CONTEXT: In my GUI library, the component's class hierarchy is defined like this (simplified for the example): COMPONENT (the base/root class) <- GRAPHICAL <- FOCUSABLE <- CONTAINER Container components contain one vector of components (actually pointers to components) for each of those types. I've made this way so the GUI don't need to search for ALL the components in a form, test if it is, for example, focusable, and finally perform a "onKeyPress" event on it (it just traverses the "focusables" vector). Currently, when a component is added to a container, the variable "Component::baseType" is used to decide to what vector(s) that component will be added. For example, a "panel" (baseType == container) will be added to all the vectors. PROBLEM: I think this approach is somewhat inelegant and prone to errors. I know of the "dynamic_cast" approach to solve this question, and also of using virtual methods to return the base type, but I don't really know what is the prefered way of doing this sort of work. QUESTION: Do you guys have any advice on how I should proceed? Is the general design decision I've made that bad?