Visualization techniques are well developed for many problem domains, but these systems break down for datasets which are very large or multidimensional. Techniques for data which is discrete rather than continuous are also less well studied. Astronomy datasets like the Sloan Digital Sky Survey are very much in this category. We propose the extension of information visualization techniques to these very large record-oriented datasets. Specifically, we describe the possible adaptation of the ``Visage'' information visualization tool to terabyte astronomy datasets.