Re: one-way sameAs and friendOf links

On 7/18/12 11:57 AM, Michiel de Jong wrote:
> I am trying to deal with the authority of sameAs and friendOf links.
> Of course, if I set up a website and on there claim to be 'sameAs'
> Bob, then a user search engine should not take that information as
> authoritative and add the information i publish to Bob's real profile.
> It should only trust outbound links.

It should do any such thing.

sameAs has semantics. A processor either understands the semantics of 
such relationships or it doesn't. Even better, if it does, reasoning can 
still be optional. That's how its done. Anything less will simply leads 
to a myriad of problems, across the board.

> Same for friendship.

Ditto.

Relationship semantics should either be used or ignored. Manufacturing 
them in this matter is DOA.

> If i claim to be a friend of Bob, then the search
> engine should interpret only that i am 'following' him.

As per comments above, all you can hope for is relationship semantics 
comprehension or incomprehension on the part of a user agent. Following 
that, your left with how the user agent deals with its reality.

>   Only if Bob
> links back to me should it be displayed as a bidirectional friendship.
Yes, that's basically reciprocity .

>
> This means we can't just take the "parent" identity as the main row in
> search results. Because anybody can add a parent to anybody else, and
> that way hijack their identity.

You are heading back to WebID ACL land. I hope you are seeing that there 
is a reason why relationship semantics that are machine readable == 
extremely important.

A chunk of data represented by a graph can deliver this clarity on a 
platter once you separate RDF, Linked Data, and Structured Data. 
Relationship semantics isn't exclusive to any of the aforementioned. 
Neither is any particular syntax.


> But we also don't want to list one
> person 7 times simply because they have accounts on 7 different
> services/social tools. I find this a difficult problem to solve.

It a co-reference problem that solved via equivalence by name or 
inverse-functional property based relationships. Again, these issues 
have been analyzed and addressed for a long time. Sadly, the letters 
"RDF" are sometimes introducing a form of expensive cognitive dissonance 
to those that assume its all about RDF/XML and resulting semantic 
goobledegook or yore.

Mistakes from the past have been corrected. I encourage everyone to look 
at the world as it exists today. RDF is a model (EAV + denotative URIs) 
that's loosely associated with RDFS ( which adds more relationship 
semantics covering subclasses, subproperties etc..) and OWL (which adds 
more fine-grained relationship semantics covering: equivalence, 
symmetry, inverses etc..).

My $0.02.
>
>
> Cheers,
> Michiel
>
>
>


-- 

Regards,

Kingsley Idehen	
Founder & CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen