W3C home > Mailing lists > Public > www-tag@w3.org > November 2007

Re: Conforming is such sweet sorrow (was: Re: The meaning of "representation"

From: Tim Berners-Lee <timbl@w3.org>
Date: Sun, 25 Nov 2007 14:27:06 -0500
Cc: Chimezie Ogbuji <chimezie@gmail.com>, Pat Hayes <phayes@ihmc.us>, <www-tag@w3.org>, Mikael Nilsson <mikael@nilsson.name>
Message-Id: <F8B3FA9D-1989-44BA-867F-331AC8F106D2@w3.org>
To: wangxiao@musc.edu


On 2007-11 -25, at 14:06, Xiaoshu Wang wrote:

> <snip>
>> On a different (perhaps more constructive note), Pat, I think you can
>> achieve your initial intent by using a standard that embeds RDF
>> directly into the HTML such that the RDF captures the declarations  
>> you
>> have recorded in natural english.  I've taken the liberty of running
>> with this idea for your sake, because I think this issue is  
>> crucial :)
>>
> It still won't conform to httpRange-14 as long as anywhere in the  
> message it says "<http://www.ihmc.us/users/phayes/PatHayes> a  
> foat:Person."

Yes.

>  In httpRange-14's eye, the meaning of a message is not solely  
> dependent on what the message is but also on how the message is  
> delivered through the web.

teh binary of  amessage cannot be self-describing unless you bookstrap  
thrugh metadata which tells you what language the  bits are in.  You  
re-implement this in the system described on your document. In HTTP,  
the metadata is in headers.

A crucial architectural invariant is that, given just a URI, with no  
context, an agent can determine the information conveyed by the  
document and its intended meaning (for data).

>  RDFa's and GRDDL's RDF is *delivered* from client side, just like  
> fragment identifier, it doesn't count.
> I am very curious about the question that I raised in my document.   
> Which one of the following assertion true?
>
> (1) <http://www.w3.org/2000/01/rdf-schema#Resource> a  
> awww:InformationResource.

Clearly not true.
<http://www.w3.org/2000/01/rdf-schema#Resource> is  Class. Classes are  
not documents.
I assume <http://www.w3.org/2000/01/rdf-schema#Resource>  
owl:disjointWith  gen:InformationResource.

> (2) <http://www.w3.org/2000/01/rdf-schema#Resource> a  
> awww:NonInformationResource.

The concept of "Non-information resource" is not very useful, as  
people tend to use the phrase to mean "some thing, not necessarily a  
document", but when pressed they realize it can be a document too, and  
they meant Resource, ow owl:Thing.

The AWWW does not talk about Non-information resources.

> Of course, I am assuming,
>
> awww:InformationResource owl:disjointWith awww:NonInformationResource.

The AWWW does not talk about Non-information resources.
You would have to define it is you wanted to use it in conversation.

> Because if it is not true, i.e., there is something that can be  
> either IR or non-IR, then the definition of IR seems already  
> irrelevant (at least if we don't find another 30x code for that  
> mixed category with regard to httpRange-14).
>
> As everything in the web is a rdfs:Resource, either (1) or (2) seems  
> running into a paradox.  (I am not a logician.  If I am wrong,  
> please point it out for me.)

1 is false.
2 is not defined in the awww.

> The question has never been asked before. I guess it is because it  
> is denoted by a hash URI.  But if IR is indeed an objective  
> attribute of a resource, then we sure could ask that question  
> regardless what its name is.

Yes, true.  There are names in HTTP space which directly imply, if  
they return 200, that the thing they denote is a document. Like the  
URI for Pats document.

There are other URIs, those which have # or those which redirect with  
303, where you can't tell, but they can certainly be documents.   I  
can say in N3 in a document

	<#thisDoc>  owl:sameAs  <>.

and now I have two URIs for the same document.

>  Put it in other way, if the RDF scheme were designed with hash

you mean slash?

> ]URI, say rdfs:resource is denoted by the following URI,
>
> http://www.w3.org/2000/01/rdf-schema/Resource
>
> Should this URI 303 or not?

Yes, it has to. So that I have separate URIs for the class and the  
ontology document.

Note the hash architecture works fine here.
IMHO  hashes should be used under almost all circumstances.
Note there is now a complete parallel dbpedia using hashes instead of  
slashes.

Tim

>
>
> Xiaoshu
>
Received on Sunday, 25 November 2007 19:27:12 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:32:54 UTC