- From: Benja Fallenstein <b.fallenstein@gmx.de>
- Date: Mon, 28 Jul 2003 18:46:35 +0200
- To: Jon Hanna <jon@spin.ie>
- CC: www-rdf-interest@w3.org
Jon Hanna wrote: >>However, if I pointed my browser to <http://example.org/1434>, I would >>not consider it correct behavior if my browser opened a HTTP connection >>to <http://example.net/~foo/bar> and showed me the content that the web >>server at that address returns -- even if I have proof that these two >>URIs denote one and the same resource! > > Why not? > > If you know they are the same resource why do you care where the > representation came from? Ah, now I think we're getting to the core of the matter. If a browser is really allowed to do this, "a URI only denotes a resource" should work as a model. A simple problem with this is: I can make an HTML page and publish it at uri1. Additionally I can state (maybe in metadata inside the page) that uri1 owl:sameAs <http://www.w3.org/>. If your browser retrieves this page through HTTP, it can conclude that uri1 *really* is owl:sameAs <http://www.w3.org/>. Why? Because the person who owns that URI (me) has made a statement to that effect. Even if everything that I claim about the resource identified by both uri1 and <http://www.w3.org/> is in fact not true, that doesn't change the fact that the two URIs identify the same resource; as the owner of uri1, I can state that axomiatically, which makes it true... Therefore, following your logic, your browser could serve *anything* I put up at uri1 when you ask it for <http://www.w3.org/>. (If <http://www.w3.org/> denotes a web page, that's not a problem; I can still create another URI to denote the same web page!) Perhaps this gaping security hole can be fixed. There are more difficult cases, though. Assume that the following URI denotes a person: http://example.org/~mjk/ This person may have an associated email address, web page, name, age and so on. It's useful then to give only this URI when speaking about the person, e.g., "I got this idea from <a href="http://example.org/~mjk/">MJK</a>." My user agent could conceivably retrieve the information that <http://example.org/~mjk/> is a person with an e-mail address, so when I right-click on that link, it could offer me to "Send e-mail to this person." Or it could have a "More information about" option which would search the Sematic Web and show me a description about the thing defined by the URI-- the name, age and so on. Thus, it's useful to link to the URI of the person, not the URI of their home page. OTOH, it could also allow me to retrieve a representation of the person through HTTP (simple left click, I guess), which would probably end up being that person's home page. Now, assume that there are other pages about this person on the Web, for example a page from their employer, a page on an "I hate these people" site, and so on. When clicking on the link, in the tradition of Web pages I'd still expect to be taken to the Web page that the person of the link referred me to, not to some other description of the same person. I think that this is described well by what Peter said-- when I make the link in HTML, it not only represents a resource, but also gives a particular point of view on that resource (in Peter's terms, an intension). - Benja
Received on Monday, 28 July 2003 12:48:18 UTC