This paper presents a survey of digital musical score databases, focusing on symbolic representation of the music content (as opposed to audio representation) in the context of Music Information Retrieval (MIR). We first provide a primer on Western classical music notation for unacquainted readers. Then, the core of our study categorizes and discusses various approaches to the data management layer of digital score libraries (DSLs), emphasizing their data models and query specifications. Special attention is given to the differences between ASCII-based, semi-structured, and graph-based data models, alongside high-level abstract models. The paper concludes with a discussion comparing these models based on several criteria, aiming to point out their respective strengths and limitations. This work synthesizes existing knowledge in the field of symbolic MIR and identifies promising areas for future research.