Re: Summary of cert : key Domain discussion

On 31 Mar 2013, at 14:13, Melvin Carvalho <melvincarvalho@gmail.com> wrote:

> 
> 
> 
> On 30 March 2013 14:59, Henry Story <henry.story@bblfish.net> wrote:
> 
> On 30 Mar 2013, at 14:45, Melvin Carvalho <melvincarvalho@gmail.com> wrote:
> 
>> Summarizing the thread.
>> 
>> I proposed that the Domain of : key be more general than Agent, since a public key is just a number, more than just an agent could have such a key. 
> 
> The :key relation relates an agent to a public key of which he has the private key. 
> This means that one can understand what it is to verify that an agent has a key. 
> 
> Yes, so I think this is a bug that goes back to the time when WebID was branded as FOAF + SSL.  

An old bug for which after years of use you are the first to "discover".  
In any case this is not a bug, it's a feature request. You need a new type of relation for a
use case you still need to develop.

> 
> So then a WebID becomes part of FOAF.  I think the community will have to decide whether it wants to closely couple the TLS spec to FOAF or not.  For example, facebook dont use FOAF in their profiles, though they did consider it.
> 
> Actually I forgot to count Mo's opinion, so at present I think rdf : Resource is slightly ahead in the straw poll.
>  
> 
>> An example I gave was that an account (e.g. belonging to an agent) could have a key pair.
> 
> the whole question is what does it mean to "have a key pair" .
> 
> I think we can use the typical OO concepts of "has-a" and "is-a" for this.  So the subject "has a" key.  

There may be any other number of properties that you can use.
If you want the has_a relation then you can use rdfs:member property

http://www.w3.org/TR/2002/WD-rdf-schema-20021112/#ch_member

"he rdfs:member property is a super-property of the container membership properties.
(ie. each numbered container membership property has a rdfs:subPropertyOf relationship to the property rdfs:member)."

<> a foaf:Account;
    rdfs:member [ a cert:RSAPublicKey;
                            cert:modulus ...;
                            cert:exponent ...; ] .

Or perhaps any other number of properties will do. It will all depend on your use case and ontology,
how much you need to put into that relation. 


>  
> 
>> 
>> Henry raised the point that you would then lose an inference that the subject of a key was an Agent.  However you should really use rdfs : type for this imho.
> 
> My point was that 
>   - changing the domain means that you have developed a completely different relation which is not the same as cert:key.
>   - that it does not make things simpler as you initially claimed it would
>   - that it is ill defined
>   - that your reasons for wanting to introduces it rely on handwaving some ontology which has not been developed
>   - that you can't tell us what it would mean for the relation to apply to inanimate objects
>   - that you don't distinguish nonsensical from sensical applications of the term, when it is applied outside of the realm of agents.
> 
> I think you have something of a case here.  However I'd caution against using the "ambiguity" argument.
> 
> This has come up on occasion in the TAG and Tim's response is that we should try and get things working rather than be "ambushed" by ambiguity.
> 
> http://lists.w3.org/Archives/Public/www-tag/2012Oct/0086.html
>  
> 
>> 
>> Henry also raised the point that it would be possible to construct triples such as the eiffel tower having a public key.  Unsure if this is a big deal, or even a demerit :)
> 
> You could not tell us why you thought this was or was not nonsensical, what it would mean, etc... 
> 
> I did, perhaps you missed it, but ive repeated above, "has a"
> 
> http://en.wikipedia.org/wiki/Has-a

rdfs:member is the "has-a" relationship. It's well known and standard. Use that and see how
far you get with your use case.

>  
> 
>> 
>> There was a small straw poll which was pretty evenly balanced, slightly in favour of keeping the status quo.
>> 
>> If there's further interest in this I'll raise an issue at some point.  But I think I have enough to model what I want to now.
>> 
> 
> Social Web Architect
> http://bblfish.net/
> 
> 

Social Web Architect
http://bblfish.net/