Composition is an important mean for creating new services based on already existing ones. Therefore services are aggregated to form a new workflow, whose input and output parameters match those of the service requested by the user. In between those services, which compose the new workflow, there have to be matches between their parameters, i.e. preceeding services have to create those output parameters that are required by the following services. Not only is it important to find matching in- and output parametertypes but also such parameters that fulfill the purpose the user needs. Therefore, besides a technically matching, semantics have to be regarded considering the parameters and their use in the services. Matching between parameter-notations that use different vocabularies but share a semantic meaning should be found and matched, so that the specific services can be composed nonetheless. The aim is to provide methods that ensure that a composition is syntactically as well as semantically correct.
While this has to be an automatic proceeding, due to the huge amount of possible workflow compositions, there may be manual parts for the human user, which ensure transparency and control. Non-functional properties are maybe not covered satisfactory, even if a sufficient service description language is provided. This may happen because the user himself has problems in articulating or valuating his non-functional preferences. A manual revision of the composition should ensure that all his needs and specific wishes are regarded. Therefore we strive for a semi-automatic composition.
Furthermore, a dynamic reconfiguration at runtime should be possible if required. This can be the case if some services that are used in the composition fail or are overloaded or because some external or internal parameters changed. The system should recognize such cases and react in that way, that it autonomously finds solutions and provides them, e.g. by replacing a part of the composition.
We regard uniform data mediation as an important basis for our composition methods. Thus, the service description language has to ensure these among other things. The compositions trustworthiness should be compatible for SOC, so that its unlimited usage can be guaranteed.