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

Re: article on URIs, is this material that can be used by the SWEO IG?

From: Tim Berners-Lee <timbl@w3.org>
Date: Sat, 2 Jun 2007 18:12:33 -0400
Message-Id: <13E64C73-988E-4844-8CB0-CF919C78645B@w3.org>
Cc: Leo Sauermann <leo.sauermann@dfki.de>, John Cowan <cowan@ccil.org>, www-tag@w3.org
To: Pat Hayes <phayes@ihmc.us>

Pat,

There are two ways of identifying things in generality, 303 and #,  
and the TAG and arch doc discuss those.  Don't hold the TAG  
accountable for the Topic Maps system of subject identifiers, or  
David Booth's related system for making predicates link a person's  
home page to their age.

Leo missed the point that in this particular example, the abstract  
thing happens to be an information resource, as well as the  
information resource about it. This happens often of course.
he was defending the point that it didn't have to be just because  
there was a 303 returned for it. So, that misunderstanding behind us,

On 2007-06 -01, at 16:08, Pat Hayes wrote:

>
>> Hi Pat,
>>
>> I try to answer the questions by referencing to the TAG  
>> resolution, I am not aware of other standards for concept URIs  
>> (besides #-uris).
>
> Agreed, there do not seem to be any. However, the TAG resolution  
> fails to address the central point, see below.
>
>>
>> It was Pat Hayes who said at the right time 01.06.2007 06:48 the  
>> following words:
>>>> Hi,
>>>>
>>>> I only kick in with some practical input ...
>>>>
>>>> It was Pat Hayes who said at the right time 31.05.2007 09:05 the  
>>>> following words:
>>>>>
>>>>>> Pat Hayes scripsit:
>>>>>>
>>>>>>>  > For example,
>>>>>>  > >http://en.wikipedia.org/wiki/Shakespeare is a subject  
>>>>>> indicator for
>>>>>>  > >Shakespeare.
>>>>>>>
>>>>>>>  But it also mentions Stratford-on-Avon, Mary Arden, and many  
>>>>>>> other
>>>>>>>  things. Why is it not just as much 'about' them?
>>>>>>
>>>>>> Because I don't so employ it.
>>>>>
>>>>> OK, fair enough. But then it follows that there is nothing  
>>>>> *intrinsic* to that resource that makes it be a subject  
>>>>> indicator for Shakespeare. It is so simply because you say it  
>>>>> is. But when I read that resource, how do I gain access to  
>>>>> *your* intention that it shall be a subject indicator? What  
>>>>> readable resource is it that tells me that http:// 
>>>>> en.wikipedia.org/wiki/Shakespeare is a subject indicator for  
>>>>> Shakespeare ?
>>>>

(I am not promoting the use of subject indicators  -- that must be  
your discussion with topic map people.)


>>>> Semantic MediaWiki has done that technically:
>>>> http://ontoworld.org/wiki/Semantic_MediaWiki
>>>>
>>>> You see how the links are done for example here:
>>>> http://ontoworld.org/wiki/Special:ExportRDF/Semantic%2BMediaWiki
>>>> <smw:Thing rdf:about="http://ontoworld.org/wiki/ 
>>>> Special:URIResolver/Semantic_MediaWiki">
>>>>  <smw:hasArticle rdf:resource="http://ontoworld.org/wiki/ 
>>>> Semantic_MediaWiki"/>
>>>
>>> No, I'm afraid I don't see. In this example, everything is an  
>>> 'information resource': both the Wiki and the article about it.
>> The URI http://ontoworld.org/wiki/Special:URIResolver/ 
>> Semantic_MediaWiki is not HTTP-200 and does a 303, therefore, by  
>> suggestion of the TAG, it may be an concept. So according to the  
>> decision of the TAG, there is a clear distinction here.
>> see
>> http://lists.w3.org/Archives/Public/www-tag/2005Jun/0039.html
>>
>>   a) If an "http" resource responds to a GET request with a
>>      2xx response, then the resource identified by that URI
>>      is an information resource;
>>
>>   b) If an "http" resource responds to a GET request with a
>>      303 (See Other) response, then the resource identified
>>      by that URI could be any resource;
>>
>> so I see no point here arguing that both are information  
>> resources, they are not, as decided above.
>
> But it does NOT decide this. What that decision says is that the  
> resource identified COULD BE ANY resource. In particular, it could  
> be an information resource. And in this case, as far as I can  
> understand the concept of "information resource", it in fact is.  
> The TAG decision provides no way to assert that a resource is NOT  
> an information resource.

yes.  But RDF allows one to assert that Shakespeare is a Person and  
OWL allows one to assert (if you need to know for some reason) that  
Information Resources and People are distinct.

