Re: [GNAP] Terminology - into Github Issues

[Francis Pouatcha <[e-mail]>]

We have been having a lot of great suggestions and discussion in the list
on terms. As we go on, a lot of knowledge gets buried in the mailing
archives. This why i suggest we use the use cases github wiki to:

- start compiling discussions on single terms into issues (tickets),
- also create a ticket for each use case, either linking the wiki page,

The wiki page will be used to hold the last consolidated state of the
definition or use case.

Using tickets to complement wiki pages will allow us to:
- move valuable contributions on each word or use case to the comment
section of the corresponding ticket.
- allow contributors or visitors to read the summary of what was discussed
on a term (resp. use case) before proceeding with additional
comments/questions.
- Help focus toward reusable content.

What do you think?

Best regards
/Francis

On Fri, Aug 14, 2020 at 10:30 AM Justin Richer <[e-mail]> wrote:

> +1 for “end user” as the human person, and perhaps “<client> operator” as
> the role they play when, you know, operating the <client>. (Where <client>
> should still have a more specific name.)
>
>  — Justin
>
> On Aug 14, 2020, at 8:23 AM, Fabien Imbault <[e-mail]>
> wrote:
>
> Hi Denis,
>
> Thanks for your feedback.
> Comments inline (marked with FI).
>
> Fabien
>
> On Fri, Aug 14, 2020 at 12:02 PM Denis <[e-mail]> wrote:
>
>> Hi Fabien,
>>
>> Thank you for your inputs, a ball is finally rolling.
>>
>> An attempt.
>>
>> I would add upfront: User =  human person
>>
>> FI : then end-user is clearer if you want to indicate specifically a
> human user. One can also create system users.
>
>> *<Client> = application that requests access to Resource Servers (RS)
>> through a Grant Server (GS). *
>> Examples: a web server, a browser-based app, a mobile app, an IoT device.
>>
>> A few explanations: "through" does not sound appropriate since it could
>> be interpreted as the GS being necessarilly placed between the Client and
>> the RS.
>>                                       In addition, more than one GS may
>> be necessary.
>>
>> My proposal:  *<Client> = application that requests access to Resource
>> Servers (RS) **on behalf of a User by using one or more Grant Servers
>> (GS) *
>> *Examples: a web server, a browser-based app, a mobile app.*
>>
>> FI: agreed.
>
>>
>> *GS = computing service that manages the grant lifecycle to a <Client> on
>> behalf of a Resource Controler (RC).*
>> Note : for privacy reasons, the GS may be issuing grants without
>> knowledge of which resources are requested.
>>
>> I dislike "*on behalf of a Resource Controler (RC)*". The GS may be
>> fully ignorant of the existence of the RSs and hence of the RCs.
>> In addition, "*grant life cycle*" is undefined.
>>
>> My proposal:  *GS = server issuing access tokens to a Client after
>> successfully authenticating the User*
>>
>>
>> *Note 1: for privacy reasons, the GS may be issuing access tokens without
>> the knowledge of which resources are requested.Note 2: a GS is able to
>> disclose to a User the User attributes that it manages.  *
>>
>> FI: I find the new definition less clear. It's not because you don't know
> which RS is called that we shouldn't say the decision is made by the RC.
> "grant life cycle" is indeed currently undefined, what i had in mind is
> basically the grant flow from the GNAP protocol, possibly also including
> revocation etc.
> Not sure why Note 2 is important to put here.
>
>
>> *RS = computing service that grants access only if its Resource Controler
>> (RC) consents.*
>> Note : the consent may involve a human interaction or may be automated
>> based on access control policies.
>>
>> I dislike "*its Resource Controler (RC) consents" *because it may let
>> think that a human person always needs to consent.
>>
>> My proposal: *R*
>> *S = server hosting protected resources, capable of accepting and
>> responding to protected resource requests
>> when access tokens are being used*
