In a well-formed XML document every element has a name and namespace even if some schema author doesn't know or doesn't care what it is. Its type is undefined, or the most general 'anyType', depending on your point of view - but of course a specialised processor can interpret the contents without any formally acknowledged type.
A dfdl parser generates a brand new 'document', and the schema is the only source of names. You're right to point out we need to think about it.
The specification already makes the 'id' attribute required, to take care of the name, but makes no provision for namespace. I suggest we treat this as a 'Qname', so that the schema author can use a prefix to specify a namespace for the generated element information item.
I'm not sure how this squares with XSD's specification of 'id', and I'm not sure how best to treat untyped items.
The schema reference should identify the wildcard component, but out specification is deliberately vague about that.
I would be reluctant to add an information set item. In particular, the infoset will become less useful and convenient if we allow anonymous nodes.
Simon
From: dfdl-wg-bounces@ogf.org on behalf of Mike Beckerle
Sent: Fri 2007-12-07 06:05
To: dfdl-wg@ogf.org
Subject: [DFDL-WG] DFDL infoset and wildcards