- From: Graham Klyne <GK@Dial.pipex.com>
- Date: Sun, 30 Jul 2000 22:25:31 +0100
- To: caro@Adobe.COM
- Cc: RDF-list <www-rdf-interest@w3.org>
Ouch indeed! I think any form of restriction on or special interpretation of the form of URI used in an RDF model flies in the face of the principle of using RDF to make statements about or linkages between any web resource. I suppose one might treat things described by an RDF schema (properties, classes) specially, but this seems to erode the generality of the RDF model. It seems to me that the requirement to find the namespace-related portion of a URI in isolation is not reasonable. It's also not clear to me what purpose it serves. #g -- At 11:41 AM 7/28/00 -0700, Perry A. Caro wrote: >Gerard Maas wrote: > > > > I have seen '/' as well. If I'm not mistaken, the rdf-api assumes either > > > a "#" or a "/" separating the namespace from the local name. > >Ouch. What about the class of schema URIs that use ":" as a separator? >URNs, for example? > >For any RDF processor that uses concatanation for expanded names, I would >recommend the following best practice: When an expanded name needs to be >decomposed into its namespace and local parts, take the URI and do a reverse >scan, character by character, from the end of the string, until a character >is found that is not legal as an XML name, OR until a colon ":" is found. >The last clause is necessary because ":" is legal in XML names. >This practice broadens the field of well-formed RDF that processors will >accept. Why rule out "+" or "?" or "="? > >It would also be nice if the RDF Schema spec made a VERY STRONG >recommendation that namespace URIs for RDF end in either "#" or "/". That, >with the addition of ":", is what our internal namespace guidelines >recommend. > >Perry ------------ Graham Klyne (GK@ACM.ORG)
Received on Monday, 31 July 2000 14:11:33 UTC