>>
>> FI : that is why I suggested a note to make sure it is correctly
> understood. I'm not sure the proposed alternative is clearer.
>
>>
>> *RC = entity which is controlling the access to a protected resource. *
>> Note : a RC may be manually operated by a human or delegated to a
>> machine, partially or completely.
>>
>> A RC is not an entity but a function. I would also place the machine case
>> first.
>>
>> My proposal:
>> *RC = function tightly coupled with a RS which controls the accesses to a
>> protected resource*                        Note : the function may be
>> operated either by a machine or by a human person or by some combination of
>> both.
>>
>> FI : your proposition on the note makes it much better. On the main
> definition, I'm not sure what you mean by function, as a result I'm not
> sure a reader would understand. Why do you need to say "tightly coupled?"
>
>> *Consent = the process of asking a RC to accept or decline based on a
>> grant request presentation, resulting in either a “yes” or “no” consent
>> decision.*
>>
>> I would instead speak of the "User Consent". The User Consent is a set of
>> choices among a proposed set of choices. It is not simply a "yes" or "no"
>> consent decision.
>>
>> My proposal: *User Consent = ability for a User, after being informed,
>> of choosing to release or not to a RS some attributes contained in one or
>> more access tokens*
>>
>>
> FI: this may be misleading I think. The consent is done by a RC (or in
> OAuth terms, RO), not the application user.
> I agree there may be a combination of consent decisions, but I think it's
> important to say that in the end for each individual choice, you do have a
> yes/no decision.
>
>>
>> Denis
>>
>>
>> Fabien
>>
>> On Thu, Aug 13, 2020 at 3:55 PM Denis <[e-mail]> wrote:
>>
>>> Fabien,
>>>
>>> IMHO, nothing is wrong with keeping "Client" since it has a wide spread
>>> usage
>>> ... but only as long as we can agree on a short and a clear definition
>>> for it.
>>>
>>> I can provide the beginning of such a definition: " application ..."
>>>
>>> If someone could go a little bit further, this would help. :-)
>>>
>>> A similar argumentation for GS.  It could be used but only as long as
>>> we can agree on a short and a clear definition for it.
>>> Any proposal ?
>>>
>>> Denis
>>>
>>> Hi,
>>>
>>> Nothing inherently wrong with Client/AS, which has worked for years in
>>> the context of OAuth2. The question is to know whether we can build the new
>>> protocol with the same concepts and deal with their known limitations, or
>>> if we're better off with more adapted concepts less prone to
>>> misunderstandings.
>>>
>>> Verb vs Noun:
>>> Problem is that the grant (noun) can only be understood if there is a
>>> grant(verb), i.e. some action that grants something.
>>> The grant (noun) definition directly derives from the verb : "something
>>> granted ..."
>>>
>>> I personally have no issue even with the fairly convoluted "The Grant
>>> Server issues a grant to the Grant Client representing access that has been
>>> granted" (except perhaps from the word Client, but that's a different
>>> issue).
>>> By the way, grant is nothing new, it's used extensively in OAuth2 as
>>> "grant types" (whatever that means).
>>>
>>> Dick summarized well the reasons why he uses GS instead of AS :
>>> 1) "grant" is in the working group name (a weaker reason, but still has
>>> been approved). Question: would our reasoning if the protocol ended up
>>> being called OAuth3?
>>> 2) grant = larger in scope than AS (not only authorization), as it at
>>> least includes idclaims + other use cases like payment (?) - no consensus,
>>> see difference in appreciation between Justin and Dick
>>>
>>> As for "Client", if most people think it is problematic, it seems a good
>>> reason to change if we find a better alternative.
>>> Quoting Dick again: "The confusion in my experience usually stems from
>>> people working with software that is acting in multiple roles. IE, the
