This talk addresses the issues of service coordination in complex business process scenarios. While existing approaches for service coordination, e.g., BPEL are powerful standards for composition of services, they lack support for a number of features highly desirable for service coordination, e.g., they hard-wire services too much, or force protocol decisions that are not required. In contrast, the declarative coordination language Reo concentrates only on important protocol decisions and defines only those restrictions that form the domain knowledge, leaving more freedom for process specification and choice of individual services compared to traditional approaches. Reo considers coordination as a first-class concept, isolating the coordination aspects away from individual services, maintaining a loosely coupled environment with services knowing little or nothing about each other. In this talk we rise a number of issues that must be solved before Reo is seriously considered as an industry-level coordination and composition language. As a part of the talk, we show how Reo can be used to support Enterprise mash-ups in the style of Yahoo Pipes.

We describe the implementation of the filter channels and how to extend them to handle arbitrary data transformations. We also discuss issues related to server-side deployment of Reo engines and an HTTP-based lightweight protocol for remote components.