The above situation is quite typical in additive prime number theory: there are a number of problems in which assuming GRH we can easily obtain a non-trivial bound for the number of integers failing to have a particular kind of an additive representation. Since those problems are usually (more sophisticated) variants of Goldbach's problem, the conventional wisdom in the area is that any unconditional result should generalize the aforementioned result of Montgomery and Vaughan. It turns out that the conventional wisdom is dead wrong! The purpose of this talk is to survey the flurry of recent work (1999--2004) that has led to the realization that in most variants of Goldbach's problem one can use the large sieve to obtain unconditional results of a comparable (and often equal) strength to those that follow from GRH.