Re: opaque uris and self-describing resources

Michaeljohn Clement wrote:
> Erik Wilde wrote:
>>>> so i assume to discover the non-http nature of the resource 
>>>> identified by u1, there must be some content within the returned 
>>>> resource that makes that statement. logically, i see three ways how the 
>>>> non-httpness of the identified resource could be established:
>>>> 1. string matching with a magic prefix
>>>> 2. the 303 returned when dereferencing the uri
>>>> 3. embedded metadata in the returned resource
> Sandro Hawke wrote:
>>> The TAG reached consensus on 15 Jun 2005 to use option 2.
> What I understood the TAG resolution to say is that the 303 implies 
> only the *possibility* that the resource might be something other than 
> an information resource.  The 303 does not indicate that the resource 
> *is not* an information resource.  (What I think Erik meant by 
> "non-httpness".)

yes, this is what i was referring to, and i understood the consensus in 
the same way. and i think there is no other way to understand it without 
breaking http/1.1.

> The 303 approach gives you a way to use HTTP URIs for non-information 
> resources, and it gives you a way to dereference those URIs and find 
> related information, but it does not give you a way to unambiguously 
> declare, merely by a status code, that that an HTTP URI identifies a 
> non-information resource.  (It also doesn't retroactively declare all 
> existing 303-returning resources to be non-information resources, 
> which would clearly be a mistake.)

yes, the 303 approach is very semantic web inspired, it allows semantic 
web users to do what they want to do, but does not at all add any 
semantics to the "plain web". which is what i would prefer.

> If that's the case, then you can't at all get from a 303 to Erik's 
> "non-httpness of the identified resource".  For that you would need 
> option 3, option 1, or some other out-of-band information.

this is what i was asking: (2) tells you that something may be some 
special resource, but since http/1.1 allows 303 to signal all kinds of 
things, there must be some additional magic to find out that a uri in 
fact identifies a non-http resource. the only possibilities i see here 
are (1) and (3), and while (1) was the approach championed by a number 
of people on this list, a true semantic web believer would probably go 
for (3). so i was simply curious to find out whether the 303 approach 
had any preference or even recommendation here...

thanks and cheers,

dret.

Received on Tuesday, 22 January 2008 03:20:54 UTC