Re: More on distinguishing information resources from other resources

On Jun 28, 2005, at 21:32, ext Roy T. Fielding wrote:

>
> Please join me in a chant of "It just doesn't matter."

It just doesn't matter.
It just doesn't matter.
It just doesn't matter.

;-)

>
> We could spend a great deal more time trying to precisely define
> the meaning of "information resource", and yet doing so solves
> no known problem in the architecture.

+1

> The goal of the resolution
> is to provide a way to supply information about a resource that
> is not actually "representable" on the Web, where representable
> is entirely dependent on what the owner of that URI intends it
> to identify, and yet do so in a way that would allow the
> Semantic Web to distinguish between the resource identified by
> the URI and the other resource providing the description.
>
> In short, it licenses the Semantic Web to consider a 200 response
> to GET to be indicative of a representable resource, whatever
> that means, while describing how to provide information about a
> non-represented resource at minimal extra cost to the old Web.
> Whether this distinction actually turns out to be needed or not
> is besides the point -- simply defining it solves the semantic
> problem at minimum cost and allows us to get back to describing
> how the system works rather than what resources mean.

And the extent to which sw agents will trust the response code
as a reliable indicator of class membership will depend on
the degree to which folks follow the best practice outlined
by the TAG.

But, again, it should be made clear in the TAG's resolution
that it is *not* an error if an existing web application
returns a 200 response for a URI which does not identify
an information resource. The TAG's resolution might make
legacy systems non-optimal, but it should not make them
broken.

>
> Whether a given resource should respond to GET with a 303
> instead of a 200 is entirely dependent on the intentions of
> the URI owner and entirely under the control of that owner,
> so there is no need for us to talk about it any further
> beyond what we have already noted: a URI should only be used
> to directly identify a single resource.  In other words, 303
> should be sent when the owner thinks the requested resource
> is not, for whatever reason, represented directly, though
> information about it can be obtained from the other resource.
>
> When a URI does get used in an ambiguous manner, the SW now
> has a defined algorithm for disambiguation that will allow it
> to note such errors accordingly. When a purely conceptual
> resource is identified using a URI, the hypertext Web can still
> be used to follow the chain of links to provide more information
> about the resource without causing ambiguity.  Everyone wins.
>

Agreed.

I'd love to see the above text included in some manner into
the TAG's resolution on httpRange-14.

Regards,

Patrick


> Cheers,
>
> Roy T. Fielding                            <http://roy.gbiv.com/>
> Chief Scientist, Day Software              <http://www.day.com/>
>
>
>

Received on Wednesday, 29 June 2005 14:35:40 UTC