Re: Conneg representation equivalence

Richard Cyganiak wrote:
> On 21 Mar 2010, at 00:37, Kingsley Idehen wrote:
>>> I got the idea of something like this:
>>>
>>> <http://example.org/>
>>>     a foaf:Document ;
>>>     rdfs:label "Article"@en , "Статья"@ru ;
>>>     dcterms:title "Article"@en , "Статья"@ru ;
>>>     dcterms:creator <http://example.org/id/joe> ;
>>>     dcterms:created "2010-03-04"^^xsd:date ;
>>>     dcterms:hasVersion <http://example.org/index.en.html> ,
>>>         <http://example.org/index.ru.html> .
>>>
>>> <http://example.org/index.en.html>
>>>     a foaf:Document ;
>>>     rdfs:label "Article"@en ;
>>>     dcterms:title "Article"@en ;
>>>     dcterms:creator <http://example.org/id/joe> ;
>>>     dcterms:created "2010-03-04"^^xsd:date ;
>>>     dcterms:language <http://www.lingvoj.org/lang/en> ;
>>>     dcterms:hasVersion <http://example.org/index.ru.html> .
>>>
>>> <http://example.org/index.ru.html>
>>>     a foaf:Document ;
>>>     rdfs:label "Статья"@ru ;
>>>     dcterms:title "Статья"@ru ;
>>>     dcterms:creator <http://example.org/id/joe> ;
>>>     dcterms:contributor <http://example.org/id/jane> ;
>>>     dcterms:created "2010-03-17"^^xsd:date ;
>>>     dcterms:language <http://www.lingvoj.org/lang/ru> .
>>>
>>> Joe is the author of the original article, and he's credited as
>>> dcterms:creator. Jane is the translator, she is mentioned as
>>> dcterms:contributor for the Russian resource.
>>>
>> Problem: the Article and the Document (the resource at the URL) are 
>> distinct. Think "Book" and "Story in the Book" they are two distinct 
>> things with different characteristics (properties) that are connected 
>> by a relation.
>
> This is debatable. It rings true in *some* cases (for example, if a 
> newspaper article appears both in print and on the web, or indeed 
> books). But for text that is only written for the purpose of being 
> published on the web, having separate URIs for the “article” and the 
> “web document containing the article” is overanalysing.
>
> For example, if I publish a blog post at 
> </blog/2010/03/21/mypost.html>, then it would be overkill to create a 
> separate URI </blog/2010/03/21/mypost.html#this> to distinguish 
> between the “web document containing a blog post” and the “blog post 
> as such”. 
Blog posts (esp. if you look at the early Blogosphere era) tended to 
have Identifiers like:
<http://example.com/post#{date-of-post}> and this was from a group of 
folks that absolutely detested RDF (back in the RDF wars era that lead 
to RSS 2.0).

> One URI (mypost.html) is enough. The only reason for having separate 
> URIs would be if the author actually wants to assert different 
> properties about the two things, but that doesn't seem likely in the 
> typical web publishing case.
Yes, describing a "Story" distinct from its "Book" isn't overkill, the 
"Story", "Book", and even "Book Cover" are distinct things. And 
increasingly, once people grok Linked Data patterns, they will want to 
describe things completely as opposed to partially.
>
> (It might be a good idea to mint a URI for the *topic* (or topics) of 
> the article, for example if the topic is a person or project or event. 
> This allows me to make statements about this person/project/event, 
> which clearly has different characteristics than the article, so it 
> cannot have the same URI. But that's just a corollary to the old 
> linked data axiom that the topic of a document is a different entity 
> from the document itself.)
The content of the document is distinct from the document itself. They 
have different characteristics. When you attempt to generate a 
representation of the description of a Web Page where each component 
(Entity-Attribute-Value) is click-able (bar literal values) you will see 
there is a problem re. the "follow-your-nose pattern"  if you ignore the 
distinction I am concerned about. 

This example that might help:

Generic HTTP URI: http://www.lingvoj.org/lang/fr

Description Bearing Resource URL: http://www.lingvoj.org/lingvo/fr.rdf


URIBurner generated description of <http://www.lingvoj.org/lang/fr> : 
http://linkeddata.uriburner.com/about/html/http://www.lingvoj.org/lang/fr

URIBurner generated description of 
<http://www.lingvoj.org/lingvo/fr.rdf>: 
http://linkeddata.uriburner.com/about/html/http://www.lingvoj.org/lingvo/fr.rdf  
-- Note the ".rdf" resource has no relation to the secondary resources 
that constitute its innards (content in the form of structured 
descriptions of one of more "things").

ODE generated descriptoin of <http://www.lingvoj.org/lingvo/fr.rdf>: 
http://linkeddata.uriburner.com/ode/?uri=http%3A%2F%2Flinkeddata.uriburner.com%2Fabout%2Fid%2Fentity%2Fhttp%2Fwww.lingvoj.org%2Flingvo%2Ffr.rdf 
-- this simply generates a presentation of the representation of the 
".rdf" resource content.

The issue at hand is the fidelity of the description of an HTTP 
accessible resource, a resource that is compound in nature (even in the 
most basic form).

Kingsley
>
> Best,
> Richard
>
>
>
>>> dcterms:created for the canonical URI reflects the creation date of the
>>> original article. For the translation, it's substituted with the 
>>> date of
>>> translation.
>>>
>>> I'm not sure if we could also say that
>>>
>>>     </index.ru.html> dcterms:hasVersion </index.en.html> .
>>>
>>> I guess it depends on one's reading of the DC spec, probably nobody
>>> would die if this triple was included, but here I omitted it for extra
>>> precision.
>>>
>>> The Lingvoj ontology[1] also has provisions for describing translations
>>> as resources in their own right, but it's a bit beyond the topic here.
>>>
>>> As for dcterms:hasFormat, I don't think it can be used here, but if we
>>> also had a PDF version of the article, we could write:
>>>
>>>     </index.en.html> dcterms:hasFormat </index.en.pdf> .
>>>     </index.ru.html> dcterms:hasFormat </index.ru.pdf> .
>>>
>>> (and vice-versa)
>>>
>>> Further corrections/additions welcome.
>>>
>>> [1] http://www.lingvoj.org/ontology.rdf
>>>
>>>
>> For simplicity sake, just at "#this" to the following URLs and then 
>> reevaluate the model i.e, <http://example.org/index.en.html#this> 
>> which is the primary resource (a representation of the document which 
>> is basically a container) associated with a secondary resource (a 
>> representation of what the document is about). At the current time, 
>> your modeling discards the container i.e., treats resource 
>> "index.html" as none existent by describing it using properties of 
>> its content.
>>
>>
>>
>> -- 
>>
>> Regards,
>>
>> Kingsley Idehen          President & CEO OpenLink Software     Web: 
>> http://www.openlinksw.com
>> Weblog: http://www.openlinksw.com/blog/~kidehen
>> Twitter/Identi.ca: kidehen
>>
>>
>>
>>
>>
>
>


-- 

Regards,

Kingsley Idehen       
President & CEO 
OpenLink Software     
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen

Received on Sunday, 21 March 2010 18:29:11 UTC