Jochen Sauter is the person working on Service Composition at Siemens. Martin Gitsels is in charge of the overall project. At this point, they have identified a set of issues that are compelling.
They seemed to concur with our view that there are going to be multiple service providers with independent services over the wide-area. They have two kinds of scenarios: wide-area service infrastructure, and local-area service infrastructure. They differ in terms of their requirements.
Service description for composition is more complex than service description for discovery. This is because the former involves specification of constraints on composition, workflows, etc. Service interface definition seemed to be a very important piece they are interested in, although they do not yet have concrete scenarios for driving such an interface definition mechanism.
They are interested in mechanisms for optimal path selection and fault tolerance as well.
It was not clear how exactly they were going to proceed, and what issues they were going to concentrate on. One thing they are interested in is building applications involving service composition. Since this is something I am also interested in, this is a possible area of collaboration (see later).
As far as service discovery is concerned, they differentiate between discovery for dynamic composition, and discovery beforehand, for forming service level agreements. For the latter kind of discovery, there is an emerging industry standard: UDDI - Universal Description, Discovery, and Integration. Its a specification for a distributed registry of web-based services. It provides business oriented descriptions. More information at http://www.uddi.org/.
They are interested in a video-conferencing like application involving service composition for Enterprise on Air. (See the notes on Enterprise on Air). The way this could work is: have a video camera at a conference room. The streaming video goes through transcoders in the network before hitting the iPAQ.
With MASH transcoders readily available, this might be something easy to put together. Siemens might be interested in setting up a testbed at Berkeley for testing the acceptance of Enterprise on Air.
They are more interested in a Text-to-Speech and Email service. A typical scenario for wide-area composition might be the case when a visitor from Berkeley goes to Siemens, and composes the Email service at Berkeley with a text-to-speech converter near Munich. Since the pieces for this are also available from ICEBERG, this might also be easy to put together.
A possibility that we talked about was setting up a 2-3 machine cluster at Siemens as a testbed for running these services. This would also give me a good platform for getting some performance numbers over the wide-area.