So by "default" format are you suggesting that
a) ZNG recommends that anyone with a vocabulary (such as Ray at present)
should use RDF to specify an XML instantiation of that vocabulary.
Providing that the structure of this XML instantiation can be given a
unique identifier (a URL), we can use this as the RecordSchema
designator in ZNG
b) the RecordSchema in the ZNG Request or Response could just reference
a vocabulary - in which case the XML instantiation is derived from the
vocabulary in RDF. This assumes that given a vocabulary it is possible
to derive a single or canonical RDF structure (i.e. how do we cope with
two servers having very different RDF expressions of the same
c) something else?