James makes a good point that limiting API features to known examples of applications potentially lowers the ceiling on how well an API can adapt to unanticipated needs.  I think he's right.  But it's only one factor.  When multiple new features are competing for a slot in the development schedule, another important factor will always be the relative demand for that feature.  So it's always good to understand what scenarios a feature will enable, whether demand already exists, and, as James mentions, the strategic value of widening the API and enabling new innovation.  (as well as other things, like cost, completeness, etc).