- From: <noah_mendelsohn@us.ibm.com>
- Date: Fri, 23 Jun 2006 11:02:57 -0400
- To: Misha Wolf <Misha.Wolf@reuters.com>
- Cc: public-rdf-in-xhtml-tf@w3.org, semantic-web@w3.org, www-tag@w3.org
Misha Wolf writes: > 7f whether the IRI mapped to the prefix is required to be dereferenceable. I'm not sure I'd put it this way, as I think it's a matter of degree rather than "whether". Specifically: * I believe that, by definition, any http URI is dereferenceable, in the sense that it is always appropriate to try a GET, except perhaps if your intentions are malicious (denial of service, etc.) Thus, for example, it's always OK for a crawler to attempt a GET or HEAD on an http URI that it stumbles upon or cobbles up. * The TAG has stated that it's always good practice for the authority to cause representations to be available. Now, here's where I think it's a matter of degree. I can imagine circumstances in which, notwithstanding the TAG's general advice, one would go the other way. If, for example, a namespace were minted deep in the heart of some system, for uses really internal to that system, and from context we knew that it would be used for one or two documents of short lifetime and never again, well maybe then it's not worth the deployment cost of responding to GET requests. Still, it's a good thing at least in principle, so that if the crawler comes along, it gets a reasonable description. So, I think the IRI mapped to the prefix is by definition dereferenceable (if in the http scheme) in the sense that you can always try, and that resource authorities SHOULD in most cases offer a representation. There is some wiggle room between that SHOULD and a MUST. -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 --------------------------------------
Received on Friday, 23 June 2006 15:03:13 UTC