The W3C XHTML 2 Working Group has posted the candidate recommendation of
CURIE Syntax 1.0:
A syntax for expressing Compact URIs. This is modeled after namespace URIs and qualified names. In brief, it defines a prefix for a known base IRI (a URI that can contain non-ASCII characters like é),
then appends a colon and a local part.
For example, the CURIE
cafe:tradeshows.xml could be shorthand for
http://www.cafeaulait.org/tradeshows.xml if the prefix
cafe were mapped to the URL
Exactly how prefixes are mapped to base IRIs is left to the specification of the documents in which the CURIEs appear. However
if the CURIEs are in an XML document, then the namespaces in scope define the
prefix mappings. The default namespace can be used for prefix-less CURIEs.
Frankly I'm surprised to see this. Namespaces and the namespace syntax are one of the notable failures of the XML ecosystem. Why someone would choose to imitate this now that we know better is beyond me. Based on experience with namespaces, I predict that the problems of moving CURIEs from one context to another are going to be especially problematic. Well, we've learned to live with (if not exactly like) namespaces. I guess we can get used to this.