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

Identifying indirectly 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 11:34:11 -0400
Message-Id: <058557EC-33B6-4F7F-8C8A-94365A247AF4@w3.org>
Cc: "Leo Sauermann" <leo.sauermann@dfki.de>, <www-tag@w3.org>
To: "Booth, David (HP Software - Boston)" <dbooth@hp.com>

While the approach of identifying people indirectly through their  
homePages works well when explicit, in the N3

    [  foaf:homePage <http://example.org/pages/alice> ]  age 10.

It is I think very detrimental in the long run to generate new  
predicates to  relate people's home pages (etc) to things, instead of  
the people themselves. It works in theory, but the RDF graph ends up  
missing out the important node about the actual person. Inference  
becomes much more complicated, unless you just regenerate the more  
logical properties.  So this technique I would restrict to use when  


<mailto:alice@exmple.com>   ex1:mAge 10.

<http://example.org/pages/alice> ex:iHeight  "1.6m".


Clearly , of all the things like home page,s FOAF pages, mailboxes  
and social security numbers which we use to identify people  
indirectly, we would only get data reuse if we stuck to the same  
one.  But for god reason, different applications use different ones.

The actual benefit of the liked data web is starting to show up, and  
it is through people using common IDs for people and common well- 
known properties.  Let's not have to make parallel sets of properties  
if we can avoid it.


On 2007-05 -30, at 17:09, Booth, David (HP Software - Boston) wrote:
> 3. There is a third good solution, which involves framing the  
> problem differently.  It is what I called the "shadow ontology"  
> approach: http://lists.w3.org/Archives/Public/public-swbp-wg/ 
> 2006Jan/0171.html
> This approach involves designing your ontology to *indirectly*  
> refer to something that is not a web document, by use of a URI for  
> a web document that describes that thing.   For example:
> 	<http://www.acme.com/people/alice>
> 		iPreferences:likesToEat
> 			<http://example.org/recipies/applePie> .
> may indicate that the person who is the primary subject of page  
> http://www.acme.com/people/alice likes to eat the food that is the  
> primary subject of page http://example.org/recipies/applePie .  In  
> essence, the iPreferences:likesToEat property has a range that is  
> "a web document whose primary topic is a person" and its range is  
> "a web document whose primary topic is a food".  In the  
> iPreferences ontology these classes might be called simply iPerson  
> and iFood -- the "i" prefix indicating that they are indirectly  
> identifying a person or food.  Of course, they could just as well  
> be called "Person" and "Food", with the explicit understanding that  
> they indirectly identify a person or food.
> In some ways this approach is easier and more straightforward than  
> maintaining related pairs of URIs, but it is only feasible if you  
> control the ontology and can design it accordingly.
> Note that this is different from the "Reference by Description"  
> approach described in section 6.2, because it does not require the  
> extra burden of an additional blank node and inverse functional  
> property.
> Note also that this approach can be used very conveniently in  
> conjunction with a 303-redirect service if it later turns out that  
> you want to use your data in conjunction with another oontology  
> like FOAF that already uses foaf:Person to refer to a person  
> directly, such as:
> http://thing-described-by.org?http://www.acme.com/people/alice .
Received on Saturday, 2 June 2007 15:34:14 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:16 UTC