We describe DPOSS, a new digital survey of the northern sky, based on the POSS-II photographic sky atlas. The survey covers the entire sky north of δ = −3 • in 3 bands, calibrated to the Gunn gri system, reaching to equivalent limiting magnitude of B lim ∼ 22 m. As a result of the state-of-the-art digitisation of the plates, detailed processing of the scans, and a very extensive CCD calibration program, the data quality exceedes that of the previous photographically-based efforts. The end product of the survey will be the Palomar-Norris Sky Catalog, anticipated to contain > 50 million galaxies and > 2 billion stars, down to the survey classification limit, ∼ 1 m above the flux detection limit. Numerous scientific projects utilising these data have been started, and we describe briefly some of them; they illustrate the scientific potential of the data, and serve as the scientific verification tests of the survey. Finally, we discuss some general issues posed by the advent of multi-terabyte data sets in astronomy. 1 Introduction The Palomar Digital Sky Survey (DPOSS) represents a digital version of the POSS-II photographic sky atlas [19]. It is based on the scans of the original plates, done at the Space Telescope Science Institute [14], [16]. The final result of this effort will be a catalog of all objects detected down to the survey limit, the Palomar-Norris Sky Catalog (PNSC). For more details, see, e.g., [5]. The goal of this project is to provide a modern, uniform digital data set covering the entire northern sky in 3 survey bands (photographic JF N, calibrated to Gunn gri), with a photometric and object-classification accuracy of sufficient quality to enable a wide range of scientific follow-up studies. The survey processing and calibration were designed to extract all of the information present in the plates. Our tests indicate that the resulting DPOSS and PNSC data are superior to other, photographically based sky surveys in the past in terms of the photometric quality, uniformity, depth, and object classification accuracy. We believe that this improvement is due to a combination of factors: (a) a superior scanning process, which minimizes scattered light problems, while maintaining the full angular resolution and dynamical range present in the plate data;