[Openid-specs-ab] client_credentials grant_type
ve7jtb at ve7jtb.com
Tue Sep 18 14:34:11 UTC 2012
It might be that SCIM is the right answer in many cases for something that is more directory access.
The advantage Connect might have is the claims request architecture, signed claims, and claims by reference.
SCIM is also clearly the thing you want if you need write access.
The use case I was comparing to was SAML Attribute Query, and that is a bit more like Connect than SCIM.
I am still quite interested in how SCIM and Connect can work together. Having a optional SCIM endpoint in Connect for on demand provisioning would be useful.
On 2012-09-17, at 2:27 PM, Dale Olds <olds at vmware.com> wrote:
> If I understand the original use case, a client wants to get claims about a specific subject, but does not have the subject's access token with scope openid, etc. Therefore, it is natural to use a client credentials grant, but the client just needs a way to specify the subject to the /userinfo endpoint.
> We see this pattern quite frequently, but we did not extend the /userinfo for per-subject client use. It would be hard to predict all the ways in which a client application may want to get information about users and potentially modify the user info. Some applications need a small bit of info about the user, some start off that way and then turn into are rather substantial user account management apps. So the usual pattern for our system is to use client credentials grant and the scim protocol for info regarding other subjects (uses), and Connect for authentication/userinfo stuff.
> I'd like to support Connect as much possible, but I guess the question is purpose of the protocol. If a client wants to get information about a user for which it does not have a token (i.e. it is not operating on behalf of that user), then I would think it's making a directory service request.
> On 09/17/2012 06:49 AM, John Bradley wrote:
>> In some ways it is standard OAuth , but standard OAuth doesn't specify how identify the subject , request claims or retrieve the result.
>> You could do all of that from scratch, but it may be useful to use what already exists in Connect. That could be done as part of Connect or a separate spec that profiles Connect.
>> John B.
>> On 2012-09-17, at 2:30 AM, Torsten Lodderstedt <torsten at lodderstedt.net> wrote:
>>> Hi John,
>>> ressource owner password credential grant makes definitely sense in my opinion. I'm not sure for client credentials, esp. w/o id_token, as this boils down to standard OAuth to get access to the user info endpoint.
>>> John Bradley <ve7jtb at ve7jtb.com> schrieb:
>>> Last week I had several conversations with FICAM people around OAuth and Connect.
>>> One thing that they do and is also not uncommon in enterprises is permission access based on client credentials.
>>> Think SAML Attribute query.
>>> We do have that in OAuth 2.0.
>>> One thing we don't say in Connect is how to support that grant_type.
>>> It seems fairly strait forward that you would have a scope of openid and any other user_info related scopes, that nonce and state are not required.
>>> Returning a id_token probably doesn't make sense.
>>> To specify the user who is the subject we already have a way of passing the required user_id in the request object.
>>> I can see this being useful to compliment or replace a SAML/SOAP flow.
>>> We don't specifically talk about this or the Resource owner Password credentials Grant.
>>> long as we don't do something in the core specs to preclude them we could put them in a separate profile as they are sort of special case.
>>> John B.
>>> Openid-specs-ab mailing list
>>> Openid-specs-ab at lists.openid.net
>> Openid-specs-ab mailing list
>> Openid-specs-ab at lists.openid.net
> Openid-specs-ab mailing list
> Openid-specs-ab at lists.openid.net
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 4937 bytes
Desc: not available
More information about the Openid-specs-ab