Collections of references are an important part of scientific and scholarly work. For many people, collections of references are the most advanced form of formal knowledge representation they are using. However, today's standards and tools for collections of references are pretty poor, providing closed environments with no or little extension mechanisms. In this paper, we describe our goal to design and implement an advanced system for collections of references. The primary goal of this system is to provide users with a tool that matches their requirements of semantic richness vs. usability, which are competing goals. As a first step towards this goal, we designed and conducted a survey among the employees of a large university, trying to find out how people are managing their references today, and what their expectations are if a new tool became available. The results of the survey are presented, followed by conclusions that are the guiding principles for the upcoming ShaRef project. The goal of this project is to design and implement a system for reference management that runs Web-based as well as stand-alone, is easy to use, supports collaborative collections of references and collection sharing, has an open and extensible data model, covers the majority of user requirements according to the 80/20 principle, and thus provides scientists and scholars with a better way to manage their collections of references.