The input feed shields the internal processing from users errors. It implements a state maschine, which checks that all documents are well formed and which does not allow users to manipulate the resulting document tree directly.
An input feed collects all data for elements and other nodes and forwards them to the normalizer. The normalizer is the first stage of the content layouting and processing.
Pagination ability is not propagated back to the caller. A caller will not normaly know whether a certain input caused a pagebreak. However, especially in the cases where only one page should be processed, we allow the detection of page breaks using a boolean flag. (Which is also used to detect loops.)
The flag is reset on each call to 'startElement', 'startDocument', 'endElement', 'endDocument' and 'addText'. Attribute modifications have no effect on that flag (as these calls are accumulated into one big supercall before passing them to the normalizer.)
Processing the meta-info also has no effect on the page-break flag, as meta- info is processed before the content is processed.
@author Thomas Morgner