> It also, more seriously, provides no way to link or connect any URI  
> to any non-information resource, so it seems to me that any claim  
> that any URI "identifies" any non-information resource is at  
> present simply bogus.

On the contrary.   <http://dbpedia.org/resource/Shakespeare> for  
example gives a 303 redirect to

<http://dbpedia.openlinksw.com:8890/sparql?default-graph-uri=http%3A% 
2F%2Fdbpedia.org&
query=DESCRIBE+%3Chttp%3A%2F%2Fdbpedia.org%2Fresource%2FShakespeare%3E>

which could assert  <http://dbpedia.org/resource/Shakespeare>  
rdf:type foaf:Person.

(In fact the data is gives s that Shakespeare was an influence on  
<http://dbpedia.org/resource/Frederic_Manning> and <http:// 
dbpedia.org/resource/Julius_Thomas_Fraser>, but that is off the point)


> [..] Bearing in mind that the latter guy has been dead now for some  
> considerable time, but that (according to the TAG), having an  
> "identifier" for him should enable us to "access" him by  
> dereferencing the URI (http://www.w3.org/TR/webarch/#dereference- 
> uri). Although many literary scholars might disagree, I rather hope  
> that the Web cannot do this, myself.

One expects to be able to get a representation of an Information  
Resource.  Not any old resource.  often in text, one uses 'resource'  
when talking about Information Resources.  That doesn't mean we will  
access the Bard.

In a discussion like this I suppose we should be careful.   I would  
say that you can 'dereference', or 'look up' the URI of a person. One  
can 'access' or 'retrieve a representation of' a document.   The way  
the architecture works is that when you dereference a URI, the de- 
hashing or the 303 (if any) gives you the URI of a document which you  
then access.

When you get the representation, if it is in RDF, it may have triples  
which tell you stuff about the
person, toaster oven, and so on.

> I can create such a URIref, yes. How to I attach it to my actual  
> oven? So that someone can use that URIref to "access" my actual  
> oven? This is apparently required for a Web identifier, see above.  
> Writing an article in English about my oven does not do this, we  
> have established. (And in any case, to hell with writing a  
> continuing blog about everything I want to refer to.)
>
>> (as Timbl favors for small ontologies), or you can install  
>> Semantic Mediawiki on any server of choice to make a page about  
>> your toaster
>
> You have GOT to be joking. How about the sand on Pensacola Beach?  
> One wikipedia entry per grain? (Look, I KNOW this is ridiculous.  
> But my point is that *this is what you guys seem to be saying*. So  
> what you are saying seems to be ridiculous. So I suspect you don't  
> really mean to be saying this at all. So, next question, what DO  
> you mean to be saying?)

Pat, if you want to write about a grain of sand, then do. Whatever  
turns you on. But if you don't write about it, it will still be  
there.  (God will write about it ;-)  JK).  Note every thing in  
existence needs to be given an ID.

> I know how to do this. But I refuse to go to all this bother, no  
> matter what the TAG says, when it does absolutely nothing whatever  
> towards solving the central problem. None of this architectural  
> doodlebugging can ever attach a name to a thing that is not  
> physically connected to the Internet. Y'all seem to be blind to  
> this very central fact.

Pat, seriously and with much friendliness, I can't make out whether  
you really see a gap in the the architecture or you are being  
philosophically pedantic.

The URI <http://dbpedia.org/resource/Tim_Berners-Lee>, when looked  
up, gives data about a person with my name and my date of birth, and  
enough other data to make a reasonable person quite convinced that  
the person is I.

Maybe you are being pedantic about grounding, and saying,

   "ah yes, but how does this person known that <http://dbpedia.org/ 
property/DATE_OF_BIRTH> really means the day someone was born on?  
(good question, it is 404).  But even if it were to dereference to an  
ontology, how do I know the ontology wasn't written by someone who  
uses english in a different way?"

   Is that the tack?  It sounds a bit like the "how to I know you  
mean green when you say green" discussion.

The philosophy of all this includes the way RDF terms we regard as  
being well-defined, are grounded though the use of  lot of English  
until, rather than talking about the sets  possible interpretations   
which would be consistent with the triples,  we relax and say, yes,  
there is one real world, and this URI identifies this one person in  
it.   Do get back out the philosophy of human language when analyzing  
breakdowns of communication, when debugging, but when not debugging  
it, permit the world to go on naming stuff with agreed names, and  
exchanging data about it.

Tim
Received on Saturday, 2 June 2007 22:12:54 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:32:52 UTC