It is a pleasure to be here at the Institute for the second meeting on Image Restoration of HST Images and Spectra. The last time I was in this building was at the first such meeting about three years ago, some months after the discovery of the problems with the HST optics. In the months between those two events, I participated in a panel to advise NASA on the potential of image processing for correcting images for the effects of the spherical aberration. My first reaction on hearing of the problems was that, yes, deconvolution as we do it in radio astronomy could be a substantial help, given enough photons and knowledge of the point spread function. Well, it turned out that the HST deconvolution problem was very tricky in a number of different ways. At the first workshop, I think there was a tendency to minimize the difficulties and to concentrate upon debates as to which was the best algorithm to use. At this workshop, I see less interest in disputes over the merits of Richardson-Lucy versus MEM and more interest in tackling the difficult aspects of the data. In this summary, I want to discuss our deepening appreciation of deconvolution algorithms, then how we did with the tricky stuff, and then conclude by addressing the questions in the title of this talk. Before any of this, I want to record some noteworthy (and telling) comments from attendees.