Re: [OAUTH-WG] Need for Extending OAuth with AuthN (was Re: Fwd: New Version Notification for draft-hunt-oauth-v2-user-a4c-00.txt)

[Torsten Lodderstedt <[e-mail]>]

 

Hi Nat, 

I think your are going in the right direction. Here are my
comments: 

- Authentication and attribute providing can be treated
separately. I therefore would recommend you move the claim stuff into a
separate specification, which includes standard claims, respective scope
values, user info endpoint and the claims parameter. This would leave
the base specification a small but reasonable extension to OAuth to
handle authentication properly. Everyone interested in claims can
implemented the complementary claims spec. Service provider wishing to
use other interfaces for claim providing can do that. I would not care.


- Use normativ language only to ensure proper implementation of the
protocol, e.g. with respect to security, not for MTI requirements. 

As
an example: "id tokens must be signed when sent via the user agent, they
may be signed when issued at the token endpoint" 

- I would recommend
to handle MTI in a separate document. Do not clutter the base spec with
MTI stuff, such as 

"OpenID Providers that are not Self-Issued OPs MUST
support this "response_type"." 

First, you are forward referencing to
concepts introduced in other documents. Moreover, there are so many
different scenarios I can think of where Connect could be used. They
have rather different requirements, so having a document with different
profile definitions is better suited (IMHO). A good starting point is
section 13, where the definition of open and closed systems and
respective MTIs is given. 

- Omit duplication of OAuth text. For
example, there is another description of the state parameter or there is
this statement: "In OpenID Connect, this communication MUST be over
TLS." As far as I remember, the same holds true for plain OAuth. 

-
3.2. Server Processings - it is the discretion of the server when the id
token is actually created. So don't make this part of the normative
text. 

- I would recommend to move "3.2.1. ID Token after the authz
response section". The description of the ID Token contents does not
help the reader at that point. 

- I would suggest to organize the doc
around the response types, i.e. describe the whole flow for code. Right
now it is distributed over 2 sections (incl. Section 4 "Tokens
Endpoint"). 

regards,
Torsten. 

Am 01.08.2013 06:27, schrieb Nat
Sakimura: 

> +1 
> 
> I am trying to figure out how we can streamline
the documentations. 
> Now that we are done with the implementer's draft
vote that diff is not that important any more as technical content is
determined and IPR is locked in, now is the time to do a major surgery
to fix the documentation clutter that is caused by its history. 
> 
>
There are several proposals on the table right now in the AB/Connect WG
at OIDF. 
> 
> My proposal at the moment is to reorganize the doc into:

> 
> * OpenID Connect Core
> * OpenID Connect Discovery
> * OpenID
Connect Dynamic Registration
> * OpenID Connect Advanced Claims
Extension
> * OpenID Connect Advanced Client Authentication Methods
Extension
> * OpenID Connect Self-Issued Provider Extension
> * OpenID
Connect JSON Based Request Extension
> 
> Currently, I am experimenting
with whether keeping the different flows in the Core makes sense or it
is better to split them out. 
> 
> Here is the link to the Core draft I
am experimenting with: http://bit.ly/19yHvJB [14] 
> XML and HTML
versions are in the same repository as well. 
> 
> Your input will be
most welcome. 
> 
> Nat 
> 
> 2013/8/1 Richer, Justin P.
<[e-mail]>
> 
>> +1 
>> 
>> On Jul 31, 2013, at 5:49 PM, Bill
Mills <[e-mail]> wrote: 
>> 
>>> Rather than extending
OAuth for something OpenID already does... why don't we get a simple
informational example doc to show how to implement the most basic OpenID
service, which is the same functionality on a standard that's already
written? 
>>> 
>>> This is sounding more and mor elike a documentation
problem. 
>>> 
>>> -------------------------
>>> FROM: Prateek Mishra
<[e-mail]>
>>> TO: Nat Sakimura <[e-mail]>

