middleware/application level to dynamically determine the applicable context for the user during her session and to perform service configuration and delivery accordingly. For instance, middleware/service components should be aware of the congestion state of both the replicated stock trading service components and the local wireless network. This awareness enables the forwarding of transaction requests to the server instances by balancing the network/service load and, therefore, by minimizing the client connection time.

In summary, the handling of context information in the wireless Internet is complicated by the frequent variations in the provisioning environment, primarily due to client mobility and heterogeneity at provision time. Context variability significantly increases the complexity and the costs of designing, developing and deploying wireless Internet services, thus slowing down their widespread diffusion. As a consequence, contextaware services call for middleware support infrastructures. There is the need for non traditional middleware with full context visibility and capable of automating service reconfiguration depending on dynamic context changes. These middleware should interact with the underlying execution environment to collect relevant information for context determination, for example, current location of users/devices, resource state, user preferences, and device characteristics. This information should be processed at provision time to identify the applicable contexts, their evolution, and the most appropriate service management operations.