[Openid-specs-ab] First Release Candidates for final OpenID Connect specifications
torsten at lodderstedt.net
Fri Oct 18 20:15:43 UTC 2013
@Mike: first of all let me thank you for taking the burden and rework
the whole document structure. I think the new structure is a major leap
forward for OpenID Connect.
I focused my review on the new core specification. Please find my
I won't attend IIW but will attend IETF-88. So, if needed, we can talk
through my comments there.
I would suggest to a reference to RFC 6749 in the first sentence. It
probably also makes sense to explicitly point out that the reader is
expected to be familiar with RFC 6749 and RFC 6750 as well as other IETF
I-Ds (notably JOSE, JWT and JWT Assertion Profile).
The flow description is a good starting point for readers. I would
suggest to add the following information in this section:
- OpenID Connect authentication is basically an extension to the
standard OAuth authorization process. This extension is defined for most
OAuth grant types.
- Clients wishing to acquire identity information indicate this by
sending the scope value "openid" as part of the authorization request
parameters. (There are much more parameters used to control the process
but this is the "main switch".)
- Such a client is also called relying party (RP). An authorization
server also supporting OpenID Connect is called OpenID Provider (OP).
Adding this information will help the reader to understand the way
connect utilizes/integrates into OAuth.
I would also suggest to move the definition (syntax and contents) of the
ID Token here and make it section 1.4 because this is THE core concepts
used throughout the specification. It's introduction in section 18.104.22.168
is to late (in my opinion) because it is cited roughly 20 times in
"The Authorization Code Flow is suitable for Clients that can securely
maintain a Client Secret between themselves and the Authorization Server
..." - this is confusing since public clients can use the code as well.
The key benefits of this grant type in my opinion are:
- AS _can_ authenticate clients
- AS _can_ return refresh tokens
- simplest way for web application backends to acquire tokens
That's why is best suited for web applications and native apps.
"The Authorization Code Flow is appropriate for web applications and
native apps as it allows to authenticate clients and obtain refresh
tokens whereas the implicit flow does not support these features."
Or just remove the assessment of OAuth grant types and leave it to the
implenentors to carry out their assessment.
2.1. Authentication using the Authorization Code Flow
OLD: "This provides the benefit of not exposing the Access Token to the
Resource Owner ..."
The same indeed holds for the ID Token, which is more important from a
NEW: "This provides the benefit of not exposing the Access Token and the
ID Token to the Resource Owner ..."
NEW (alternative): "This provides the benefit of not exposing any Token
to the Resource Owner ..."
2.1.1. Authorization Code Flow Steps
OLD: "8. Client validates the tokens and retrieves the End-User's
I assume the client is supposed to validate the ID token, only?
NEW: "8. Client validates the ID token and retrieves the End-User's
22.214.171.124. Authorization Request
"When the Client wishes to access a Protected Resource and the End-User
Authorization has not yet been obtained, the Client prepares an
Authorization Request to the Authorization Endpoint" - Why is this
relevent in this context? I suggest to remove this sentence.
"An Authorization Request is a message sent from an RP to the OP's
Authorization Endpoint. It is an extended OAuth 2.0 [RFC6749]
Authorization Request. Section 4.1.1 and 4.2.1 of OAuth 2.0 [RFC6749]
define the OAuth 2.0 Authorization Request parameters." - Why
Authorization Request? Shouldn't this be called an "Authentication Request"?
"Communication with the Authorization Endpoint MUST utilize TLS. See
Section 15.17 for more information on using TLS.
Authorization Servers MUST support the use of the HTTP GET and POST
methods defined in RFC 2616 [RFC2616] at the Authorization
Endpoint.Clients MAY use the HTTP GET or POST methods to send the
Authorization Request to the Authorization Server. If using the HTTP GET
method, the request parameters are serialized using URI Query String
Serialization, perSection 12.1. If using the HTTP POST method, the
request parameters are serialized using Form Serialization, per Section
Seems to be standard OAuth stuff, I suggest to remove it.
- redirect_uri Parameter
"This URI MUST exactly match one of the redirect_uris registered for the
Client" - Why is the redirect_uri validation more restrictive than
"When using this flow, the redirection URI MAY use the http scheme,
provided that the Client Type is confidential, as defined in Section 2.1
of OAuth 2.0; otherwise, it MUST use the https scheme" - This,
surprisingly enough, is relaxed in comparison to
RFC 6749 states: "Authorization codes operate as plaintext bearer
credentials, used to verify that the resource owner who granted
authorization at the authorization server is the same resource owner
returning to the client to complete the process. Therefore, if the
client relies on the authorization code for its own resource owner
authentication, the client redirection endpoint MUST require the use of
Why is Connect, in this particular case, less restrictive than OAuth?
