[Openid-specs-ab] OpenID Connect Federation updated in preparation for third Implementer’s Draft review

Roland Hedberg roland at catalogix.se
Mon Jun 28 07:49:06 UTC 2021

Hi !

Thanks Torsten for your comments. I’ll start the answer with the design criteria we had:

- So far we’ve seen a small number of federation models. One-to-many (Google's 1 OP, many RPs or Amazon's 1 RP and many OPs) and small to fairly large multi lateral federations like EduGAIN (~4400 OPs and 3300 RPs). All of them based on centralised static registration. In order to allow multi lateral federations to grow in size we think that it’s necessary to move to decentralised dynamic registration (imaging if SIOP takes off). 

- One-to-many OIDC federations normally uses dynamic provider info discovery but not dynamic client registration. 
Which is not that surprising since classic OIDC client registration in essence is a leap of faith. There is no way you as an OP can 
verify that the client metadata is correct. We would like to make client registration more robust and allow OPs to verify
the correctness of the client metadata.

- Federation policies will change over time (like moving from SHA1 to SHA256) we would like to support that and to have built-in
support for policies to change dynamically. Also, having decentralised entity registration we need a way to enforce federation policies.

- OIDC and OAuth2 both have defined APIs for provider info discovery and client registration the federation specification should 
work equally well for both.

- The messages pushed around in this specification should not depend on TLS for their protection. 

- We should when possible use functionally already present in OIDC libraries (like key handling, signed JWT verifications, JWKS, ..)

- We should only touch the initial OIDC RP<->OP communication phases (provider info discovery and client registration).
Now, this changed during the work of the specification so there now is one use case where we touch the authorization request.

- An entity (OP or RP) should be able to belong to more the one federation.

> On 30 May 2021, at 18:38, Torsten Lodderstedt <torsten at lodderstedt.net> wrote:
> I think an overview describing and motivating the design concepts and principles would be helpful to readers. 
> I would also appreciate an explanation why the federation draft design is better suited for the envisioned use cases than X.509 certificates. Deployments need to be convinced to invest into a pretty new solution with a lot of runtime overhead (latency and availability implications!) while X.509 is used for the same/similar (?) applications in the wild. I’m pretty sure there a good arguments ;-). 

— Roland

Were it left to me to decide whether we should have a government without newspapers, or newspapers without a government, I should not hesitate a moment to prefer the latter. -Thomas Jefferson, third US president, architect, and author (1743-1826) 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openid.net/pipermail/openid-specs-ab/attachments/20210628/7e0bf632/attachment-0001.html>

More information about the Openid-specs-ab mailing list