| Selecting among a set of refinement alternatives is an important step in systematically analyzing non-functional requirements (NFRs). Such selection decisions, in the past have focused only on alternatives' contributions (synergistic vs. conflicting) towards NFRs. The concept of dependency, where functioning of one component(s) is necessary for the functioning of another component, has not been systematically analyzed, possibly leading to omissions or commissions of selection alternatives, hence affecting the concerned NFRs. In this dissertation, we present dependency analysis to further classify various types of dependencies across dimensions (e.g., partial vs. total, mandatory vs. optional), study their propagation when composed together and deduce properties that allow us to make effective selections among alternatives, for better meeting NFRs. Focusing on a smartphone operating environment, we couple the concept of redundancy to mitigate failure of alternatives. We illustrate the utility of such an analysis through a fall detection-response scenario, which has gone through an experiment with elderly people. |
