The default content generator produces CSS-compliant content structures (also known as DOM-Trees). This model explicitly allows block elements to contain other block elements and to do all kinds of weird mixing of elements.
The content generator already has all knowledge to compute the manual pagebreaks. So this class could already fire pagebreaks. The next element in the layouting chain will be the Layouter itself. The renderer computes the element positions,knows about the page sizes (and whether we have pages anyway) and more important - keeps track of the current page and the page fill state.
The content generator is conceptionally a part of the normalizer step, and thererfore has access to the raw display model. The renderer receives a more generic view with no object hierarchies at all - if that whould be needed, it has to be rebuilt. (This simplifies serialization, so that most remote calls are now self-contained and do not supply a possibly contradictionary context (in terms of object-identity)).
@author Thomas Morgner