- nonce Parameter
"One method to achieve this is to store a random value as a signed
session cookie, and pass the value in the nonce parameter. In that case,
the nonce in the returned ID Token can be compared to the signed session
cookie to detect ID Token replay by third parties." - I would recommend
to move this text into an "implementation note" section
id_token_hint Parameter - "Previously issued ID Token passed to the
Authorization Server .." issued by the AS being requested? or any AS? I
assume by the same AS
NEW: "ID Token previously issued by this Authorization server to the
"... it SHOULD return a login_required error." - Does this mean the OP
shall try to authenticate the user account identified by the ID token
and refuses authentication otherwise? This sounds more like a
requirement than a hint.
"When possible, an id_token_hint SHOULD be present when prompt=none is
used and an invalid_request error MAY be returned if it is not; however,
the server SHOULD respond successfully when possible, even if it is not
present." - Why is the login hint recommended for this prompt value?
checkid_immediate in OpenID 2.0 worked very well w/o a hint?
126.96.36.199. Authorization Request Validation
"3. If the sub (subject) Claim is requested with a specific value for
the ID Token ...." The meaning of the text is unclear to me. How is a
specific sub value requested? by the login_hint or the id_token_hint?
"As specified in OAuth 2.0 [RFC6749], Authorization Servers SHOULD
ignore unrecognized request parameters.
If the Authorization Server encounters any error, it MUST return an
Standard OAuth stuff, I recommend to remove it.
188.8.131.52. Authorization Server Obtains End-User Consent/Authorization
"When permitted by the request parameters used, this MAY be done through
an interactive dialogue with the End-User ..." - What if the parameters
do not allow for an interactive dialogue, e.g. prompt==none? I assume an
error response with return code consent_required or interaction_required
is appropriate. I would prefer interaction_required because to RP does
not need to know more.
184.108.40.206. Authorization Successful Response
This is a vanilla OAuth 2.0 response, right? I would suggest to just say so.
BTW: This piece of text is not applicable to the code grant type: "This
specification only describes OAuth 2.0 Bearer Token Usage [RFC6750]. The
OAuth 2.0 response parameter token_type MUST be set to Bearer unless
another Token Type has been negotiated with the Client."
2.1.3. Tokens Endpoint
"Clients MUST use the HTTP POST method to make requests to the Token
Endpoint. Request parameters are added using Form Serialization, per
Section 12.2. The Token Endpoint MUST support the use of the HTTP POST
method defined in RFC 2616 [RFC2616] at the Token Endpoint.
Communication with the Token Endpoint MUST utilize TLS. See Section
15.17 for more information on using TLS.
All Token Endpoint responses that contain tokens, secrets, or other
sensitive information MUST include the following HTTP response header
fields and values: ..."
This seems to be standard OAuth stuff. I recommend to remove it.
220.127.116.11. Token Request
"To obtain an ID Token, Access Token, or Refresh Token, the Client MUST
authenticate to the Token Endpoint using the authentication method
registered for its client_id, as described in Section 8 ..." - At this
point the reader is not familiar with the different authentication
methods supported by an OpenID OP. I therefore suggest to move the
client authentication section before the authentication section (e.g.
make it a section 1.5).
18.104.22.168. Token Request Validation
The whole sections seems to re-phrase standard OAuth stuff. I recommend
to remove it.
22.214.171.124. Token Successful Response
"Servers SHOULD support OAuth 2.0 Bearer Token Usage [RFC6750] for
interoperability" - I think this topic is better covered in the MTI section.
"the following parameters MUST be included in the response if the
grant_type value is authorization_code and the Authorization Request
scope parameter contains openid:" - Seems to be redundant since this
whole section is about exactly this use case. I recommend to remove this
text. Same holds true for the following text
"An id_token MUST be returned when the grant_type value is
authorization_code and MAY be returned when other grant types are used."
126.96.36.199. ID Token
That's the key elements of OpenID Connect! As already stated, I
recommend to move the (full) description of its content and syntax to
section 1 (section 1.4). I think this will facilitate readability. The
validation rules should stay in the sections of the respective grant types.
188.8.131.52. ID Token Validation
"1. If the Client has provided an id_token_encrypted_response_alg
parameter during Registration, decrypt the ID Token using the key pair
specified during Registration." - text depends on dynamic registration
and should therefore be generalized.
NEW: "1. If the ID Token is encrypted, the Client first decrypts it
using the key agreed upon with the authorization server."
