New algorithms and applications of cyclic reference counting

Abstract

Reference counting is a simple and efficient way of performing graph transformation and management in which each graph node stores the number of pointers to it. Graph operations are performed in such a way to keep this property invariant. The major drawback of standard reference counting is its inability to work with cyclic structures, which appear ever so often in real applications. The author of this talk developed a series of cyclic reference counting algorithms whose applicability goes far beyond the implementation of garbage collectors in programming languages. This paper presents the milestones in the history of cyclic reference counting followed by two new applications: the consistent management of Web pages in the Internet and the correctly handling of processes in clusters and grids. © Springer-Verlag Berlin Heidelberg 2006.