In this paper, we present the work done for a small company who launched a reengineering process on an existing LMS called SERPOLET in order to allow flexibility in design and interoperability with other LMS. This paper proposes a set of metadata for describing learning resources. For that purpose, we analyse the many data which are processed both by humans and machines about a digital learning resource and to use existing models from the learning sciences field to organise these data from an education perspective as well as from a technical perspective.