Legacy and rdf:XMLLiteral

This message is really an appendix to the discussion on the pros and cons of 
the various rdf:parseType="Literal" designs.

One of the concerns expressed against the merging of plain and XML literals 
design [1] was its impact on legacy data.

My view is as follows:

All the designs we have considered, except that one, have no real impact on 
plain literals. 
For that design [1] most members of the WG seemed to see the legacy impact as 
very large, and those few that didn't admitted to doing intellectual 
somersaults in order to preserve the legacy. (I only really noted one 
participant who argued that such somersaults were possible).

However, it could be argued that the design in the september 5th WDs has a 
negative legacy impact on XML literals - since data for which the lang tag 
was intended as significant has to be reworked to include the lang tag within 
the XML Literal.

My view is that, that is indeed a problem - but that most use of 
rdf:parseType="Literal" needs to be revisited for all the designs.

As an example, the mathml example in RDF Model &Syntax Recommendation works 
somewhat differently in any of our designs than in any RDF implementation 
prior to the work of this Working Group. Any particularly behaviour or 
understanding intended for such an example is not necessarily preserved.
This is a consequence of M&S leaving aspects underspecified, such as equality 
between such literals.

As a further example, consider embedding HTML within a 
rdf:parseType="Literal". By clarifying how namespaces work we have ensured 
that the literal '<b>foo</b>'^^rdf:XMLLiteral is different from
'<b xmlns="http://www.w3.org/xhtml">foo</b>'^^rdf:XMLLiteral.
This cannot be read into Model and Syntax. It tripped up one of my colleagues 
this week.

As a final example, any data from an rdf:parseType="Literal" construct now 
stored in a non-RDF/XML format will need to be revisited, since all the live 
designs require such data to be in exc-C14N.

Hence, in the summary in my report (forthcoming), I indicate that *none* of 
the designs provide a wholly satisfactory solution for legacy 
rdf:parseType="Literal" data, but that all but one are satisfactory for 
legacy plain literals.

Jeremy

[1]
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Sep/0099

Received on Friday, 26 September 2003 05:11:28 UTC