Using the multi-layer model for building interactive graphical applications

Most interactive graphical applications that use direct manip- ulation are built with low-level libraries such as Xlib because the graphic and interaction models of higher-level toolkits such as Motif are not extensible. This results in high de- sign, development and maintenance costs and encourages the development of stereotyped applications based on buttons, menus and dialogue boxes instead of direct manipulation of the applications objects. There have been several attempts to provide high level tools for building such applications, including popular toolkits such as Garnet [26], Unidraw [33], Fresco [21, 32] and Open- Inventor [28]. Unfortunately, these tools are not adapted to the development of sophisticated graphical editors because of their lack of extensibility: In this article we argue that these drawbacks come from the fact that high-level toolkits rely on a visualization model to manage interaction. We introduce a model that uses several graphical layers to separate the graphic entities involved in visualization from those involved in feedback and interaction management. We describe the implementation of this Multi- Layer Model and we show how it can take advantage of soft- ware and hardware graphic extensions to provide good per- formance. We also show how it supports multiple input de- vices and simplifies the description of a wide variety of in- teraction styles. Finally, we describe our experience in using this model to implement a set of editors for a professional an- imation system.