- From: Joshua Allen <joshuaa@microsoft.com>
- Date: Wed, 17 Jul 2002 00:35:12 -0700
- To: "Tim Bray" <tbray@textuality.com>, "www-tag" <www-tag@w3.org>
Start with: http://lists.w3.org/Archives/Public/www-rdf-interest/2002Apr/0124.html But just to clarify what we are NOT arguing about: 1. Everyone (even RDF) agrees that "http: URIs" should be used to identify resources which are generally lumped together and called alternately: hypermedia, hypertext, documents, web pages, and so on. 2. Everyone also agrees that "http: URIs" should be strongly preferred for identifying resources, IF those resources are most naturally dealt with through transfers of representational state. (In other words, if you envision interacting with the resource primarily through a web browser UI and synchronous request+response pairs, use the http: scheme) Those use cases for http: identifiers are well-established. The proponents of expanding the range of http are making three generalized arguments: A. Some people claim that *all* resources which one would care to identify can (and should) be dealt with through REST, and therefore rule #2 applies. And even if you think that a web browser UI and request+response interaction makes absolutely no sense to your class of resources, these people want you to use http: identifiers *anyway* -- Web browsers are only good at doing http: so you might as well name all things in the world http: "just in case" it turns out that they could be useful in http -- that way there is a slight chance it could one day work in a web browser. B. Some people claim that the idea of "hypermedia" can (and should) be ambiguous enough to encompass all named things, and therefore rule #1 applies. Just give it a new mime-type (object/car). C. Some people claim that identity is inherently ambiguous, and therefore URIs are meaningless to begin with. Since a URI doesn't *really* identify anything, it doesn't matter what scheme you use. This is the perversion of "minimally constraining". A, B, and C are what people are arguing about. (minor comments inline below) ---------------------------------------------------------------------- > problem is. In RDF, I can assert that the resource > http://example.com/23ru30u2 has a property named "Title" whose value is > "Lorem ipsum". It may be the case that an HTML representation of that You are describing scenario #1, so there is no disagreement. It is fine to make RDF assertions about things that are identified using http: identifiers. RDF does not discriminate based on scheme. You will get challenged if you decide to use an http: identifier to identify something that is clearly NOT in scenario #1 or #2 above. However, from the little we can glean from the example, using an http: identifier seems appropriate according to those two generally accepted conventions. > My own view of what a resource is may be found in > http://www.textuality.com/tag/s1.1.html That seems OK to me. > out: given a URI, while you can potentially retrieve a representation of > the resource, you can't find out what the resource is. There is no From the POV of the semantic web, it's the opposite. You can potentially find out what the resource is, but there is a good chance that you will NOT be able to dereference it. In other words, synchronously retrieving a representational state of a resource is not the only thing that you can do with a resource identifier. In fact, I expect that the vast majority of entities which are identified on the semantic web will NOT support synchronous retrieval of representational state. Of course, the subset of entities which *are* aptly identified with http: identifiers will be equal citizens in the land of RDF, but it is falling into trap "A" above to think that *all* entities should be identified with http: identifiers. > http://weather.yahoo.com/forecast/MXOA0069.html and realize that the > resource is really "Yahoo's weather forecast for Oaxaca". In fact, this > is why we need RDF or equivalent - to provide a standardized way to make Again, this is scenario #1. People normally think of a "weather report" as being something that they read, and it fits the "hypertext" category very nicely. RDF can certainly make assertions about these resources. However, RDF is not constrained to only making assertions about things that easily fit into the HTTP mould.
Received on Wednesday, 17 July 2002 03:36:34 UTC