In Push-MVC the model data is given to the view layer by putting it in scoped variables like request or session. Typical example is Spring MVC and Struts1.
Pull-MVC supposedly puts the model data in a common place i.e. in actions, which then gets rendered by view layer. Struts2 is supposedly a Pull-MVC based architecture, in which all data is stored in the Value Stack and retrieved by the view layer for rendering. I don't agree with this viewpoint since by definition of "view", it must be composed of pure HTML and/or JavaScript. Thus niether tags nor the Value Stack counts as "view layer" as both of these are Java or handled by Java before.