[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: RE: [xri] Proposed XRD schema
Could you kindly give me an example of <Link> in the OpenID Provider's XRD?
=nat
差出人: Eran Hammer-Lahav [eran@hueniverse.com] 送信日時: 2009年2月18日 1:11 宛先: John Bradley CC: Dirk Balfanz; Brian Eaton; njones@ouno.com; xri@lists.oasis-open.org 件名: RE: [xri] Proposed XRD schema ResourceType replaces (Service-level) Type, and demoted to a hint instead of authoritative information. Rel is new and is authoritative. SubjectType is new and is authoritative.
ResourceType and SubjectType will use the same type values.
Rel can use the same type values but it has a different meaning (basically add a “my” before, as in “my openid provider”).
So on your blog page you can have:
<XRD> <Link> <Rel>http://specs.openid.net/relation/provider</Rel> <URI>…</URI> <ResourceType>http://specs.openid.net/auth/2.0/server</ResourceType> <ResourceType>some openid extensions</ResourceType>
And on the openid provider (pointed to by the URI above) you can have:
<XRD> <SubjectType>http://specs.openid.net/auth/2.0/server</ SubjectType > <SubjectType>some openid extensions</ SubjectType >
So the Rel just tells you it is a provider (not which version or other features). The rest in the first XRD are hints.
EHL
From: John Bradley [mailto:jbradley@mac.com]
Eran,
So I take it that you are thinking that Rel is the replacement for Type.
That an application searches for one or more Rel links in some application specific order.
Are you imagining that an App like openID would have two relationship types one for a service provider and one for the service itself?
I would like to go through some example resolution flows before coming to a conclusion on the schema.
=jbradley On 17-Feb-09, at 12:41 PM, Eran Hammer-Lahav wrote:
As I previously wrote (re: both Rel and ResourceType):
The idea was to preserve the functionality offered by Service/Type but to redefine it as a hint (in the same sense that MediaType is now a hint per link semantics). So that the same way to can give a parser a hint that the URI linked to will serve a JPEG, you can hint that the OpenID provider will support PAPE. As we know, there are cases where we will trust hints, allowing us to be more efficient.
EHL
From: John
Bradley [mailto:jbradley@mac.com]
Dirk,ers on the
I am also -1 to the idea of encoding parameters in the Subject type.
My example of <Rel>http://Subject tysome.URI.that.indicates.relationsip.is.self.or.sub.service.com</Rel>- Was just a example of one way of doing it if we don't want to make SubjectType extensible.
When the Meta Data is part of a Link statement it is clear that it is part of the Link however in the case of the final XRD (the one with the API URI as the subject. It is less clear where that goes.
If we are describing an optional part of an API then it is something that the API URI has a relationship with arguably and could be described in a Link. I don't know if it should. That depends on what sort of entity model we are creating for these related XRD.
I am still not clear on the difference between ResourceType and Rel. I suspect one may need to go, because I suspect that they are probably the same thing in Eran's mind.
If we are going to keep both, I think one needs to be a hint about the type of XRD that the URI is pointing at, if the URI is discoverable at all.
I think there are four of these sort of relationships: 1 The subject is a individual (pointing at someone's blog XRD) 2 The subject is a Service Provider (The XRD will contain multiple Rel types) 3 The subject is a Service (OpenID as an example) 4 The Subject is a API endpoint and this is the end of discovery
Perhaps there is a relationship between this and the Subject-Type of the target XRD?
I am unclear on what the XRD of a Service provider would have as it's Subject type? Are people thinking that it would have multiple SubjectTypes one for each service it provides. Then have Links to the individual services like OpenID 2.0.
I understand that a large part of discovery will be application specific. More than in XRI 2.0 at least. However I think we do need to define a common entity model so that applications can have a common view of what the XRD are describing.
=jbradley
On 17-Feb-09, at 4:14 AM, Dirk Balfanz wrote:
On Sun, Feb 15, 2009 at 11:08 PM, Brian Eaton <beaton@google.com> wrote: On Fri, Feb 13, 2009 at 7:28 PM, <njones@ouno.com> wrote: Libraries would wander through the XML looking for the appropriate
I agree that works inside Link elements, where the URI is tucked away neatly inside the Rel element, and thus separated from other potential metadata. But inside the SubjectType, it doesn't seem to work as nicely. It would look something like this:
<SubjectType> <height>500px</height> <width>450px</width> </SubjectType>
Is that the same as this:
<SubjectType> <height>500px</height> <width>450px</width> th/2.0/ux/popup </SubjectType>
?
That doesn't seem right...
+1 on using XML parsers, though, and -1 on funny business with query parameters.
Dunno how I feel about this <Rel>http://some.URI.that.indicates.relationsip.is.self.or.sub.service.com</Rel>-stuff, though. What's the purpose of that? That's what the XRD top-level section is for, right?
Dirk.
|
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]