Re: [webid spec] overview section

On 20 Nov 2012, at 21:23, Andrei SAMBRA <andrei.sambra@gmail.com> wrote:

> On Tue, Nov 20, 2012 at 2:58 PM, Henry Story <henry.story@bblfish.net> wrote:
> 
> On 20 Nov 2012, at 20:50, Andrei SAMBRA <andrei.sambra@gmail.com> wrote:
> 
>> On Tue, Nov 20, 2012 at 2:36 PM, Henry Story <henry.story@bblfish.net> wrote:
>> The graphic showing the picture of TimBL should I think belong to an
>> overview section ( replacing the current section 3 ) that would explain:
>> 
>> I agree. I'll get right on it.
>>  
>>  1. The full uri denotes the agent TimBl
>>  2. that the uri minus the hash denotes the document
>>  3. that the document SHOULD describe the agent in a uniquely identifiable way, so that the
>>     agent can be distinguished from every other agent via this definition.
>>     ( so here one can specify that this is very general: a public key, a link to the profile document,
>>       a link to an e-mail address, any or more will do )
>>     ( the reason is that otherwise one would need a backchannel to know what the WebID refers to )
>> 
>> for 1 and 2 refer to the URI rfc spec on fragment identifiers.
>> 
>> ---
>> 
>>   I think it may be useful also to explain how these WebIDs can then be used to
>> create social networks - across servers. ( I can send a graphic of interlinked WebIDs for
>> that ).
>> 
>> Isn't this out of scope for the spec? It's up to each of us to decide how we want to use WebIDs.
> 
> Once you publish a WebID you cannot stop people from linking to it.  The point of WebIDs is
> to allow the build a linked data space: a social web.  This is not something that is a side issue,
> it is pretty core. We don't need to go into great details about this. We don't have to explain which
> relations must or should be used, but pointing out that one can link is important.
> 
> I agree, and I think that happens by default in the world of linked data, where URIs are dereferanceable instead of being an arbitrary graph name (like it happens for static RDF). By definition, the WebID URI is dereferenceable, so it's like saying that you have HTTP links links. Maybe I'm not expressing myself clearly...

I am not sure if you are disagreeing with me anymore.

It may be the default, but explaining it is important. Just like you also show how one can write a
profile document, or how URIs work. All of this is evident in some way.

From experience I know that I have spent a lot of time explaining to people that foaf profiles did
not have to be public. This not having been explained is part of what lead the OpenID group to
do their complicated attributed exchange protocol, where the user could choose which attributes
which user would see. 

The easy linkability of WebIDs is what makes them so useful.

>  
> 
> Whether the claims these people make are true or not is something else. I think we should say
> something on this topic too. If someone links to your webId saying they foaf:know you then an
> agent that reads that should only belieave that insofar as they believe the original document. If 
> there is a link back, then that counts as confirmation.
> 
> Agreed, but at this point, the agent has passed the WebID point, and reached the point where it uses returned data for a specific purpose.

I am just saying that we can hint at these things lightly in non normative contexts.


>>  
>> 
>> ---
>> 
>>   It may also be useful to show how one can have a WebID profile link to protected
>> profile document. ( an adapted version of the WebID-TLS spec graphics ), as otherwise
>> people will cry out that this creates anonymity problems.
>> 
>> 
>> Mentioning protected documents (ACL) at this point will open the door to a lot of questions. I guess it's the same reason LDP doesn't mention it either, and they really need it.
> 
> We don't have to say how the documents are authenticate an agent that requests it, just that
> there can be protected documents. Otherwise you will just get people thinking that all information
> has to be public and you will get the privacy folsk asking questions every day. So if you want
> to be here for the forseable future answering the questions of every privacy person who comes along...
> 
> Right, I see. So we could mention that one may split the profile document into multiple documents, and then add seeAlso relations to them. Would that be acceptable in a non-normative context?  

yes, in non-normative contexts. What the relation should be would be interesting to know but need not be settled. a sub relation of rdf:seeAlso I suppose.


> 
> Andrei
>>  
>> Andrei
>> 
>> 
>> 
>> 
>> Henry
>> 
>> 
>> 
>> Social Web Architect
>> http://bblfish.net/
>> 
>> 
> 
> 
> Social Web Architect
> http://bblfish.net/
> 
> 

Social Web Architect
http://bblfish.net/