Engineering the PageSpace

The Java language is rapidly becoming a de facto standard for scripting client-side applications on the World Wide Web. Java is a sort of simplified object oriented language easily portable across several architectures. Many commercial and research browsers have announced Java support.

Java, although powerful, lacks of a general coordination model. Thus, the coordination of active agents within the PageSpace shall be expressed and implemented using a certain type of parallel programming languages, called coordination languages. Our basic idea is to augment Java with the coordination model offered by some coordination language. We are exploring two possible candidates for this task:

• Shade, an object-based coordination language developed in Bologna [5], which includes the concepts of a shared object dataspace and objects-as-multisets. It has been designed to offer the abstraction of a shared object space that is both a repository for objects and a coordination media of services and of their clients. Inheritance is implemented by delegation. Coordination is driven by rules and is based on associative communication between an object and the object space.

• Laura, a coordination language developed in Berlin [11] which extends the Linda's Tuple Space with the concept of services that can be associatively requested. Laura has been designed to meet the requirements of open distributed systems, such as heterogeneity of hardware, networks, and software as well as to potential high dynamics of the system caused by joining and leaving components. It provides three coordination primitives that manipulate forms in a shared data space, called the service-space. These forms contain service-offers, -requests and -results. Services are identified by the types of their interfaces, matched according to a subtype relation. Laura is implemented on top of the ISIS toolkit [3] and has been embedded into C and csh.

The PageSpace concept should be a conservative extension of the mechanisms used in World Wide Web and retain their elegant and efficient simplicity; only minor extensions to the coordination languages chosen for the experiment are foreseen, and implementation will be based on existing standard components, such as communication toolkits, information brokers, public-domain World Wide Web tools, etc.