- From: Roy T. Fielding <fielding@apache.org>
- Date: Tue, 23 Jul 2002 23:06:15 -0700
- To: "Joshua Allen" <joshuaa@microsoft.com>
- Cc: "Jonathan Borden" <jonathan@openhealth.org>, "Graham Klyne" <GK@NineByNine.org>, <www-tag@w3.org>, <timbl@w3.org>
On Tuesday, July 23, 2002, at 10:14 PM, Joshua Allen wrote: > It's a "should", not a "must". Systems that will "break" are systems > which assume that an http: identifier identifies a hypermedia resource > meant to be accessed via a web browser. > > People SHOULD use http: identifiers only to refer to resources which > they intend to be accessed through HTTP. That statement is not the same as "http URI SHOULD identify documents", which is what we are having a disagreement about. > Sorry to sound frustrated, but this is THE most important issue to the > semantic web. There is nothing more important than agreeing on an > identification scheme that unambiguously (to the extent possible) > identifies things. How can it be THE most important issue when nobody can explain what difference it makes to the Semantic Web? Just saying it is important doesn't get you anywhere. I have no problem unambiguously identifying things with any URI, including http, so I simply cannot fathom why the topic keeps getting repeated. > I have commented on this issue on my personal web log: > http://www.netcrucible.com/blog/2002/07/20.html#a225 Oh, for crying out loud! I am going to try to explain this one more time to see if it gets through and then I'll give up. A URI is an identifier. The semantics of the resource it identifies are defined by the sameness of representations of that resource over time, not by any property of the identification system used to create that identifier. In short, a URI reference has the semantics that other people assign to it when they persist in trying to use it to refer to something useful. This isn't a property of the technology; it is a property of how people use the technology. The same is true of "words", whether or not you are prepared to admit it. That is the basis of every modern human language except French. Some identifiers are better (more persistent, more available, more whatever-you-like) than others. http identifiers are ideal for resources that can be accessed through HTTP via TCP/IP. Other identifiers may be better for some cases, but only if they come with a support infrastructure that makes them sufficiently useful for people. People decide what identifier is best for any given situation, and it is a fundamental principle of the Web that any URI scheme can be used as an identifier within any context or protocol element that requires a URI. XML namespaces, by the way, can be accessed through HTTP, as can robots, climate controls, beaches, and any other thing for which someone cares to provide a meaningful representation via HTTP. > (Here is the exact same argument in human terms. Sound like a Beckett > play, doesn't it?) > TimBL: People shouldn't say "bad" when they mean "good" > Roy: But people CAN use "bad" when they mean "good"; I heard > Michael Jackson do that before. I'd appreciate it if you didn't use my name in a context that is a deliberate deception on your part, since it will confuse readers of the archive. Besides, it makes you look like a total jack-ass. ....Roy
Received on Wednesday, 24 July 2002 02:06:54 UTC