>>> software that is acting as a client in once context, is also acting as
>>> an RS in other contexts, or even acting as an AS. The other confusion is
>>> that people view clients as being the software the user is using --
>>> although it may not be acting as a client in the oauth context." and
>>> later "I do agree that it is not the best term in GNAP. Primarily because
>>> GNAP is a combination of the client from OAuth 2, and the relying party
>>> from OIDC."
>>>
>>> So far there's no consensus however, recent tries: Initiating
>>> Application (Denis), Orchestrator (Francis).
>>>
>>> Cheers
>>> Fabien
>>>
>>>
>>>
>>>
>>> On Thu, Aug 13, 2020 at 2:59 PM Dave Tonge <[e-mail]>
>>> wrote:
>>>
>>>> I would be against using "grant" as both a verb and a noun and stick
>>>> purely with one or the other. (In the charter the only use of "grant" is in
>>>> the verb: "granted").
>>>>
>>>> Using it as both a verb and a noun will be confusing and less
>>>> accessible.
>>>>
>>>> I think it will be confusing to say "The Grant Server issues a grant to
>>>> the Grant Client representing access that has been granted"
>>>>
>>>> Whether the access takes place via a token being returned and used at a
>>>> resource server, or "claims" that are directly returned from the "Grant
>>>> Server" I think should be largely irrelevant when it comes to the naming of
>>>> the roles.
>>>>
>>>> In almost all use cases that I have seen the "Grant Server" is making a
>>>> policy based decision "to grant" access to requested resource(s). To me,
>>>> that is the fundamental operation happening. I think nearly all use cases
>>>> can be applied to that, e.g. the GS grants access to
>>>>  - identity attributes for the end user
>>>>  - verify an identity attribute (e.g. that user is over 18)
>>>>  - all users photos at resource server X
>>>>  - a single photo with id 12345 at resource server Y
>>>>  - resource of type X at any resource server that trusts the Grant
>>>> Server
>>>>  - call a payment API with specific properties (e.g. amount < 5)
>>>>  - call a file storage API
>>>>  - call a "contract signing" API with specific properties (e.g.
>>>> contract hash of xxx,)
>>>>
>>>> While "client" is problematic, it does now have wide spread usage, so
>>>> perhaps we are stuck with it.
>>>> However I agree with Justin and think "Grant Client" makes things more
>>>> confusing.
>>>>
>>>> What is wrong with keeping "Client" and "Authorization Server"?
>>>>
>>>> Dave
>>>>
>>>>
>>>>
>>>>
>>>> Moneyhub Enterprise is a trading style of Moneyhub Financial Technology
>>>> Limited which is authorised and regulated by the Financial Conduct
>>>> Authority ("FCA"). Moneyhub Financial Technology is entered on the
>>>> Financial Services Register (FRN 809360) at
>>>> https://register.fca.org.uk/. Moneyhub Financial Technology is
>>>> registered in England & Wales, company registration number 06909772.
>>>> Moneyhub Financial Technology Limited 2020 © Moneyhub Enterprise, Regus
>>>> Building, Temple Quay, 1 Friary, Bristol, BS1 6EA.
>>>>
>>>> DISCLAIMER: This email (including any attachments) is subject to
>>>> copyright, and the information in it is confidential. Use of this email or
>>>> of any information in it other than by the addressee is unauthorised and
>>>> unlawful. Whilst reasonable efforts are made to ensure that any attachments
>>>> are virus-free, it is the recipient's sole responsibility to scan all
>>>> attachments for viruses. All calls and emails to and from this company may
>>>> be monitored and recorded for legitimate purposes relating to this
>>>> company's business. Any opinions expressed in this email (or in any
>>>> attachments) are those of the author and do not necessarily represent the
>>>> opinions of Moneyhub Financial Technology Limited or of any other group
>>>> company.
>>>>
>>>>
>>>
>> --
>> TXAuth mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/txauth
>
>
>

-- 
Francis Pouatcha
Co-Founder and Technical Lead
adorsys GmbH & Co. KG
https://adorsys-platform.de/solutions/