Abstract
Since Banning first introduced flow sensitivity in 1978, the term has been used to indicate hard or complex data flow problems, but there is no consensus as to its precise meaning. We look at Banning's original uses of the term and some interpretations they have generated. Then we consider the multiplicity of meanings in more recent interprocedural analyses, categorizing a number of data flow problems. We also classify several recent interprocedural approximation techniques with respect to properties related to sensitivity and discuss additional data flow problem properties. Finally, we propose a definition for flow sensitivity that appears to capture much of the original intent and current use.