Building software with off-the-shelf (OTS) components is an attractive alternative to bespoke development in terms of initial development cost. Dependability assurance of such software, however, appears more difficult than that of the bespoke alternative since it is rarely possible to use evidence about the development process of OTS components. Software development with OTS components is centered upon selecting the best components in the particular development context. If dependability itself or evidence about high dependability is insufficient, often the only available option for improvement is deploying fault tolerance based on design diversity. In such cases the importance of selecting the best components is even greater and more difficult, than for non-fault tolerant solutions: the best set of components is not necessarily the components which best on their own, how effective the fault-tolerant solution is depends on diversity of the chosen components. In the paper I critically review the concept
