The structure of a music piece is a concept which is often referred to in various areas of music sciences and technologies, but for which there is no commonly agreed definition. This raises a methodological issue in MIR, when designing and evaluating automatic structure inference algorithms. It also strongly limits the possibility to produce consistent large-scale annotation datasets in a cooperative manner. This article proposes an approach called decomposition into autonomous and comparable blocks, based on principles inspired from structuralism and generativism. It specifies a methodology for producing music structure annotation by human listeners based on simple criteria and resorting solely to the listening experience of the annotator. We show on a development set that the proposed approach can provide a reasonable level of concordance across annotators and we introduce a set of annotations on the RWC database, intended to be released to the MIR community.