Re: [webfinger] Absolute URI vs URI

[John Bradley <[e-mail]>]

I don't personally think the additional flexibility is worth the interop issues likely to be created by different parsers of the JRD attempting to interpret the base.

John B.

On 2013-07-24, at 3:13 PM, Melvin Carvalho <[e-mail]> wrote:

> 
> 
> 
> On 24 July 2013 19:19, John Bradley <[e-mail]> wrote:
> Relative URI are not useful without a base to be relative to.  Unless we say how to determine the base of a relative URI they should not be allowed.
> 
> I an imagine several ways to determine a base but without specifying it applications will not behave consistently.   The JRD is not a HTML document.
> 
> Sure.  I wasnt trying to imply the JRD was the same as an HTML document.  Relative URIs, can be used in many contexts such as html/json/xml do indeed require a base.  I think the following text from RFC 3986, "5.1.  Establishing a Base URI" would be relevant here, and it's referred to by JSON LD too
> 
> http://tools.ietf.org/html/rfc3986#section-5.1
>  
> 
> John B.
> 
> On 2013-07-24, at 12:58 PM, Melvin Carvalho <[e-mail]> wrote:
> 
>> 
>> 
>> 
>> On 22 July 2013 16:10, Paul E. Jones <[e-mail]> wrote:
>> Melvin,
>> 
>>  
>> 
>> It’s easy enough to replace strings in URIs should data need to be ported.  I suggest we remove the word “absolute” where we currently have “absolute URI” and then introduce a new paragraph in the terminology section as follows:
>> 
>>  
>> 
>> The use of URIs throughout this document refers to URIs following the syntax specified in Section 3 of RFC 3986.  Relative URIs, having syntax following that of Section 4.2 or RFC 3986, are not used with WebFinger.
>> 
>>  
>> 
>> Is that clear?
>> 
>> 
>> Hi Paul, currently "Absolute URI" occurs only in section 4.4 (ie the description of JRD).  Certainly I think the 'absolute' should be removed in section 4.4.  
>> 
>> The RFC in essence describes two things, 
>> 
>> 1. the webfinger protocol
>> 2. the JRD definition
>> 
>> This restriction on relative URIs applies to (2), rather than, the whole document, but as it happens that turns out to be the same thing
>> 
>> Side note: in other situations I like relative URIs because they can be moved from one location to another in a non-breaking way.  Say I have /about/ /photos/ /blog/ at one location, I can move them to another without having to worry too much.  Therefore, if JRD was ever used in a context outside of webfinger, that would be a slight advantage.
>>  
>> 
>>  
>> 
>> Paul
>> 
>>  
>> 
>> From: Melvin Carvalho [mailto:[e-mail]] 
>> Sent: Monday, July 22, 2013 4:09 AM
>> To: Paul E. Jones
>> Cc: webfinger
>> 
>> 
>> Subject: Re: [webfinger] Absolute URI vs URI
>> 
>>  
>> 
>>  
>> 
>>  
>> 
>> On 22 July 2013 09:06, Paul E. Jones <[e-mail]> wrote:
>> 
>> In not suggesting we use or allow relative URIs. The immediate question would be "relative to what". I think the URIs should be absolute, but not following the syntax of "absolute-URI", but instead just following "URI".
>> 
>> Yes, that's a good question.  Generally relative URIs are relative to the document, but they need not be.  For example they could be relative by default to the [supposed] webfinger registry.  In JSON LD they have a @context variable which allows quite a bit of flexibility in this regard.  I'm unsure if relative URIs are considered a must in standards based serialization (I suspect not) but they have the advantage of slightly increased data portability.
>> 
>>  
>> 
>> Paul
>> 
>>  
>> 
>> From: Melvin Carvalho <[e-mail]>
>> Sent: Mon Jul 22 02:46:25 EDT 2013
>> 
>> 
>> To: "Paul E. Jones" <[e-mail]>
>> 
>> Cc: webfinger <[e-mail]>
>> 
>> 
>> Subject: Re: [webfinger] Absolute URI vs URI
>> 
>>  
>> 
>>  
>> 
>>  
>> 
>>  
>> 
>> On 22 July 2013 07:47, Paul E. Jones <[e-mail]> wrote:
>> 
>> Folks,
>> 
>> One of the requirements in the JRD spec is that certain URIs (e.g., those
>> identifying link relation types and properties) be "absolute URIs".  This
>> term has been a point of confusion, since RFC 3986 uses the term to have a
>> concrete meaning, namely this:
>> 
>>      absolute-URI  = scheme ":" hier-part [ "?" query ]
>> 
>> And the term is used, because that's the term used in the OASIS XRD
>> specification.  However, it was not clear to me whether that referred to the
>> above (which I assumed) or referred to URIs that are not relative URIs
>> (i.e., those lacking a scheme specified).
>> 
>> I had an exchange with Eran Hammer and Mark Nottingham.  I believe the
>> intent of that language was not to require the above constrained syntax,
>> but to require the standard URI syntax:
>> 
>>      URI = scheme ":" hier-part [ "?" query ] [ "#" fragment ]
>> 
>>  
>> 
>> There are many ways that people do this, see:
>> 
>> http://tantek.com/2011/238/b1/many-ways-slice-url-name-pieces
>> 
>>  
>> 
>> 
>> Given that understanding, I believe we should remove the word "absolute"
>> that appears in front of "URI" in the WebFinger spec.  Do others have an
>> opinion on this?
>> 
>>  
>> 
>> +1 on allowing relative URIs
>> 
>>  
>> 
>> 
>> 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
> 
>