It is now almost 200 years since Gauss, a teenager at the time, formul
ated his famous principle of least-squares and used it to determine, f
or the first time, the orbit of one of the asteroids, a problem which
had defeated astronomers for years. When applied to the crystallograph
ic phase problem, least-squares leads directly to the formulation of t
he minimal principle, which effectively replaces the phase problem by
one of constrained global minimization. Shake-and-Bake, the computer s
oftware package which implements this formulation of the phase problem
, provides a completely automatic solution of this problem. The progra
m has solved ca two dozen structures, of which five or six had been pr
eviously unknown, with no failure. Three of these structures - gramici
din A, crambin and rubredoxin - in the 300-500 atom range, and six or
seven others, in the 100-200 atom range, were all routinely solved Wit
h this background it now appears likely that, provided data to atomic
resolution is available, structures having as many as 1000 atoms or mo
re will prove to be solvable by this technique. What if only diffracti
on data to less than atomic resolution is available? Here the crystal
ball becomes murky and the evidence less than compelling. Nevertheless
, in view of recent experience, it is again conjectured that, building
on existing techniques, complex structures will eventually prove to b
e routinely solvable, even with data to only 1.5 Angstrom resolution.