6. Validation Goal and Claims
The main contribution of this thesis is the provisioning of a method for the semantic integration of service-oriented applications. The key feature of the proposed method is that semantically enriched service models are be employed at different levels of abstraction to develop flexible, end-to-end integration solutions from business requirements to software realization.
In Chapter 1, we identified requirements for integration methods in general. To validate whether or not the solution proposed in Chapter 5 meets these requirements, in the following, we make a number of claims and provide arguments for their validity. This is done by applying our integration method in a particular context (i.e., solving two characteristic integration problems presented in Chapter 7 and 8, respectively). When applying our integration method we observed a number of effects. Chapter 9 analyses these observations and argues to what extent our integration method meets the requirements defined in Chapter 1. In particular, our analysis provides more insight into the following issues:
– What effects did we observe when applying our integration method in a particular context?
– Did these observations show that the proposed integration method really meets the requirements defined in Chapter 1? To what extent?
– How similar are the presented cases? Can we generalise the observations made to a broader context? What did we learn when applying our integration method to the cases presented in Chapter 7 and 8 that is relevant for other cases?
In the following, we present the validation claims. Each claim corresponds to a particular requirement.
Claim 1: Service PIMs can be derived from their respective PSMs. This claim is to validate whether our method meets Requirement R1. We verify it in Chapter 7 by providing a concrete model transformation that abstracts a service specification in terms of WSDL to service PIMs in terms of COSMO. In addition, Chapter 8 shows how a service PIM can be obtained if no explicit service PSM is available.
Claim 2: COSMO provides all required concepts to model platform-independent integration solutions. This claim is to validate whether our method meets Requirement R1. We verify it in Chapter 7 and 8 by specifying the PIM of the integration solution using COSMO. In Chapter 7, we focus on the concepts for modeling service aspects whereas in Chapter 8 we also show the usefulness of COMSO perspectives and abstraction levels.
Claim 3: The service models of the systems to be integrated can be semantically enriched. This claim is to validate whether our method meets Requirement R2.
We verify it in Chapter 7 by defining additional service interactions and causality relations in the behavior models of the systems to be integrated. In addition, we demonstrate how an information model can be mapped to a domain-specific ontology, namely Universal Data Element Framework (UDEF (UDEF, 2006)).
Claim 4: The necessary conditions for interoperability can be formally checked. This claim is to validate whether our method meets Requirement R3. We verify it in Chapter 7 and 8 by providing concrete mappings from COSMO to OWL and Petri Nets and verifying the necessary conditions for interoperability (defined in Chapter 5) using logical reasoners.
Claim 5: The same solution PIM can be used to derive different solution PSMs. This claim is to validate whether our method meets Requirement R4. To verify it, we present a hypothetic scenario in which the requirements for the implementation technology of the case, presented in Chapter 8, change. To address the new requirements we specify a new model transformation from PIM to PSM and show how the same service PIM of the integration solution can be reused to derive new service PSM (in terms of the new technology).
Claim 6. The same model transformations can be used to solve different integration problems. This claim is to validate whether our method meets Requirement R5.
To verify the claim, we present a variation of the case, presented in Chapter 7, in which the business requirements change. To address the new business requirements we update the service PIM of the existing integration solution and then reuse the same model transformations to derive the new service PSM.
The table below presents the correspondence between method properties and claims. In addition, we give references to sections, in which the claims are validated.
Requirement Claim Validated in
Claim C1: Service PIMs can be derived from their respective PSMs.
Chapter 7, Section 7.2.1 and 7.3.1, and Chapter 8, Section 8.2.1 Requirement R1: The method
should provide for defining the integration solutions in terms of the problem domain, rather than in terms of
solution technologies. Claim C2: COSMO provides all required concepts to model platform-independent
Claim C3: The service models of the systems to be enable the formal verification of the integration solution.
Claim C4: The necessary conditions for interoperability can be formally checked.
Chapter 7, Section allow for changes in the implementation technology.
Claim C5: The same solution PIM can be used to derive allow for changes of the business requirements.
Claim C6. The same model transformations can be used to solve different integration problems.
Chapter 7, Section 7.3
The remainder of this validation part of the thesis is organised as follows: In Chapter 7 we apply our integration method to solve a reference integration problem, defined in the Semantic Web Service Challenge (SWSC). SWSC provides an infrastructure for testing semantic web service technologies and a forum for discussion based on a common application base. In Chapter 8, we apply our integration method to solve a real-world integration problem
from the travel domain. More precisely, we perform a lab experiment, i.e., we apply our integration method using real-world data in a lab setting.
Finally, Chapter 9 analyses the cases presented in Chapter 7 and 8, and identifies similarities and differences. By doing this, we want to provide further insight into the applicability of our integration method in more general context. In addition, we present a number of challenges we faced when solving the integration problems in Chapter 7 and 8. Finally, we summarise important lessons learnt.
Chapter