CURIEs: how we got here

CURIEs have come about from the work to integrate RDF in XHTML. I shall  
try and summarise some of the threads here.

For instance to generate the necessary triples from

	<meta property="dc:creator">Dan Connolly</meta>

it is necessary to recognise 'dc:creator' is a URI. Is this a QName?  
Pretty much, sort of. But a QName can only generate a subset of the URIs,  
since the suffix can only be a name (whilst the prefix can represent any  
URI-head). The obvious step is to allow the suffix to be any URI-tail, and  
then you can contract any URI.

We have long discussed the issues of syntax. For instance, once you  
recognise that 'dc:creator' represents a URI, then the obvious chain of  
thought is to allow them in other places where URIs are acceptable:

	<a href="cc:license" ...

Unfortunately, the QName syntax clashes with the URI syntax.

	<a href="http:get" ...

Is that a URI or a QName? We tried '[dc]creator' ':dc:creator' (and  
others) as examples of syntax that were distinguishable from URIs, but  
have always returned to the idea that RDF tends to use the dc:creator  
form, and that is what people expect. It also allows for a compatibility  
path from QName to CURIE without having to modify any documents. So to  
distinguish CURIEs from full URIs in places where both are acceptable, the  
decision was made to use the form href="[cc:license]".

Finally, we wanted people who didn't care about RDF to still write the  
same things, and still get the same effect, while still generating the  
triples, so that

	<link rel="index" href="index.html"/>

was still written the same.

Steven Pemberton

Received on Tuesday, 27 June 2006 12:36:14 UTC