>>> CC: "[e-mail] WG" <[e-mail]> 
>>> SENT: Wednesday, July
31, 2013 2:38 PM
>>> SUBJECT: [OAUTH-WG] Need for Extending OAuth with
AuthN (was Re: Fwd: New Version Notification for
draft-hunt-oauth-v2-user-a4c-00.txt)
>>> 
>>> Nat - 
>>> 
>>> thanks for
the detailed response. I did review the links you sent out but it
remained unclear to me which
>>> features are MTI and which are not. For
example, there is nothing in the Basic Client Profile that suggests
>>>
that Section 2.3 is optional. I also could not find any definition for "
non-dynamic OpenID Connect Server".
>>> 
>>> I dont think there is a
need to duplicate portions of the draft specification text in a new
document. One solution
>>> that was used in SAML 2.0 was to define a
conformance document which described several different 
>>> operational
modes and explained how only a small set of features needed to be
implemented in certain modes.
>>> 
>>>
http://docs.oasis-open.org/security/saml/v2.0/saml-conformance-2.0-os.pdf
[12]
>>> 
>>> There are probably other smarter ways to achieve the same
effect.
>>> 
>>> Given this situation, I do think its a reasonable task
for the OAuth community to consider the need for 
>>> a minimal
extension to OAuth that accommodates authentication. The community
should be made aware that 
>>> RFC 6749 is being misused for federated
authentication, as explained in - 
>>> 
>>>
http://www.independentid.com/2013/07/simple-authentication-for-oauth-2-what.html
[13] 
>>> 
>>> and that there doesn't appear to be a simple solution
that is currently available. It would be great if it turned
>>> out that
OpenID Connect offered such a solution but that isn't clear to me.
>>>

>>> Thx,
>>> prateek
>>> 
>>>> Inline: 
>>>> 
>>>> 2013/7/31 Prateek
Mishra <[e-mail]>
>>>> 
>>>>> Nat - 
>>>>> 
>>>>> your
blog posting is helpful to those of us who are looking for a minimal
extension of OAuth with 
>>>>> an authenticator. Many implementors are
seeking a modest extension of OAuth, not an entire new protocol
>>>>>
stack. I believe that is the point of Phil Hunt's proposal to the OAuth
committee.
>>>>> 
>>>>> I do have some questions for about the
statements made in the blog - 
>>>>> 
>>>>> A) Can you direct me to a
single OpenID Connect draft specification document where steps 1 and 2
are described?
>>>> 
>>>> Actually, it is not a single spec, that the
Standard is referencing others. 
>>>> The Standard is kind of cluttered
because it has 6 response types and three request types in it. 
>>>> I
suppose it would be much easier for the readers to split them into
coherent pieces, though that means duplicate texts. 
>>>> 
>>>> The
easiest approach here is to read the Basic Client Profile.
http://openid.net/specs/openid-connect-basic-1_0-28.html [10] 
>>>>
Then, read OAuth 2.0 Multiple Response Type Encoding Practices
http://openid.net/specs/oauth-v2-multiple-response-types-1_0-08.html
[11] . 
>>>> 
>>>>> B) If I implement steps 1 and 2, do I then have a
conformant OpenID Connect implementation? Are there no 
>>>>> other MTI
protocol exchanges in OpenID Connect?
>>>> 
>>>> Yes, for a non-dynamic
OpenID Connect Server. 
>>>> 
>>>> Nat 
>>>> 
>>>> Thanks,
>>>> 
>>>>>
px solid; margin-left:5px; width:100%"> 
>>>>> I have written a short
blog post titled "Write an OpenID Connect server in three simple steps
[8]". 
>>>>> 
>>>>> Really, there is not much you need to on top of
OAuth 2.0. 
>>>>> 
>>>>> It puzzles me why you need to create a draft
with only minor variances in parameter names. 
>>>>> 
>>>>>> e.g.,

>>>>>> session instead of id_token 
>>>>>> lat instead of iat 
>>>>>>
alv instead of acr 
>>>>>> etc.
>>>>> 
>>>>> If you change those
parameter names, you will have a conformant profile of OpenID Connect.

>>>>> 
>>>>> Nat 
>>>>> 
>>>>> 2013/7/31 John Bradley
<[e-mail]>
>>>>> 
>>>>>> Connect dosen't require a userinfo
endpoint. It is required for interoperability if you are building an
open IdP. For an enterprise type deployment discovery, registration,
userifo are all optional. 
>>>>>> 
>>>>>> The server is required to pass
the nonce which is equivalent to a request ID through to the JWT if the
client sends it in the request. 
>>>>>> 
>>>>>> Justin is correct.

