- From: Roy T. Fielding <fielding@apache.org>
- Date: Fri, 4 Oct 2002 14:30:32 -0700
- To: "Joshua Allen" <joshuaa@microsoft.com>
- Cc: <www-tag@w3.org>
> "Via that interface" is particularly useful in clarifying the range of > HTTP. Clearly, the URL which HTTP uses identifies the *interface*, and > not anything else. Clearly, you haven't been paying attention. The fact that the identifier is an address to an interface does not prevent it from being used as an identifier for anything at all. It only provides one default opportunity for use of that identifier, and one that isn't even guaranteed to respond with 200 OK. There are a million other uses that could only use it as a name and have every single property of persistence as any other name assignment. > Sometimes people might point to a family portrait hanging on the wall to > identify a particular person in conversation. The picture frame is > clearly *not* the person, it simply serves as an identification proxy. > When people want to be far less ambiguous in identifying a particular > individual, they tend to use identity proxies which are specialized for > unambiguously identifying individuals (SSN, etc.) SSN is an address for an account on the Social Security Administration's database. The fact that the IRS and MIT both use the same address to identify people, in spite of the fact that one person can have multiple SSN, should demonstrate to you that the mechanism for defining the name has very little to do with the mechanisms for using the name, other than the fact that the naming authority determines how much trust you can have in the mapping over time. Besides, anyone who owns a domain name can use their namespace as a one-to-one mapping to SSN, which produces identifiers which are every bit as useful for identifying people as the original SSN. All identifiers are identity proxies. You claim that people use the "http" URI to identify the interface. I claim, and have demonstrated with MOMspider, that authors use most "http" URI in the hopes of identifying a sameness of information over time -- otherwise, there would be no need for them to correct references when the information changes to something they didn't intend to reference. Furthermore, the action GET does not retrieve the resource, but only a representation of the resource, and thus no assumptions can be made about the resource based purely on its responses to the GET method. Finally, the fact that most http URI do identify information sources does not in any way prevent some http URI from identifying things that are not information sources, as is the case for information sinks and resources that have no interface at all, and there is no reason whatsoever to claim that such resources are any less useful to the Web than those that seem to be "documents". ....Roy
Received on Friday, 4 October 2002 17:33:08 UTC