Current-generation constraint programming languages are considered by many, especially in industry, to be too low-level, difficult, and large. We argue that solver-independent, high-level relational constraint modelling leads to a simpler and smaller language, to more concise, intuitive, and analysable models, as well as to more efficient and effective model formulation, maintenance, reformulation, and verification, and all this without sacrificing the possibility of efficient solving, so that even lazy or less competent modellers can be well assisted. Towards this, we propose the ESRA relational constraint modelling language, showcase its elegance on some well-known problems, and outline a compilation philosophy for such languages.