"5. If the id_token is received via direct communication between the
Client and the Token Endpoint, the TLS server validation MAY be used to
validate the issuer in place of checking the token signature. The Client
MUST validate the signature of all other ID Tokens according to JWS
[JWS] using the algorithm specified in the alg parameter of the JWT
header." - text seems rather generic. As this is about the code flow,
the ID token is received via direct communication, so the text can be
NEW: "5. Since the ID token is received via direct communication, the
TLS server validation MUST be used to validate the issuer in place of
checking the token signature."
Steps 6-8 can be removed.
184.108.40.206. Access Token Validation
I would recommend to add the text from section 220.127.116.11. because this is
the first point where the concept is used.
a_hash validation is mentioned. What's about c_hash validation?
2.2. Authentication using the Implicit Flow
"...which may expose them to the Resource Owner and other applications
that have access to the Resource Owner's User-Agent." - I recommend to
add: "In contrast to the authorization code flow, this requires
additional security mechanisms (described below) to detect falsified ID
18.104.22.168. Authorization Request
see comments regarding redirect_uri and nonce for section 22.214.171.124
126.96.36.199 - 188.8.131.52 - I recommend to state in section 2.2.2 that those
steps are performed in the same manner as for the code flow and to
remove these section?
184.108.40.206. Authorization Successful Response
access_token Parameter - "Access Token for the UserInfo Endpoint." - In
contrast to section 220.127.116.11, this section also described standard OAuth
response parameters. I don't think this is needed. Moreover, the term
"User Info" is used before it is really introduced. In my opinion,
authentication should not talk about user info. The access token
returned as part of the authentication result might suited for
interactions with any protected resource, including user info.
18.104.22.168. Redirect URI Fragment Handling
This section needs a bit more of context and description. While the
introduction of 2.2 states: "The Implicit Flow is mainly used by Clients
implemented in a browser using a scripting language", this section
suddenly _requires_ the client to send data to a web server ("The Client
MUST provide ...").
First of all, I don't understand the MUST.
Second, a description is needed of the different patterns, scripting
clients vs. scripted frontends of server-based web-application (here the
implicit grant offers better performance and scalability at the cost of
a more complex client implementation. Therefore, implementation advices
I also think this section should be part of an "implementation note"
22.214.171.124. ID Token Validation
Text around signature validation should be moved from 126.96.36.199 to this
section as it is required for the implicit grant (in contrast to code).
2.3. Authentication using the Hybrid Flow
188.8.131.52.-184.108.40.206. can be removed
220.127.116.11. Authorization Successful Response
Again, I would recommend to focus on additional response parameters, as
in section 18.104.22.168
22.214.171.124. Redirect URI Fragment Handling
see comment at section 126.96.36.199
I think 188.8.131.52.-184.108.40.206 can be removed - Instead it should be stated
that for the hybrid flow client and AS must conform to ALL requirements
for code and implicit.
3. Initiating Login from a Third Party
I assume this is mainly intended to support OP initiated logins? I don't
think it deserves a top-level section. I would recommend to make it part
of the Authentication section.
"The Client MAY optionally register [OpenID.Registration] an
initiate_login_uri that can be used by the Authorization Server or
another party to initiate a login for an End-User at the Client." I
assume this feature shall also be supported by OPs w/o dynamic
registration? I therefore suggest to move this text to the dynamic
registration spec. Instead one could state: "The approach utilized by
the 3rd party or the OP to determine the client's respective URL is
out-of-scope of this document."
Generally, I don't think any meta data element registered via dynamic
registration or discovered via OpenID Discovery should be
specified/mentioned in the core spec. I think the core function must
work w/o both functions in place.
"This section specifies how the Client can obtain Claims about the
End-User ..." claims about the authentication process are supported as well.
NEW: "This section specifies how the Client can obtain Claims about the
End-User and the authentication process ...
4.1. Requesting Claims using Scope Values
This is an extension to the authentication part, it should be specified
that way. For example, there is no need to specify the use of the scope
value "openid" again. IMHO it suffices to say that clients may request
access to user data by adding more scope values in conjunction with
4.2. Standard Claims
I think this section should be the first section as it describes
standard claims on a conceptual level and which ways exist for a client
to retrieve them.
4.3. UserInfo Endpoint
I think this section should be merged with Section 4.1 as the scope
values defined there control access to this resource, only.
4.4. Requesting Claims Locales with the "claims_locales" Request Parameter
I would suggest to move this either before 4.1. or after 4.5. as it
seems to be orthogonal to the functions described there.
4.6. Claim Types
This seems to be out of order because after a description of how a
client may allocate claims to ID token and user info, this section again
exclusively talks about UserInfo. Is it really the intention to support
aggregated and distributed claims at the User Info endpoint, only? If so
I recommend to move this section before 4.4. Otherwise, please state
that such claims can be provided in the ID Token as well.
