Re: [Txauth] Reviewing draft-hardt-xauth-protocol-11 - Dictionary

Justin Richer <> Thu, 23 July 2020 15:07 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 8FF0B3A091F for <>; Thu, 23 Jul 2020 08:07:39 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -1.918
X-Spam-Status: No, score=-1.918 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, HTML_MESSAGE=0.001, RCVD_IN_MSPIKE_H4=-0.01, RCVD_IN_MSPIKE_WL=-0.01, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id Zqb3rEMOKYBc for <>; Thu, 23 Jul 2020 08:07:37 -0700 (PDT)
Received: from ( []) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by (Postfix) with ESMTPS id 06DCB3A0920 for <>; Thu, 23 Jul 2020 08:07:32 -0700 (PDT)
Received: from [] ( []) (authenticated bits=0) (User authenticated as jricher@ATHENA.MIT.EDU) by (8.14.7/8.12.4) with ESMTP id 06NF7Q9I029663 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Thu, 23 Jul 2020 11:07:27 -0400
From: Justin Richer <>
Message-Id: <>
Content-Type: multipart/alternative; boundary="Apple-Mail=_8C0070B5-1405-47DD-B72C-687FC347EF37"
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.\))
Date: Thu, 23 Jul 2020 11:07:26 -0400
In-Reply-To: <>
Cc:, Dick Hardt <>, Tom Jones <>, Denis <>
To: Francis Pouatcha <>
References: <> <> <> <> <> <> <> <>
X-Mailer: Apple Mail (2.3608.
Archived-At: <>
Subject: Re: [Txauth] Reviewing draft-hardt-xauth-protocol-11 - Dictionary
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Thu, 23 Jul 2020 15:07:40 -0000

Up-front question to the Chairs, do we have a good place to capture this kind of thing? A non-spec document like this would be a good candidate for a wiki page. Do we have plans to set up a GitHub repository or similar? If so we could use the wiki capabilities there to at least get this stuff written down. Use cases would be good to capture there as well. 


First, I want to thank you for putting the work in to getting some vocabulary down. This is not an easy task! Some notes on specific items inline below.

> On Jul 21, 2020, at 8:33 PM, Francis Pouatcha <> wrote:
> Hello Dick, Justin, Tom, Denis,
> Below is a new version of the dictionary includings comments from different threads. Change log includes:
> - avoided the expression "owns a resource" in favor of the expression "controls a resource"
> - included the clarification "resource access" and "claim release"
> - included a clarification User vs. RO
> - tried to clarify the word User-Agent
> - added the definition of a Claim
> Terms
> =====
> Party - represents a role in the GNAP protocol flow. A party can take the form of a web service, an application installed on the user device and acting autonomously or the form of a natural person (resp. of an autonomous device) using an application to interact with other parties.
> Resource - a piece of data or web service
>       - controlled by a Resource Owner (RO),
>       - held and guarded by a Resource Server (RS) and
>       - serviced by the RS to a Client, if the Client provides a valid Authorization.
> Claim - a piece of data
>       - controlled by a Resource Owner (RO),
>       - held and guarded by the Grant Server (GS) and
>       - released by the GS to a Client as part of an Authorization.

The differentiation between “resource” and “claim” is important since it separates items that get tied to an access token’s rights (“resources”) with items that come back directly to the client without going through a separate API. The term “claim” is problematic here though because it is used broadly in the community to mean, roughly, “a statement about the end user”. It’s therefore both too broad and too narrow in the wrong ways: a “claim” could come back as part of an API response, like the OIDC UserInfo Endpoint, and so it’s not a good name for this item. Even the OIDC “claims” query language defines several different targets for returning the “claims” about the user, and none of them match the method that GNAP is looking to use here. A “claim” is also generally understood in the identity community to be a statement about the user, and the data coming back directly to the client might be about the user or something else, especially as we look forward to extensions of GNAP. 

I think we should focus on the separation of the delivery mechanism, but I’m not sure what good alternatives would be. Perhaps:

 - “access right” for things tied to the API/token
 - “data response” for things coming back directly to the client

I’m not particularly thrilled by these names, but I think recent list conversation shows that we need to get away from “claim” as it’s overloaded.

