Re: [OAUTH-WG] A question on token revocation.

[Torsten Lodderstedt <[e-mail]>]

Hi Prabath,

we tried to address both use cases in the first revisions of the draft. The API was well suited for client-driven revocation but not the resource owner - driven use case. There are definitely differences with respect to the protocol design, at least regarding authentication and authorization of the respective actors. This made the spec more complex and caused ambiguities and confusion. Moreover, the working group seemed not convinced by the the latter use case.

Therefore the working group decided to focus on the single use cases of the revocation by clients. This makes a lot of sense since this interface is most important with respect to interoperability.

I'm focusing right now on finishing this draft. And the open issues discussed on the list in the last couple of weeks illustrate that even this poses a considerable amount of work. So I'm very reluctant to add support for a whole new use case at that point of the process.

If you feel this is an important use case worth an RFC, don't hesitate to publish a new I-D.

regards,
Torsten.

Am 06.02.2013 um 16:37 schrieb Prabath Siriwardena <[e-mail]>:

> 
> 
> On Wed, Feb 6, 2013 at 9:04 PM, Todd W Lainhart <[e-mail]> wrote:
>> > Resource owner needs to know the consumer key (represents the OAuth Client app) & scope to revoke the access token for a given client.  
>> 
>> I see - you're saying that requiring client credentials on the end point is the problem?
> 
> In fact what I meant was - when RO authorizes the an access token for client for particular scope. Those information are kept at the AS.
> 
> Now - if the RO want to revoke access from the client - the RO needs to authenticate him self to the AS and pass the consumer key and the scope. So AS can revoke access.
> 
> Thanks & regards,
> -Prabath
>  
>> 
>> 
>> 
>> 
>> Todd Lainhart
>> Rational software
>> IBM Corporation
>> 550 King Street, Littleton, MA 01460-1250
>> 1-978-899-4705
>> 2-276-4705 (T/L)
>> [e-mail]
>> 
>> 
>> 
>> 
>> 
>> From:        Prabath Siriwardena <[e-mail]> 
>> To:        Justin Richer <[e-mail]>, 
>> Cc:        "[e-mail] WG" <[e-mail]> 
>> Date:        02/06/2013 10:31 AM 
>> Subject:        Re: [OAUTH-WG] A question on token revocation. 
>> Sent by:        [e-mail] 
>> 
>> 
>> 
>> 
>> 
>> On Wed, Feb 6, 2013 at 8:49 PM, Justin Richer <[e-mail]> wrote: 
>> 
>> On 02/06/2013 10:13 AM, Prabath Siriwardena wrote: 
>> 
>> 
>> On Wed, Feb 6, 2013 at 8:19 PM, Justin Richer <[e-mail]> wrote: 
>> These are generally handled through a user interface where the RO is authenticated directly to the AS, and there's not much need for a "protocol" here, in practice. 
>> 
>> Why do you think leaving access token revocation by RO to a proprietary API is a good practice ? IMO this an essential requirement in API security. 
>> I think it makes more sense in the same way that having a "proprietary" UI/API for managing the user consent makes sense: unless you're doing a fully dynamic end-to-end system like UMA, then there's not much value in trying to squeeze disparate systems into the same mold, since they won't be talking to each other anyway. 
>> 
>> This is required in distributed setup for each API platform from different vendors to perform in an interop manner. 
>>   
>> 
>> And since you refer to it as an "API", what will the RO be using to call this API? Is there a token management client that's separate from the OAuth client? 
>> 
>> I didn't get your question right... If you meant the how to invoke revocation end point, the the resource owner needs to know the consumer key (represents the OAuth Client app) & scope to revoke the access token for a given client.  
>> 
>> 
>> 
>> IMHO token revocation done my RO is more practical than token revocation done by the Client. 
>> They're both valid but require different kinds of protocols and considerations. This token revocation draft is meant to solve one problem, and that doesn't mean it can or should solve other seemingly related problems.
>> 
>> If you would like to see the RO-initiated token revocation go through (not grant revocation, mind you -- that's related, but different), then I would suggest that you start specifying exactly how that works. I predict it will be problematic in practice, though, as the RO often doesn't actually have direct access to the token itself. 
>> 
>> Resource owner needs to know the consumer key (represents the OAuth Client app) & scope to revoke the access token for a given client.  
>> 
>>   
>> 
>>   
>> There are larger applications, like UMA, that have client and PR provisioning that would allow for this to be managed somewhat programmatically, but even in that case you're still generally doing token revocation by a "client" in some fashion. In UMA, though, several different pieces can play the role of a "client" at different parts of the process.
>> 
>> Revoking a scope is a whole different mess. Generally, you'd want to just revoke an existing token and make a new authorization grant with lower access if you don't want that client getting that scope anymore. If you just want to downscope for a single transaction, you can already do that with either the refresh token or token chaining approaches, depending on where in the process you are. The latter of these are both client-focused, though, and the RO doesn't have a direct hand in it at this point. 
>> 
>> Why do you think it a mess. If you revoke the entire token then Client needs to go through the complete OAuth flow - and also needs to get the user consent. If RO can  downgrade the scope, then we restrict API access by the client at RS end and its transparent to the client. 
>> 
>> 
>> Downgrading the scope of tokens in the wild is hardly transparent to the client (stuff that it expects to work will suddenly start to fail, meaning that most clients will throw out the token and try to get a new one), and in a distributed system you've got to propagate that change to the RS. If you bake the scopes into the token itself (which many do) then you actually *can't* downgrade a single token, anyway. 
>> 
>> Yes.. that is the expected behavior. I mean the process is transparent. Client will notice at runtime. 
>> 
>> Thanks & regards, 
>> -Prabath 
>>   
>> 
>>  -- Justin 
>> 
>> 
>> Thanks & regards, 
>> -Prabath 
>>   
>> 
>> 
>>  -- Justin 
>> 
>> 
>> On 02/06/2013 04:35 AM, Prabath Siriwardena wrote: 
>> I am sorry if this was already discussed in this list..  
>> 
>> Looking at [1] it only talks about revoking the access token from the client. 
>> 
>> How about the resource owner..? 
>> 
>> There can be cases where resource owner needs to revoke an authorized access token from a given client. Or revoke an scope.. 
>> 
>> How are we going to address these requirements..? Thoughts appreciated... 
>> 
>> [1] http://tools.ietf.org/html/draft-ietf-oauth-revocation-04 
>> 
>> -- 
>> Thanks & Regards,
>> Prabath 
>> 
>> Mobile : [phone] 
>> 
>> http://blog.facilelogin.com
>> http://RampartFAQ.com 
>> 
>> 
>> _______________________________________________
>> OAuth mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/oauth
>> 
>> 
>> 
>> 
>> 
>> -- 
>> Thanks & Regards,
>> Prabath 
>> 
>> Mobile : [phone] 
>> 
>> http://blog.facilelogin.com
>> http://RampartFAQ.com 
>> 
>> 
>> 
>> 
>> -- 
>> Thanks & Regards,
>> Prabath 
>> 
>> Mobile : [phone] 
>> 
>> http://blog.facilelogin.com
>> http://RampartFAQ.com_______________________________________________
>> OAuth mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/oauth
> 
> 
> 
> -- 
> Thanks & Regards,
> Prabath
> 
> Mobile : [phone] 
> 
> http://blog.facilelogin.com
> http://RampartFAQ.com
> _______________________________________________
> OAuth mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/oauth