Re: [xri] Version Control Commit by blade

[Will Norris <will@willnorris.com>]

inline

On Aug 20, 2009, at 11:51 AM, Eran Hammer-Lahav wrote:

>
>
>> -----Original Message-----
>> From: Will Norris [mailto:will@willnorris.com]
>> Sent: Thursday, August 20, 2009 11:47 AM
>> To: xri@lists.oasis-open.org
>> Subject: Re: [xri] Version Control Commit by blade
>>
>> not sure which commit these were actually on, but i have two
>> comments.  You modified Section 2.4.1 to read:
>>
>>> The one distinction is that link relationships described by the
>>> <Link> element are between the resource described by the XRD (the
>>> context resource) and the linked resources (the target resources),
>>> and not between the XRD document itself and the linked resource.
>>
>> The addition here are the two parenthetical statements, "context
>> resource" and "target resource".  Both of these are new terms that  
>> are
>> not present anywhere else in the spec.  Some time ago we did away  
>> with
>> the term "target resource" in favor of "linked resource".  We also
>> consistently use the phrase "resource described by the XRD"  
>> throughout
>> the spec.  It is a mouthful, and I'm not terribly fond of it, but it
>> is accurate and the best we could come up with.  I don't disagree  
>> that
>> this sentence can be a little confusing with all the "resources"  
>> being
>> thrown about, but I'm concerned about throwing in new terminology  
>> here
>> that isn't at all consistent with the rest of the spec.
>
> It is how the HTTP-Link spec defines typed links... It might be a  
> good idea to borrow a bit of that language here to make it easy for  
> those familiar with the link framework.

That may be.  I just want to make sure we're consistent.


>> You added the following paragraph to the end of Section 2.5.1:
>>
>>> New relation types between resources must follow the extensibility
>>> and registration requirements defined in [HTTP Link Header].
>>
>> While I don't disagree with this, I'm curious if it's necessary.  The
>> whole reason link-header had to create a registry is because it is
>> using tokens for registered values, and only using URIs for extended
>> values.  XRD is specifically using URIs only, so we don't really care
>> about the token values.  The universal rule of URIs is that you don't
>> make up new URIs in namespaces you don't own.  That would imply that
>> you can't make up a new relationship in the
>> "http://www.iana.org/assignment/relation
>> " unless you've properly registered it.  Now, I do like having some
>> kind of reference because it instructs publishers how to format those
>> token values as URIs, but the wording of the above paragraph seems  
>> out
>> of place, given how XRD uses these values.
>
> We are not using just URIs. Note we don't have any requirement for  
> absolute URIs for <Rel>. A 'describedby' rel type is perfectly valid.

Now that's really interesting... I hadn't considered "describedby" to  
be a valid value for <Rel>.  I'm actually not too fond of that, and am  
now wondering if we *should* in fact mandate an absolute URI here.


> I think it is important to point out to developers that if they want  
> to create new relation types, they should consult the link spec for  
> directions and guidelines on when to mint a short name and when a  
> URI extension.


Fair enough.  Perhaps move this paragraph to the extensibility section  
of the spec then?