Thin-Client Programming Model

  The key aspects of the Metis thin-client programming model are that the client downloads AFEs from the network; that these AFEs rely only on services found on network servers; that the services are bound as late as possible; and that AFEs interact with each other within the confines of a workspace. AFEs are securely installed and downloaded using a code service. They are also made `thin' by implementing most of the application logic as one or more services. Late binding to these services provides:

• manageability, because services can be moved across server machines without impacting AFEs;
• flexibility, because services can be selected based on server load; and
• fault-tolerance, because a service can be obtained from an alternate server.

The workspace is a container for AFEs, allowing for interaction, as well as providing a shared environment. One important part of that environment is the authorization information that can be read from a smart card or provided as part of a logon process from an authentication service. The authorization information is used first to determine if a user is allowed to use the system, and then to identify the user's access rights to available AFEs. Afterwards, this information can be used directly by the AFEs to authenticate themselves to the service providers.

Figure 1 outlines the various building blocks of the thin-client programming model. It shows three important parts - the client workspace, application-specific services, and support services needed to provide full thin-client functionality.

 

 

Figure 1: The schematic shows the three important parts of the thin-client application model: the client workspace, support services, and application-specific services. The client workspace contains user profile and authentication objects, objects to find and bind to services, and controllers for data and print. AFEs execute in the context of the client workspace.

• Client Workspace
• Application-Specific Services
• Metis Support Services