How is a aggregated or distributed claim requested by a client?
5. Passing Request Parameters as JWTs
I would suggest to move this topic into another document. The features
described here allow implementors to achieve higher security levels and
may reduce the request size but I consider them beyond the scope of a
6. Self-Issued OpenID Provider
How mature is the concept? How many implementations exist? Is this
really part of a core specification or rather an extension for
mobile/personal devices? I would opt for making this section a separate
6.2. Self-Issued OpenID Provider Registration
"When using a Self-Issued OP, the Client is deemed to have registered
with the OP and obtained following Client Registration Response." - Does
this mean dynamic registration is required for self-issued ID providing?
7. Subject Identifier Types
This section is completely about discovery and registration - it should
go there, consequently.
8. Client Authentication
This section provides fundamental information about the client
authentication methods supported/introduced by OpenID Connect. I would
suggest to move it into section 1 (e.g. section 1.5).
"During Client Registration, the RP (Client) MAY register an
authentication method. " I assume all client authentication methods
shall work even if the OP does not support dynamic registration.
Consequently, this text must be replaced by a general description of
what methods are supported and how the OP is supposed to determine the
right method for a particular client.
OLD: "During Client Registration, the RP (Client) MAY register an
authentication method. If no method is registered, the default method of
client_secret_basic MUST be used."
NEW: "OpenID connect supports the authentication methods listed below.
The authorization server determines the authentication method to be used
in a particular authorization transaction based on the client_id. The
way client and authorization server negotiate the authentication method
is out of scope of this specification."
9. Signatures and Encryption
Most of this section talks about discovery and dynamic registration - it
should go there, consequently.
I think the core spec needs MTI algorithms and clear definition when
signatures are required. In my opinion, there are two use cases:
- Login via implicit grant
- azp - Login at another AS via ID token
In both cases, Signing ID Tokens with RSA SHA-256 could be good
baseline. This could be documented in the ID Tokens section or the MTI
9.3.1. Rotation of Asymmetric Signing Keys
Isn't this an implementation note?
10. Offline Access
Given the description (" that grants access to the End-User's UserInfo
Endpoint ..."), I would say this text can go to the User Info section.
11. Using Refresh Tokens
I think this should go to the authentication section (2.4?), as it
describes usage of the refresh token grant type in the Connect context.
11.2. Refresh Successful Response
"If an ID Token is returned as a result of a token refresh request ..."
- Can't we specify the conditions under which an ID token is returned?
Otherwise, an interoperable client does not know what to expect or how
to control the outcome of this request. For the standard authentication
flow, presence of the scope value "openid" is the trigger. I would
suggest to use the same logic for the refresh token grant type.
14. Implementation Considerations
I would suggest to move MTI requirements to another document as it
refers to nearly every document of the OpenID Connect suite. For the
core document, I would suggest to (at most) specify the MTI requirements
for closed systems, only:
- code flow extension
14.1. Mandatory to Implement Features for All OpenID Providers
"Signing ID Tokens with RSA SHA-256" - I think, in Berlin, we agreed to
not require signatures for simple, static setups using authz code. We
agreed to move this to the MTI section for dynamic OPs.
14.5. Compatibility Notes
Isn't such information better covered on the working group page. Such
information typically change more often than the specification itself.
14.6. Related Specifications and Implementer's Guides
Am 15.10.2013 17:18, schrieb Mike Jones:
> I'm pleased to announce that the first release candidate versions for
> final OpenID Connect specifications have been published. The complete
> set of specifications has been updated to resolve all issues that had
> been filed against the specs being finished.
> Please review these this week, in time for the in-person working group
> meeting on Monday <http://openid-wg-oct-2013.eventbrite.com/>.
> Besides publishing the specs in the usual formats, I've also created a
> Word version of the core spec with tracked changes turned on to
> facilitate people marking it up with specific proposed text changes.
> If you're in the working group, please download it
> <http://self-issued.info/docs/openid-connect-core-1_0-14.docx> and
> make any corrections or changes you'd like to propose for the final
> The release candidate spec versions are:
> Also, two implementer's guides are also available to serve as
> self-contained references for implementers of basic Web-based Relying
> Thanks to Nat for the early feedback. The structure of Core has been
> changed somewhat since -13 to adopt some of his suggestions.
> -- Mike
> P.S. I also posted about this at http://self-issued.info/?p=1140
> <http://self-issued.info/?p=1140> and as @selfissued.
> Openid-specs-ab mailing list
> Openid-specs-ab at lists.openid.net
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openid-specs-ab