The definition of conforming validating and non-validating processors in
section 5 is reasonably OK, subject to "a reasonable man's" interpretations
of certain words, especially the verbs "to report" (is an XML processor
allowed to allow the application to suppress the error reports?) and "to
process" (is the XML processor required to reveal the results of its
processing or can it throw it all away, or mangle it in some way after
processing?). A formal spec of an API such as SAX or DOM would be much
cleaner.
By contrast the defining occurrences of the terms processor and application
in section 1 (to which hyperlinks point) are hopeless, since they don't
define the boundary between the XML processor and the application: they
don't tell me, for example, whether my SAXON library is part of the XML
processor or part of the application.
Mike Kay