The old man tallies up his colored beads; He fits a blue one here, a white one there, Makes sure a large one, or a small, precedes, And shapes his Game ring with devoted care. Hermann Hesse, The Glass Bead Game I n 1959, in a talk to the American Physical Society, Richard Feynman talked about amassing the contents of Encyclopedia Britannica on the surface of a pin head. Inspired by the ability of biological systems to encode information on a small scale, the Nobel laureate described his vision

more »

... f a miniature, atom-size computer with the processing ability of the human brain. 1 In 1998, in his capstone address at the IEEE Visualization conference, Turner Whitted envisioned projecting information on large surfaces. Inspired by artistic drawings on cave walls, he advanced the idea of large-format computer displays that surround us and facilitate collaboration. 2 Today we are in the privileged position of witnessing these visionary predictions come true. Although there is still "plenty of room at the bottom," as Feynman once said, and not every man-made surface is a display yet, the recent explosion of computing devices large and small is proof that good progress is under way toward both ends of the scale spectrum. When small, large, and everything in between coexist in the same networked environment, we are faced with the challenge of providing customized access to information, depending on the resources available. While the hardware technology that makes feasible cre-ating small computers or driving large displays has known considerable advances in recent years, software applications have remained for the most part rather generic and insensitive to device diversity. Figure 1a shows a prototype Linux-based wristwatch computer featuring a 320 × 240 dot monochrome VGA display hardly bigger than a postal stamp, a wireless link, and 8 Mbytes of RAM. 3 Figure 1b shows a 3840 × 2400 pixel display connected to a network-attached frame buffer capable of double buffering up to 16 million pixels and a cluster of eight workstations each featuring dual 866-MHz processors, 1 Gbyte of RAM, and high-speed Ethernet connections. 4 Imagine both of these devices attempting to access and visualize stock market data over the network in real time. While the brute-force approach of downloading all the data and rendering it locally might arguably work for the cluster, it is clearly inappropriate for the wristwatch. In fact, given the sheer volume of stock market data and its dynamic nature, both devices could benefit from an intelligent selection mechanism that accounts for the needs of the application running on each device and its resources. This article presents the idea of a unifying framework that allows visual representations of information to be customized and mixed together into new ones. The net result is a fine-grained approach to representing data, better suited to accessing and rendering it over networks. Although my focus is on geometric models and 3D shape representations, many issues I discuss here are relevant to network-based visualization in general.