Re: [webfinger] Registration of a URN for WebFinger Properties

["Paul E. Jones" <[e-mail]>]

Will,

The reason is to have one set of identifiers that people can agree to 
use.  There are two types of identifiers, too:
     * Link relations types
     * Properties

As you point out, there are several link relations already defined.  
I've cataloged some of them:
http://www.packetizer.com/webfinger/link_relations.html

Some that are interesting to me did not exist,  so I made up values.

There are also properties, and I've not seen any described formally 
anywhere.  I've defined a ones I find useful:
http://www.packetizer.com/webfinger/properties.html

There are more defined out there, I'm sure.  And it's perfectly OK for 
applications to define whatever they want.

However, there are a few that I think really should be defined by the 
IETF.  If we don't do that, then we at least need to have an informal 
agreement on, for example, how to get a person's name.  Otherwise, some 
link relation types or property identifiers are likely to get defined 
with 10 different identifiers, yet carry the same semantics.

Paul

------ Original Message ------
From: "Will Norris" <[e-mail]>
To: "Paul E. Jones" <[e-mail]>
Cc: "Melvin Carvalho" <[e-mail]>; "webfinger" 
<[e-mail]>
Sent: 10/10/2013 1:48:59 PM
Subject: Re: [webfinger] Registration of a URN for WebFinger Properties
>There irony here is that WebFinger itself was created to solve the 
>problem of resolving an otherwise unresolvable URI (originally mailto 
>and later acct URIs).  URNs have the same problem, and since they have 
>no well-defined host, you can't even use WebFinger to resolve them.  I 
>know there have been some efforts to define URN resolution (rfc2483 et 
>al), but as far as I know none are very well adopted.  Using HTTP URIs 
>makes the most sense to me.
>
>But do we really need yet another registry of properties, the vast 
>majority of which I'm sure have been defined in a dozen other places?  
>Is there a reason why reusing one of these existing namespaces would 
>not work?  (and if something WebFinger specific really is desirable, 
>then we can continue using webfinger.net, which has already been used 
>for http://webfinger.net/rel/avatar/ and 
>http://webfinger.net/rel/profile-page/.  That was kind of the idea of 
>running it as a static site out of the GitHub "webfinger" org; it's 
>very easy to give others access to everything.  That's what we've done 
>with activitystrea.ms for several years now)
>
>
>On Sun, Oct 6, 2013 at 8:16 AM, Paul E. Jones <[e-mail]> 
>wrote:
>>Yes, any URI can be used to identify a property. However, there has to 
>>be some agreed scheme and structure for things defined in the IETF. 
>>The point of the suggestion was to specify that.
>>
>>We could use HTTP, but I've never seen that scheme used in IETF 
>>documents for this type of thing. I've seen URNs, though.
>>
>>Paul
>>
>>
>>
>>--------------------------------------------------------------------------------
>>From: Melvin Carvalho <[e-mail]>
>>Sent: Sun Oct 06 09:19:42 EDT 2013
>>To: "Paul E. Jones" <[e-mail]>
>>Cc: webfinger <[e-mail]>
>>Subject: Re: [webfinger] Registration of a URN for WebFinger 
>>Properties
>>
>>
>>
>>
>>On 5 October 2013 03:22, Paul E. Jones <[e-mail]> wrote:
>>>Folks,
>>>
>>>As you know, properties (both link and subject-specific properties) 
>>>are identified via a URI.  For applications that are defined outside 
>>>the IETF, those organizations are able to define any URI they wish to 
>>>use.  For any we might define within the IETF, however, we need 
>>>something.
>>>
>>>In anticipation of having such a need, I think we should create a 
>>>document along the lines of RFC 6755 for WebFinger.  Specifically, we 
>>>would define a URN sub-namespace as:
>>>
>>>    urn:ietf:params:webfinger
>>>
>>>One such potentially-useful subject-specific properties are "name", 
>>>which would be the the subject's name intended for human consumption. 
>>>  If you've queried my WebFinger server, you would know I currently 
>>>advertise my name in English and Chinese.  For the "default" name, 
>>>the URN might be:
>>>
>>>    urn:ietf:params:webfinger:name
>>>
>>>For language-specific variants, it might be:
>>>
>>>    urn:ietf:params:webfinger:name:zh-CN
>>>
>>>Defining the various properties and their meaning is an exercise for 
>>>another day, but I hope you see the value in defining the URN 
>>>sub-namespace.
>>
>>Cant this be done using traditional HTTP keys.  The advantage being 
>>that they can be systematically dereferenced using http GET, rather 
>>than having to look it up in a central registry in a non machine 
>>readable way.  This is what has been going on for 10+ years, with 
>>FOAF, schema.org, open graph protocol and others.  Have I missed 
>>something?
>>
>>>
>>>Thoughts?
>>>
>>>Paul
>>>
>>>
>>>_______________________________________________
>>>webfinger mailing list
>>>[e-mail]
>>>https://www.ietf.org/mailman/listinfo/webfinger
>>>
>>
>>
>>_______________________________________________
>>webfinger mailing list
>>[e-mail]
>>https://www.ietf.org/mailman/listinfo/webfinger
>>
>