I'd actually view the fact that the PI's are ignored by a validating
parser as an advantage. It allows the formatting to be ignored
downstream and if one decides to re-use a piece of content using conref
or some other mechanism, depending on how we build these applications,
we can safely throw out the PI's (knowing that it won't change the
validity of the topic). Unless we need to have some structural rules
about page breaks, i.e. only occur within paras or min/maxOccurs in our
schema, why not just keep them out of the DTDs/Schemas?
