- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Mon, 11 Mar 2002 16:28:28 -0000
- To: "Dave Beckett" <dave.beckett@bristol.ac.uk>
- Cc: "w3c-rdfcore-wg" <w3c-rdfcore-wg@w3.org>
The other extreme that I see in the possibility space is a proposal like: ======== Propose that: - the exact form of the string value corresponding to any given XML Literal within RDF/XML is implementation dependent. - the string value is well-balanced XML that can be inserted as the elment content between two tags: <foo></foo> to form an XML document, satisfying both XML and XML Namespaces. - taking the exclusive canonicalization of both the original XML Literal in its containing document, and the string value within the dummy "<foo></foo>" document produce the same character string. Equality between xml literals is defined on this basis, but only for the purpose of exercising the test cases. - that the canonicalization above is without comments - we create some test cases (e.g. with qnames in attribute values) for which interoperability difficulties may be observed ======== What this means is that implementations must: - preserve all visibly used namespaces - not leave dangling qnames in element tags or attribute names - not leave dangling entity references - preserve whitespace (except in tags) - preserve processing instructions are free to: - ignore or keep comments - use any form of the XML that is Infoset equivalent - include any namespaces (whatsoever) that are not visibly used, e.g. to treat a declaration as a visible use. - be written in XSLT, and/or have XSLT preprocessors (not compatible with treating declarations as visible use) - not implement C14N This does not suffer from namespace pollution. Jeremy
Received on Monday, 11 March 2002 11:28:48 UTC