Undo-Based Access Control for Distributed Collaborative Editors

Abstract

When adding access control layer to a replication based Distributed Collaborative Editor (DCE), ensuring convergence to the same copy of the shared document becomes a challenging problem. We consider here an optimistic access control in the sense that temporarily access right violation is tolerated [3]. This leads to data divergence. To maintain convergence, updates violating access rights must be undone. However, undo approach may itself lead to divergence cases called undo puzzles [6]. In this paper, we address undo as the main feature in an optimistic access-control-based DCE. We also show how we can avoid several known undo puzzles and present additional ones. We propose a new generic solution for these puzzles and provide performance measurements of our undo command.