Re: Issue-57

Tim Berners-Lee wrote:
> My software would assume that <http://www.knox.edu/Images/_News/news_media/img/2005/obama-barack-1ss.jpg>
> was a document before doing any HTTP request, just as it assumes that <mailto:timbl@w3.org> is a mailbox
> before sending any mail.  However, others wanted to be able to use http: URIs like that for arbitrary things.
> The TAG defined a compromise, so that you can only assume that it is a document if you get back 200.
> I changed my code.  I would have preferred otherwise, but I went along with the compromise 
> in order to get consensus and get interoperability.

I think I understand and agree with the thrust of this, but the wording seems 
open to interpretation when you say "you can only assume that it is a document 
if you get back 200".  I think that getting an HTTP 200 response is just *one* 
way that allows you to assume it is a document (if you trust the origin server).

If one had (and trusted) an RDF document containing:

   <http://www.knox.edu/Images/_News/news_media/img/2005/obama-barack-1ss.jpg> a 
ex:Photo .

(where ex:Photo denotes a class of things we know of as photographs, which are 
understood to be documents), then that is another way to justify this assertion.

...

Thus, what I think we have here is that by performing an HTTP GET and receiving 
a 200 response allows us to infer:

[ a ex:webService ; ex:runBy [ ex:withDomain "www.knox.edu" ] ]
   ex:claims
{ <http://www.knox.edu/Images/_News/news_media/img/2005/obama-barack-1ss.jpg> a 
ex:Photo }

which under appropriate assumptions of trustworthiness may allow you to accept 
the inference:

<http://www.knox.edu/Images/_News/news_media/img/2005/obama-barack-1ss.jpg> a 
ex:Photo .

...

As far as I can tell, a 303 response doesn't of itself suggest any specific 
inferences, not even a negation of the above.

I'm not up to speed on discussions of a 209 response, but I'm guessing the 
intent would be to license some different inferences.

...

The bottom line, I think, is that the form of an HTTP URI (without fragment) 
tells us nothing about its denotation, but that we have network mechanisms to 
help us find out.

#g
--



Related, I think:

[[
Constraint: Web software MUST NOT depend on the correctness of metadata inferred 
from a URI, except when the encoding of such metadata is documented by 
applicable standards and specifications.
]]
-- http://www.w3.org/2001/tag/doc/metaDataInURI-31

Received on Sunday, 26 June 2011 14:24:56 UTC