- From: Roy T. Fielding <fielding@gbiv.com>
- Date: Mon, 7 Jun 2004 16:47:01 -0700
- To: "John A. Kunze" <jak@ucop.edu>
- Cc: uri@w3.org
On Thursday, June 3, 2004, at 10:57 AM, John A. Kunze wrote: > The RFC2396 "identifier" definition, in its various forms through the > years, has not helped the discussion of persistent identifiers (URNs). > Here's the only definition I've seen that allows you to talk about an > identifier breaking without immediately tripping over metaphor bugs > that have to do with broken character sequences: > > An identifier is an association between a string and a resource. > > It is a feature, not a bug, that under this definition a URL ceases to > be an identifier -- becoming downgraded to a mere string of data -- > when that server later responds with a 404 Not Found. This would not > be a problem for RFC2396 because an association can be broken and you > still have a URI and its syntax, whether it identifies anything or not. There are a lot of myths about "broken identifiers". HTTP's 404 (not found) response code does not mean the identifier is broken. The 410 (gone) response does. 404 just means that no representation is available at the time requested. It is a feature of the Web that references can be made, via identifiers, prior to completion of the referent and independent of its continuity as a mapping. Confusion is due to thinking that the resource is the thing returned in response to a GET request; it is not. The resource in HTTP, regardless of scheme, is an interface against which methods can be applied -- it is presumed to exist until the origin server says otherwise. References to information supplied by that interface may become "broken" in any number of ways, not limited to 404 and 410, including unanticipated changes in the content between different 200 responses. Such references are inherently indirect: they depend on the people maintaining the mapping of URI to representation to retain an expected notion of sameness. This is no different from any other name resolution system, including DNS and the various methods for resolving URNs. Does microsoft.com cease to be an identifier when their DNS goes down? No, it merely ceases to be useful for some actions that require DNS resolution -- others, including those that have cached prior resolution attempts, remain unaffected by such temporary glitches. Cheers, Roy T. Fielding <http://roy.gbiv.com/> Chief Scientist, Day Software <http://www.day.com/>
Received on Monday, 7 June 2004 19:48:57 UTC