>>>>>> 
>>>>>> John B. 
>>>>>> 
>>>>>> On 2013-07-30, at 5:30 PM, Phil
Hunt <[e-mail]> wrote: 
>>>>>> 
>>>>>>> Forgot reply
all.
>>>>>>> 
>>>>>>> Phil 
>>>>>>> 
>>>>>>> Begin forwarded
message:
>>>>>>> 
>>>>>>>> FROM: Phil Hunt
<[e-mail]>
>>>>>>>> DATE: 30 July, 2013 17:25:46
GMT+02:00
>>>>>>>> TO: "Richer, Justin P." <[e-mail]>
>>>>>>>>
SUBJECT: RE: [OAUTH-WG] NEW VERSION NOTIFICATION FOR
DRAFT-HUNT-OAUTH-V2-USER-A4C-00.TXT
>>>>>>> 
>>>>>>>> The whole point is
authn only. Many do not want or need the userinfo endpoint. 
>>>>>>>>

>>>>>>>> Phil 
>>>>>>>> 
>>>>>>>> On 2013-07-30, at 17:17, "Richer,
Justin P." <[e-mail]> wrote:
>>>>>>>> 
>>>>>>>>> What do you
mean? You absolutely can implement a compliant OIDC server nearly as
simply as this. The things that you're missing I think are necessary for
basic interoperable functionality, and are things that other folks using
OAuth for authentication have also implemented. Namely: 
>>>>>>>>>

>>>>>>>>> - Signing the ID token (OIDC specifies the RS256 flavor of
JWS, which is easy to do with JWT). Without a signed and verifiable ID
token or equivalent, you're asking for all kinds of token injection
problems. 
>>>>>>>>> - Session management requests (max auth age, auth
time) 
>>>>>>>>> - Not fall over with other parameters that you don't
support (display, prompt, etc). 
>>>>>>>>> 
>>>>>>>>> See here for more
information: 
>>>>>>>>> 
>>>>>>>>>
http://openid.net/specs/openid-connect-messages-1_0.html#ServerMTI [7]

>>>>>>>>> 
>>>>>>>>> Additionally, something that's really important to
support is the User Info Endpoint, so you can actually get user profile
information beyond just the simple "someone was here" claim -- this was
the real value of Facebook Connect from an RP's perspective. Some people
will probably want to use SCIM for this, too, and that's fine.

>>>>>>>>> 
>>>>>>>>> -- Justin 
>>>>>>>>> 
>>>>>>>>> On Jul 30, 2013,
at 10:54 AM, Phil Hunt <[e-mail]> 
>>>>>>>>> wrote:

>>>>>>>>> 
>>>>>>>>>> The oidc specs do not allow this simple an
implementation. The spec members have not shown interest in making
changes as they say they are too far down the road. 
>>>>>>>>>>

>>>>>>>>>> I have tried to make my draft as close as possible to oidc
but maybe it shouldn't be clarity wise. I am interested in what the
group feels is clearest. 
>>>>>>>>>> 
>>>>>>>>>> From an ietf
perspective the concern is improper use of the 6749 for authn. Is this a
bug or gap we need to address?
>>>>>>>>>> 
>>>>>>>>>> Phil 
>>>>>>>>>>

>>>>>>>>>> On 2013-07-30, at 16:46, "Richer, Justin P."
<[e-mail]> wrote:
>>>>>>>>>> 
>>>>>>>>>>> From what I read,
you've defined something that uses an OAuth 2 code flow to get an extra
token which is specified as a JWT. You named it "session_token" instead
of "id_token", and you've left off the User Information Endpoint -- but
other than that, this is exactly the Basic Client for OpenID Connect. In
other words, if you change the names on things you've got OIDC, but
without the capabilities to go beyond a very basic "hey there's a user
here" claim. This is the same place that OpenID 2.0 started, and it was
very, very quickly extended with SREG, AX, PAPE, and others for it to be
useful in the real world of distributed logins. You've also left out
discovery and registration which are required for distributed
deployments, but I'm guessing that those would be modular components
that could be added in (like they are in OIDC). 
>>>>>>>>>>>

