Though it is a stupid question, is there any advantage i will get if i use boost property_map instead of std::map for my name value property pairs ? I am working for a statistical system, where i have certain parameter names (20-50 names at max) mostly fixed defined in terms of compile time constants (enums) but sometimes are from a variable set, and values which changes during adaptation (a small statistical distribution struct contains mean & sigma usually). so far i have used things like, enum height{ class1, class2,class3}; struct dist{ float mean; float sigma; }; typedef std::map<height,dist> height_prop; ( i also thought of having a compile time lookup based on a lookup table, but not sure if that is worth esp when search is among small no of elements) and similarly other properties. so will i get any benefit using boost::property_map instead of std::map (or unordered_map ) . in future i may want these properties to be associated with a Boost graph node & edge. thanks abir