Process discovery uses event data generated by information systems to construct process models that describe system behavior. This paper revisits the problem from a conceptual standpoint and clarifies its core elements, including the nature of the input event data, the intended characteristics and purposes of the resulting models, the variants of the discovery problem, the key challenges associated with solving it, and the principles that guide the evaluation of the quality of discovered models. We further position process discovery in relation to related problems in adjacent research areas and identify open issues and research opportunities. Through this analysis, the paper aims to provide a solid reference point for future methodological and tool developments in process discovery.