version numbers and infosets

The alert reader will have noticed an inconsistency, or at least a
tension, between my last two articles.

In one I suggested that a given document should have a single infoset.

In the other I suggested - in line with the current XML 1.0 spec -
that an XML 1.0 parser would accept XML 1.1 documents that were
well-formed XML 1.0 apart from the version number.

But of course an XML 1.0-only parser cannot generate the correct
infoset for all such documents, because of the changed whitespace
rules.  And it may generate an infoset for a document that doesn't
have one if XML 1.1 is not a superset of 1.0 (eg if it makes
unnormalized unicode a well-formedness error).

This might be taken as an argument for requiring 1.0 parsers to reject
1.1 documents.

-- Richard