>>>>>>>>>>> I've heard complaints that OIDC is complicated, but it's
really not. Yes, I agree that the giant stack of documents is
intimidating and in my opinion it's a bit of a mess with Messages and
Standard split up (but I lost that argument years ago). However, at the
core, you've got an OAuth2 authorization server that spits out access
tokens and id tokens. The id token is a JWT with some known claims (iss,
sub, etc) and is issued along side the access token, and its audience is
the *client* and not the *protected resource*. The access token is a
regular old access token and its format is undefined (so you can use it
with an existing OAuth2 server setup, like we have), and it can be used
at the User Info Endpoint to get profile information about the user who
authenticated. It could also be used for other services if your AS/IdP
protects multiple things. 
>>>>>>>>>>> 
>>>>>>>>>>> So I guess what I'm
missing is what's the value proposition in this spec when we have
something that can do this already? And this doesn't seem to do anything
different (apart from syntax changes)? 
>>>>>>>>>>> 
>>>>>>>>>>> --
Justin 
>>>>>>>>>>> 
>>>>>>>>>>> On Jul 29, 2013, at 4:14 AM, Phil Hunt
<[e-mail]> wrote: 
>>>>>>>>>>> 
>>>>>>>>>>>> FYI. I have
been noticing a substantial number of sites acting as OAuth Clients
using OAuth to authenticate users. 
>>>>>>>>>>>> 
>>>>>>>>>>>> I know
several of us have blogged on the issue over the past year so I won't
re-hash it here. In short, many of us recommended OIDC as the correct
methodology. 
>>>>>>>>>>>> 
>>>>>>>>>>>> Never-the-less, I've spoken
with a number of service providers who indicate they are not ready to
make the jump to OIDC, yet they agree there is a desire to support
authentication only (where as OIDC does IDP-like services).

>>>>>>>>>>>> 
>>>>>>>>>>>> This draft is intended as a minimum
authentication only specification. I've tried to make it as compatible
as possible with OIDC. 
>>>>>>>>>>>> 
>>>>>>>>>>>> For now, I've just
posted to keep track of the issue so we can address at the next
re-chartering. 
>>>>>>>>>>>> 
>>>>>>>>>>>> Happy to answer questions and
discuss. 
>>>>>>>>>>>> 
>>>>>>>>>>>> Phil 
>>>>>>>>>>>> 
>>>>>>>>>>>>
@independentid 
>>>>>>>>>>>> www.independentid.com [5]
[e-mail]
>>>>>>>>>>>> 
>>>>>>>>>>>> Begin forwarded message:

