We are interested in modelling constraint solving as a deduction process. In previous works we have described how a constraint solver can be viewed as a strategy that specifies the order of application of a set of rules, to transform a set of constraints into a solved form. In this framework, the computation process is associated to a constructive proof of a formula. On one hand, the use of transformation rules allows to prove termination in a very easy way. On the other, the use of strategies allows to describe constraint handling in a very abstract way, prototype new heuristics almost by modifying only the choice of rules, and combine constraint solver in the same theoretical framework. In this paper, we introduce COLETTE, an implementation of these ideas using ELAN, a rule-based language. We exemplify step by step how a flexible strategy language allows to prototype existing algorithms for solving CSPs. The theoretical and practical results show that this approach could be useful for better understand constraint solving.