- From: Tim Bray <tbray@textuality.com>
- Date: Mon, 20 Mar 2006 07:42:57 -0800
- To: Dick Hardt <dick@sxip.com>
- Cc: Digital Identity Exchange <dix@ietf.org>, John Merrells <merrells@sxip.com>, Lisa Dusseault <lisa@osafoundation.org>, uri@w3.org
On Mar 20, 2006, at 6:54 AM, Dick Hardt wrote: > Agree that we need to make good use of scarce community resources. On top of which, using an "http:" URI has the huge advantage that you can put something useful there that a human being can point their browser at and learn something about what this string means. I'd advise those who are promoting the new URI scheme to check out the remarks about URI schemes in http://www.w3.org/TR/webarch/ and in particular http://www.w3.org/TR/webarch/#URI-scheme - when you are doing something that flies in the face of a SHOULD recommendation in the "Architecture of the World Wide Web", you need to have really thought it through. > The reasoning behind introducing a new scheme was we need an escape > sequence for processing the name/value pairs and to differentiate > data from constants etc.. Anything starting with "dix:" is known to > be a constant. Anything else is not. :) -- one of the reasons for > this is that we want to be able to pass through name/value pairs > that a web application may be using to preserve state. We think the > likelihood that any existing app have strings that start with > "dix:/" to be, well, really really small. You're saying that you need a reliable way to ascertain that this identifier contains name-value pairs that should be processed per DIX? Can this be established by context? Any other signaling mechanisms? If not, then why are you buying into all the expectations and culture that go with using a URI, why not just use an opaque string or a chunk of XML or ";"-separated name-value pairs or whatever? A URI is supposed to identify a resource, and a common expectation is that one way or another, you can use it to retrieve information. If you avoid specifying one particular scheme, you leave the door open for different modes of information retrieval down the road. > Does this make sense? Do you have a suggestion for another approach > that provides an escape mechanism and allows decentralized > property /capability extension? In general, naming things with URIs is a good and Web-friendly thing to do; as you point out, the use of the authority field allows distributed authority over parts of the namespace. In particular, HTTP URIs have a lot of advantages. -Tim
Received on Monday, 20 March 2006 15:42:49 UTC