>>>>>>>>>>>> 
>>>>>>>>>>>>> FROM:
[e-mail]
>>>>>>>>>>>>> 
>>>>>>>>>>>>> SUBJECT: NEW
VERSION NOTIFICATION FOR
DRAFT-HUNT-OAUTH-V2-USER-A4C-00.TXT
>>>>>>>>>>>>> 
>>>>>>>>>>>>> DATE:
29 July, 2013 9:49:41 AM GMT+02:00
>>>>>>>>>>>>> 
>>>>>>>>>>>>> TO: Phil
Hunt <[e-mail]>, Phil Hunt <[e-mail]>, Phil Hunt
<>
>>>>>>>>>>>>> 
>>>>>>>>>>>>> A new version of I-D,
draft-hunt-oauth-v2-user-a4c-00.txt
>>>>>>>>>>>>> has been successfully
submitted by Phil Hunt and posted to the
>>>>>>>>>>>>> IETF
repository.
>>>>>>>>>>>>> 
>>>>>>>>>>>>> Filename:
draft-hunt-oauth-v2-user-a4c
>>>>>>>>>>>>> Revision: 00
>>>>>>>>>>>>>
Title: OAuth 2.0 User Authentication For Client
>>>>>>>>>>>>> Creation
date: 2013-07-29
>>>>>>>>>>>>> Group: Individual
Submission
>>>>>>>>>>>>> Number of pages: 9
>>>>>>>>>>>>> URL:
http://www.ietf.org/internet-drafts/draft-hunt-oauth-v2-user-a4c-00.txt
[1]
>>>>>>>>>>>>> Status:
http://datatracker.ietf.org/doc/draft-hunt-oauth-v2-user-a4c
[2]
>>>>>>>>>>>>> Htmlized:
http://tools.ietf.org/html/draft-hunt-oauth-v2-user-a4c-00
[3]
>>>>>>>>>>>>> 
>>>>>>>>>>>>> Abstract:
>>>>>>>>>>>>> This
specification defines a new OAuth2 endpoint that enables
user
>>>>>>>>>>>>> authentication session information to be shared with
client
>>>>>>>>>>>>> applications.
>>>>>>>>>>>>> 
>>>>>>>>>>>>> Please
note that it may take a couple of minutes from the time of
submission
>>>>>>>>>>>>> until the htmlized version and diff are
available at tools.ietf.org [4].
>>>>>>>>>>>>> 
>>>>>>>>>>>>> The IETF
Secretariat
>>>>>>>>>>>>
_______________________________________________
>>>>>>>>>>>> OAuth
mailing list
>>>>>>>>>>>> [e-mail]
>>>>>>>>>>>>
https://www.ietf.org/mailman/listinfo/oauth [6]
>>>>>>>
_______________________________________________
>>>>>>> OAuth mailing
list
>>>>>>> [e-mail]
>>>>>>>
https://www.ietf.org/mailman/listinfo/oauth [6]
>>>>>> 
>>>>>>
_______________________________________________
>>>>>> OAuth mailing
list
>>>>>> [e-mail]
>>>>>>
https://www.ietf.org/mailman/listinfo/oauth [6]
>>>>> 
>>>>> -- 
>>>>>
Nat Sakimura (=nat) 
>>>>> Chairman, OpenID Foundation
>>>>>
http://nat.sakimura.org/ [9]
>>>>> @_nat_en 
>>>>> 
>>>>>
_______________________________________________
>>>>> OAuth mailing
list
>>>>> [e-mail]
>>>>>
https://www.ietf.org/mailman/listinfo/oauth [6]
>>>>> 
>>>>> -- 
>>>>>
Nat Sakimura (=nat) 
>>>>> Chairman, OpenID Foundation
>>>>> h
>>>>
imura.org/
>>>> @_nat_en
>>> 
>>>
_______________________________________________
>>> OAuth mailing
list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/oauth
[6]
>>> 
>>> _______________________________________________
>>> OAuth
mailing list
>>> [e-mail]
>>>
https://www.ietf.org/mailman/listinfo/oauth [6]
> 
> -- 
> Nat Sakimura
(=nat) 
> Chairman, OpenID Foundation
> http://nat.sakimura.org/ [9]
>
@_nat_en 
> 
> _______________________________________________
> OAuth
mailing list
> [e-mail]
>
https://www.ietf.org/mailman/listinfo/oauth [6]

 

Links:
------
[1]
http://www.ietf.org/internet-drafts/draft-hunt-oauth-v2-user-a4c-00.txt
[2]
http://datatracker.ietf.org/doc/draft-hunt-oauth-v2-user-a4c
[3]
http://tools.ietf.org/html/draft-hunt-oauth-v2-user-a4c-00
[4]
http://tools.ietf.org/
[5] http://www.independentid.com/
[6]
https://www.ietf.org/mailman/listinfo/oauth
[7]
http://openid.net/specs/openid-connect-messages-1_0.html#ServerMTI
[8]
http://nat.sakimura.org/2013/07/28/write-openid-connect-server-in-three-simple-steps/
[9]
http://nat.sakimura.org/
[10]
http://openid.net/specs/openid-connect-basic-1_0-28.html
[11]
http://openid.net/specs/oauth-v2-multiple-response-types-1_0-08.html
[12]
http://docs.oasis-open.org/security/saml/v2.0/saml-conformance-2.0-os.pdf
[13]
http://www.independentid.com/2013/07/simple-authentication-for-oauth-2-what.html
[14]
http://bit.ly/19yHvJB