In this work, we present our experience when developing the Matching Framework (MaF), a framework for matching ontologies that allows users to configure their own ontology matching algorithms and it allows developers to perform research on new complex algorithms. MaF provides numerical results instead of logic results provided by other kinds of algorithms. The framework can be configured by selecting the simple algorithms which will be used from a set of 136 basic algorithms, indicating exactly how many and how these algorithms will be composed and selecting the thresholds for retrieving the most promising mappings. Output results are provided in a standard format so that they can be used in many existing tools (evaluators, mediators, viewers, and so on) which follow this standard. The main goal of our work is not to better the existing solutions for ontology matching, but to help research new ways of combining algorithms in order to meet specific needs. In fact, the system can test more than 6 * 136! possible combinations of algorithms, but the graphical interface is designed to simplify the matching process.