- From: Jeremy Carroll <jjc@hpl.hp.com>
- Date: Thu, 3 Jul 2003 11:46:13 +0300
- To: w3c-rdfcore-wg@w3.org, duerst@w3.org
- Cc: w3c-i18n-ig@w3.org, "Ralph R. Swick" <swick@w3.org>
Martin: [[ > http://www.w3.org/2000/03/rdf-tracking/#rdfms-xml-literal-namespaces This is related in that the RDF Core WG, by using exclusive canonicalization, actually cares about namespaces inherited from (maybe far) outside the actual literal (apart from namespaces implied by qnames in attributes and element content, which is not RDF's fault). It seems very odd that at the same time, inheritance of xml:lang was thrown out, even though the later is clearly used in M&S, whereas the former is definitely unclear in M&S. ]] Exc-c14n is used to decide what is locally explicit and what is not: So, if for example there is an entity reference, this is expanded in accordance with exc-c14n, and hence the defn of that entity is relevant no matter where it was. Similarly the namespace nodes which in the infoset and the xpath nodeset are inherited, are locally explicit going by the rules of exc-c14n, i.e. if the namespace prefix is visibly used. The behaviour here is essentially local <eg:prop> locally uses the binding of the eg prefix, and that's why the namespace node is relevant and the binding will explicitly appear in the corresponding XMLLiteral. This sort of behaviour is necessary to cope with the mathml example in M&S. The WG could have decided that the MathML example was simply wrong, but as it is we found a design that worked for that example. As for not including the xml:lang, we are simply following exc-c14n. I would suggest looking at their documents for the rationale for that decision. Jeremy
Received on Thursday, 3 July 2003 05:46:38 UTC