This paper presents a method for merging UML models which takes place in a quality evaluation framework for Web Services (WS). This framework, called iTac-QoS, is an extended UDDI server (a yellow pages system dedicated to WS), using model based testing to assess quality. WS vendors have to create UML model of their product and our framework extracts tests from it. Depending on the results of the test execution, a mark is given to WS. This mark permits to customers to have an idea about the quality of WS they find on our UDDI server. Up today, our framework was limited to WS which did not use other WS. This was justified by the fact that it is impossible for vendors to create a good model of a foreign product. Our method for model merging solves this problem: each vendor produces models of its own product, and we automatically merge the different models. The resulting model from this merging represents the composition of the different WS. For each type of diagram present in the models (class, instance or state-chart diagram), a method is proposed in order to produce a unique model. In addition to this, a solution is proposed to merge all OCL code in the class modeling the WS under test. Unfortunately, this process introduces inconsistencies in the resulting model, that falsify the results of the subsequent test generation phase. We thus propose to detect such inconsistencies in order to distinguish inconsistent and unreachable test targets.

No abstract

International audienceThis paper presents a method for merging UML models which takes place in a quality evaluation framework for Web Services (WS). This framework, called iTac-QoS, is an extended UDDI server (a yellow pages system dedicated to WS), using model based testing to assess quality. WS vendors have to create UML model of their product and our framework extracts tests from it. Depending on the results of the test execution, a mark is given to WS. This mark gives to the cus- tomers an idea about the quality of WS they find on our UDDI server. Up today, our framework was limited to WS which did not use other WS. This was justified by the fact that it is impossible for vendors to cre- ate a good model of a foreign product. Our method for model merging solves this problem: each vendor produces models of its own product, and we automatically merge the different models. The resulting model from this merging represents the composition of the different WS. For each type of diagram present in the models (class, instance or state- chart diagram), a method is proposed in order to produce a unique model. In addition to this, a solution is proposed to merge all OCL code in the class modeling the WS under test. Unfortunately, this pro- cess introduces inconsistencies in the resulting model, that falsify the results of the subsequent test generation phase. We thus propose to detect such inconsistencies in order to distinguish inconsistent and un- reachable test targets