Navigation and Coordination Primitives for Multidimensional Visual Browsers

Allison Woodruff, Alan Su, Michael Stonebraker, Caroline Paxson, Jolly Chen, Alexander Aiken, Peter Wisnovsky, Cimarron TaylorVDB 1995

Introduction

In the mid-1990s, database interfaces were primarily restricted to textual queries and rigid forms, presenting a significant barrier for non-expert users—particularly scientists managing massive multidimensional datasets. While the "Tioga" project introduced a visual programming model using "boxes and arrows" to visualize data, it lacked sophisticated navigation capabilities. This paper addresses the need for intuitive exploration of complex data spaces, focusing on three core primitives: semantic zooming, wormholes, and browser coordination.

Core Navigation Primitives

Semantic Zooming via Elevation Maps

Traditional graphical magnification simply enlarges pixels. The authors propose "semantic zooming," where the level of detail changes based on a logical dimension called "elevation." This is managed through elevation maps, which dictate which data "recipes" (queries) are active at specific heights.

Elevation maps define data visibility based on logical height.

As shown in Figure 3, developers can define "required," "optional," or "exclusive" data layers. For instance, at a high elevation, a user might see a state outline; as they "descend," the system automatically transitions to showing census tracts, then river networks, and finally individual structures. This ensures that the user is presented with contextually appropriate information at every scale.

A wormhole acts as a connector between disparate multidimensional spaces, generalizing the concept of a hyperlink. Unlike traditional 2D hyperlinks that connect static objects, a wormhole allows a user to "tunnel" from an object in one coordinate system (e.g., a specific house on a map) into an entirely different browser environment (e.g., a document viewer showing the residents' records). This enables seamless transitions between related but structurally different datasets.

Browser Coordination: Slaving and Cloning

To maintain orientation during deep exploration, the system allows browsers to be "slaved" to one another. By utilizing translation functions such as LOC_CALC and ELEV_CALC, a master browser can control the view of a slave browser.

Coordination between inner and outer browsers within a multidimensional data space.

This mechanism, illustrated in Figure 5, enables sophisticated UI constructs like "magnifying glasses" or "Magic Lenses." An inner window can show a high-detail or alternative data layer (like population density) while remaining spatially synchronized with an outer window providing broader geographic context.

Implementation and Applications

The system is powered by the POSTGRES DBMS, which optimizes and executes the underlying visual programs (recipes). The authors demonstrated these primitives using several practical datasets:

  • California Geographic Data: Managing state outlines, census tracts, and highway systems.
  • Digital Libraries: Navigating a 35mm slide library and multi-resolution Kodak PhotoCD images (ranging from 128x192 thumbnails to 2Kx3K full-resolution images).
  • Scientific Visualization: Supporting Earth science data as part of the SEQUOIA 2000 project.

Key Insights and Conclusion

The primary contribution of this work is the shift from purely visual zooming to logical, semantic navigation. By treating zooming as a change in data representation rather than just scale, and by providing a unified execution model for coordinated browsers, the authors created a flexible framework that unites Geographic Information Systems (GIS), hypermedia, and scientific visualization.

While the paper presents a robust architectural model, it notes that full integration into the Tioga system was ongoing at the time of publication. Future work remains necessary to evaluate the computational overhead of concurrent coordinated recipes and to conduct user studies verifying that these primitives effectively reduce disorientation in large-scale multidimensional environments.

Summary updated

BibTeX
@inproceedings{woodruff1995navigation,
  author = {Allison Woodruff and Alan Su and Michael Stonebraker and Caroline Paxson and Jolly Chen and Alexander Aiken and Peter Wisnovsky and Cimarron Taylor},
  title = {{Navigation and Coordination Primitives for Multidimensional Visual Browsers}},
  booktitle = {VDB},
  year = {1995},
}
Figure