generally speaking, Ive been following how microsoft cloud and directory is rolling out its first incarnation of openid connect for commodity developers to use. And its very simple (with openid connect being merely one of several protocols that come together as a websso/webapi family). Im hardly smart, and I CAN get it (in that form).
There, openid connect is presented as little more than oauth2 with an id-token (assuming that the access token might be encrypted one day, denying its repurposing as an id token with claims for UIs). The openid connect world shares some control plane material with the non openid-connect profiles of the family (which includes ws-fedp, and non-openid connect profiles of oauth2). This uses old SAML-proof token features like scopes and audiences and claim names standardization. These, whether in the openid connect compatability modes or not, can be orchestrated so the control and management does the technically-govern multi-server interactions - as a distributed/partitioned TCB (NCSC Red Book model for secure networking).
If it helps, I for one cannot relate much of what I read in the standard to what I see being practiced. It all feels somewhat like the PKIX spec, that took 100 pages to define what certs do (which only required 15 ISO pages in 1988, and 30 pages in 2000). The number of folks using the advanced profiles of PKIX must be about 0.000001% percent of the internet, with the 80% using the bottom set of features that have not really changed much in 20 years.
Now, you might come back at me and say: this is NOT for general internet apps building. This is REALLY all for telcos, and the world like Neustar run in which huge numbers of management servers control the security policy that govern interworking, roaming of users in such as GSM. And thus THAT is where openid foundation is REALLY focused (and not on the likes of me, in the traditional web world).
have I captured the right way to think - about where emphasis of all the work is targeted?
This list is for general information about openID.
That group/list has no IPR restrictions.
You must sign a non assert IPR agreement to contribute to the specification, as all the other WG participants have.
The version of Dynamic Client registration being developed by the OAuth WG in the IETF based on the Connect specification can be discussed at:
To your question Registration's application type is not the same as RFC 6749 Sec 2.1 Client Types.
It is dynamic registration itself that allows a native app to be confidential so that section of RFC 6749 may be changed after the IETF approves dynamic client registration.
That section also states
A client may be implemented as a distributed set of components, each
with a different client type and security context (e.g., a
distributed client with both a confidential server-based component
and a public browser-based component).
Connect supports this with the hybrid response types.
A Web application with a single client_id may contain both a public component and a confidential component.
The element response_type determines the flows the client can preform.
The default for that is the client can only preform the code flow and must use a client credential to authenticate itself to the token endpoint.
The response types are a finer grained and imply public and private for the hybrid components.
What is not possible without extending dynamic client registration are public clients that use a code grant_type.
One of the main reasons for dynamic client registration is to eliminate the need for an AS to support public clients using the code grant_type.
In any event we can continue this on one of the other lists.
Thank you for your comment.
If both native clients and web clients can be public or confidential,
application_type which is described in Client Metadata section of
OpenID Connect Dynamic Client Registration 1.0 is DIFFERENT from the
client type described in RFC 6749.
Considering the following two points:
(1) RFC 6749 (OAuth 2.0) requires a client type to be registered.
(RFC 6749, 2. Client Registration)
(2) OpenID Connect Relying Party is a kind of OAuth 2.0 client app.
(OpenID Connect Core 1.0, 1.2. Terminology)
there should be a client metadata entry to specify a client type (e.g.
client_type), and it should be treated separately from application_type.
Information about Client Type (not Application Type) is really needed
to implement OAuth 2.0 authorization server correctly. Especially it
is needed to determine (1) whether client authentication is required
at the token endpoint (3.2.1. Client Authentication) and (2) whether
registration of redirect URIs can be omitted (18.104.22.168. Registration
Without a client metadata entry to specify a client type, the default
value of client type cannot help but be left to implementations of
OpenID Connect Dynamic Client Registration 1.0.
IMHO, if such an entry (e.g. client_type) should be added, 'public'
would be appropriate as the default value because 'client_secret' in
Client Registration Response is OPTIONAL.