The emerging technology of large (approximately 10,000 pixel) submillimeter-wave bolometer arrays presents a novel optical design problem -- how can such arrays be fed by diffraction- limited telescope optics where the primary mirror is less than 100,000 wavelengths in diameter? Standard Cassegrain designs for radiotelescope optics exhibit focal surface curvature so large that detectors cannot be placed more than 25 beam diameters from the central ray. The problem is worse for Ritchey-Chretien designs, because these minimize coma while increasing field curvature. Classical aberrations, including coma, are usually dominated by diffraction in submillimeter- wave single dish telescopes. The telescope designer must consider (1) diffraction, (2) aberration, (3) curvature of field, (4) cross-polarization, (5) internal reflections, (6) the effect of blockages, (7) means of beam chopping on- and off-source, (8) gravitational and thermal deformations of the primary mirror, (9) the physical mounting of large detector packages, and (10) the effect of gravity and (11) vibration on those detectors. Simultaneous optimization of these considerations in the case of large detector arrays leads to telescopes that differ considerably from standard radiotelescope designs. Offset optics provide flexibility for mounting detectors, while eliminating blockage and internal reflections. Aberrations and cross-polarization can be the same as on-axis designs having the same diameter and focal length. Trade-offs include the complication of primary mirror homology and an increase in overall cost. A dramatic increase in usable field of view can be achieved using shaped optics. Solutions having one to six mirrors will be discussed, including possible six-mirror design for the proposed South Pole 10 m telescope.