Re: Dons flame resistant (3 hours) interface about Linked Data URIs

> On 10.07.2009 10:53:32, Toby Inkster wrote:
> >What would it mean for the file to have a dc:created property? Would the
> >value of that property be my date of birth, or would it be the date I
> >first uploaded my data?
> >
> >The classic example is that if I use the same URL to represent myself
> >and my web page, then how can I state that I am the creator of my web
> >page without also asserting that I'm my own father.
> 
> By simply using two different properties?
> 
> These are the typical (and correct) arguments, but they are grounded 
> in an AI/logics purism(?) that *maybe* shouldn't be taken too seriously
> on the public SemWeb. They are of course practically motivated as well,
> but the practitioner here is someone with a reasoning background, not 
> necessarily a web developer in a web agency.
> 
> We could most probably use Hugh's approach/idea and still solve all 
> our practical problems.
> 
> Why did we give URIs to properties? To tell us what types of resources 
> they relate. They should support us, not restrict us. So, 
> 
> twitter:bengee is me (in Web 2.0 speak)
> 
> The page has a creation date: 
>    twitter:bengee ex1:created "2007" .
>    (ex1:created relates a document to a date)
> 
> I have a birthday: 
>    twitter:bengee ex2:birthday "08-14" .
>    (ex2:birthday relates an agent to a date)
> 
> The page has a creator:
>    twitter:bengee ex1:author twitter:bengee .
>    (ex1:author relates a document to an agent)
> 
> I have a father: 
>    twitter:bengee ex2:father "Bodo" .
>    (ex2:father relates an agent to an agent)
> 
> Now, this is totally blasphemic RDF *in our current view*, but
> heck would it make publishing easy. And with properly annotated
> properties, it would be easy to detect whether a term refers to
> a document or a NIR, and the syntax is pretty obvious about whether
> we are talking about a resource or the label of a resource. And hey,
> no more arguing about whether a vcard is a person or not. And we 
> could get rid of our =FCber-complicated XFN converters ;)
> 
> Simple querying works easily, directly on the instance data,
> and ontologies could be used for more automatic disambiguation.
> 
> So, dc:created can't tell you whether it refers to a person or a 
> document? Predicate FAIL, not Subject fail, maybe?
> 
> This is all rather tongue-in-cheek, of course, we've been here a couple 
> of times, I'm happy with the current specs, and different URIs for NIRs 
> and docs make a lot of sense, but we as a community should be prepared 
> that people will just use their homepages and OpenIDs as direct 
> identifiers (XFN, anyone?). Our apps will have to deal with the situation,
> and it's actually not too difficult to implement such a disambiguation 
> step. When I read a blog post and drag an author link on my address book,
> I want to add a person, not a page, and my address book should not say 
> "Ey dude, not a person" (well, would be cool if it could, though). 

I think we can make ex2:birthday work within the current understanding
of how URIs work in RDF, with all the correct formality.  We just say
that the ex2:birthday is doing a kind of indirect reference, which is
perfectly equivalent to some other RDF graph which uses the more common
direct style.

Specifically, the semantics of ex2:birthday are given by this rule:

  forall ?page ?date
      { ?page ex2:birthday ?date. } 
    iff
      exists ?person
         { ?page rx:primarySubject ?person.
           ?person ex:rdf-birthday ?date. }

where rx is http://www.w3.org/2008/09/rx, or some equivalent, and
ex:rdf-birthday is the more commonly found kind of birthday predicate.

I'm not specifically advocating doing this -- it has some drawbacks --
but I see the appeal.

     -- Sandro

Received on Friday, 10 July 2009 16:26:56 UTC