> Resource Owner (RO) - the party that
>       - controls a Resource,
>       - relies on the services the GS to manage access to that Resource,
>       - relies on the services of a RS to hold the Resource and guard access to that Resource,
>       - controls a Claim,
>       - relies on the services the GS to release that Claim to a Client.
> Resource Server (RS) - the party that 
>       - holds a resource and guards access to that resource on behalf of the RO,
>       - services the Resource to the requesting Client, provided the Client presents a valid Authorization.
>       The RS is generally deployed as a web service.
> Grant Server (GS) - the party that 
>       - manages access to a Resource on behalf of the RO,
>       - holds Claims and releases them when consented by the RO.
>       For each Resource access request, the GS might request the Consent of the RO and produce a corresponding Authorization that is given to the requesting Client.

I don’t like the shift from “Authorization Server” to “Grant Server”, and it especially doesn’t make sense in light of the definitions of “grant” and “authorization” above. I believe we should stick with “Authorization Server” if it’s a single role. I would also propose that we have a separate term for the RO-interactive portion vs. the client-facing portion, as has been brought up in a few threads. Maybe these get different names entirely, or maybe they are “aspects” of the AS, I’m not sure.

> Consent - act of an RO :
>       - approving access to a Resource, means approval that a Resource he controls can be given to the Client by the RS.
>       - approving release of a Claim, means approval that a Claim he controls can be included by the GS in an Authorization to be returned to the Client.
> Grant - material form of an RO Consent. In order not to interact with the RO on each Resource access request, the GS might store the RO Consent in the form of a Grant for reuse.
> Authorization - externalized form of a Grant as known to the GS and the RS.
>       - The GS converts a Grant into an Authorization for use in a Resource access request.
>       - The RS evaluates an Authorization to decide on whether or not to release the Resource to the Client.
> Client - the party that provides the infrastructure used by a User to access a Resource. The client infrastructure is designed to:
>       - Receive the resource access request from the User,
>       - Interact with the RS to discover authorization requirements,
>       - Interact with the GS to obtain an Authorization to access the Resource,
>       - Interact with the RS to access the Resource on behalf of the User.

I think we should be clear that what we call the “client” is a piece of software operated by the user. It’s true that this software might have multiple parts of its “infrastructure” but it’s always software.

> User - the party using the infrastructure of the Client to gain access to a Resource or a Claim.

In UMA, we call this the “Requesting Party”, or RqP. We might want to consider adopting this terminology here as well.

> Clarifications:
> ==========
> User vs. RO : 
>       - the User is the party using the infrastructure of the Client to gain access to a Resource or a Claim,
>       - the RO is the party controlling access to a Resource or controlling the release of a Claim.
> In many use cases, User and RO will be represented by the same party (see oAuth2), but GNAP exposes use cases where User and RO are represented by different parties.
> User-Agent :
> User and RO are parties that might be represented by a natural person or an autonomous device. In those cases, User (resp. RO) might need the help of an application to interact with other parties. Such an application is generally referred to as the "User-Agent".

The term “user agent” is often used to be equivalent to the web browser, and not the more general use here. I think we’re seeing more “agent” terminology for helper applications like digital wallets and AI bots, which would seem to be covered under your definition here.

> Feedbacks are welcome.

Again, thanks so much for getting this going!

 — Justin

> Best regards,
> /Francis
> On Tue, Jul 21, 2020 at 2:27 PM Francis Pouatcha < <>> wrote:
> On Tue, Jul 21, 2020 at 2:15 PM Dick Hardt < <>> wrote:
> On Tue, Jul 21, 2020 at 11:09 AM Francis Pouatcha < <>> wrote:
> On Tue, Jul 21, 2020 at 1:59 PM Dick Hardt < <>> wrote:
> FYI: I consider the release of claims to be an "authorization" as well. 
> Great. Then it is included. 
> I pivoted to using the term Grant rather than authorization so that it included both authorizing access to a resource, and authorizing the release of an identity claim.
> Perhaps we should not use the word "authorization"? 
> Authorization is currently used to refer to the token issued by the GSand presented to the RS by the Client. I have no alternative word for now.
> "resource access" and "claim release"
> A Grant then covers both?
> Yes, Great! The word Grant fits best for both. I suggest adding this clarification to that dictionary.
> IE, the user authorizes the GS to release a DOB claim to a Client.
> I guess you mean the "RO" authorizes the GS to release a DOB!
> Yes, thanks for clarifying!
> /Dick
> -- 
> Francis Pouatcha
> Co-Founder and Technical Lead
> adorsys GmbH & Co. KG
> <>
> -- 
> Francis Pouatcha
> Co-Founder and Technical Lead
> adorsys GmbH & Co. KG
> <>