[Openid-specs-ab] Spec call notes 7-Jun-12
Dingwell, Robert A.
bobd at mitre.org
Fri Jun 8 15:43:18 UTC 2012
I would also like to echo Justin sentiments on this issue, this seems like a step backwards and does not allow for a easy path forward to integrate OIDC idP's into systems that already rely on a username type claim.
While I work at MITRE I am actually on a different project than that of Justin. My project is currently using/profiling the use of OpenID Connect as a means to provide distributed identity and authentication across the Healthcare Domain for the Office of the National Coordinator for Health IT (ONC) under the Nationwide Health Information Network (NwHIN ). As part of our profile for OIDC there will be a required claim of username for identifying users across multiple local health domains. This lets us integrate and use existing identifiers already declared in other NwHIN projects based on username at healthdomain semantics. For us a username is an absolutely must have claim and our profile will require it's use. It's that or we are forced to use the user_id field for such requirements because nobody wants to use lkasdf66778asdf6asd55ksdksdf7 at example.com as an identifier to remember.
Additionally I would like to add that there is already implicitly declared the need for a username in the discovery protocol and specification for that has examples showing it's use. Whether or not it has been called that or not that is essentially what it is . What else would the joe in joe at example.com be. The discover protocol states that while using an email address for discovery the principle is the full email address and that idP is the host section of that address. So what does that make "joe".
You cannot just pull this out of the email address for a user. The email addresses are not scoped to the domain of the idP and multiple users could have the same username portion of an email address just at different hosts, joe at hotmail.com vs joe @gmail.com vs joe at example.com. I cannot see how including a username claim adds more confusion when not having one makes parts of the specs already written vastly more confusing without it.
Also on a side tangent though somewhat related by way of the discovery specs. how do you verify a person is who they have claimed to be during discovery? If I put into a form an email address identifier that then on the backend uses discovery to find the idP and redirect me to authenticate how does the RP know that is who I actually authenticated as at the idP? There appears to be a step missing here that would allow the RP to know who actually authenticated. Otherwise I could put in joe at example.com but authenticate as joes_evil_twin at example.com and there does not appear to be away to verify that I actually authenticated as joe and not his evil twin. If there was a username claim that could resolve some of this, but is should be baked into the authorization request somehow.
From: Justin Richer <jricher at mitre.org<mailto:jricher at mitre.org>>
Date: Friday, June 8, 2012 9:57 AM
To: Mike Jones <Michael.Jones at microsoft.com<mailto:Michael.Jones at microsoft.com>>
Cc: "openid-specs-ab at lists.openid.net<mailto:openid-specs-ab at lists.openid.net>" <openid-specs-ab at lists.openid.net<mailto:openid-specs-ab at lists.openid.net>>
Subject: Re: [Openid-specs-ab] Spec call notes 7-Jun-12
I think that the resolution to issue #584 (Username claim) is a mistake. We're regressing from capability available and used in OpenID 2.0 via SREG and AX. We're ignoring what an existing IdP, Facebook, already does.
Telling clients to parse out the email address and pull in a name there is a ludicrous suggestion. Why don't we also drop given_name and family_name since the RP could just try and parse those out of "name"? It's absurd to make something this simple so complicated.
I'm sorry that I can't make the telephone calls for the WG (for several personal reasons), especially since that seems to be where decisions like this get made. I would be happy to grab some other time to discuss this with people if more discussion is warranted. I'll even gladly submit a pull request of the spec changes for the issue. It will be about five lines tops, including all the XML.
I can also say that our IdP implementation *will* have a username claim, and our in-house RP's *will* rely on it. I imagine that it other IdPs in the wild will as well. We might as well just decide here and now to standardize on it.
And yes, I do think this is important.
On 06/07/2012 08:07 PM, Mike Jones wrote:
Spec call notes 7-Jun-12
Edits and Release
We agreed that we need to clone the OC3 interop to create OC4 before adding
We will need Pam's buy-in and time to do this
We agreed that it is easier to reach consensus when you can hear one another's voices than via e-mail
We will try to resolve this issue soon that way
Nat will send out a Doodle poll for special call time to discuss this issue
We will schedule the call at a time that makes it convenient for Europeans
We discussed that at John's proposal to have 4 new scope elements makes the scope elements not stateful
profile_idt email_idt address_idt phone_idt
Developers are rejecting claims_in_id_token because it made the 4 scope definitions stateful
Nat began transcribing the Session Management notes
He sent out initial notes that he needs feedback on
He noted that this spec will be quite different than the others, as it contains local APIs
Edits and Release:
The self-issued edits may be ready for review tomorrow - issue #566
After that, we'll do a release with that functionality only
We discussed some of the open issues. There were no new ones.
#576: Discovery - Monitor IETF discovery spec decisions
We're seeing no movement on WebFinger
We will need to decide soon what to do
Stay with SWD
Try to profile a subset of WebFinger
#257: Acknowledgements and other sections need review
Nat complied a proposed list of acknowledgements
#539 Messages - 0. Add scope for offline access
AOL's approach requires stateful session state, Google's is stateless
This issue needs an owner and a concrete proposal
Assigned to George, who will work with Breno
#584 Messages - Username claim
We discussed whether syntax restrictions should be imposed and if so, what
No one on the call was particularly enamored with the claim
Relying parties could just use the part of the e-mail address before the @ to achieve approximately the same thing
The consensus on the call was that this field is adding more confusion than value
Resolved as WontFix
#593 Standard: redirect_uri registration & matching
George argued that requiring matching query parameters is overkill
We only got through the issues up to #593
Openid-specs-ab mailing list
Openid-specs-ab at lists.openid.net<mailto:Openid-specs-ab at lists.openid.net>http://lists.openid.net/mailman/listinfo/openid-specs-ab
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openid-specs-ab