Re: [OAUTH-WG] JWT access tokens and the revocation endpoint

[Seán Kelleher <[e-mail]>]

Hi all,

Long time lurker, first time poster, glad to be finally getting involved!

In terms of weighing in on the revocation practice, I don't think this
document needs to address it as JWT ATs don't seem to require special
handling in this case. I think a general coverage of approaches to token
revocation would be more appropriate in a BCP.

One thing I'm unsure of when reading this is how a RS can depend on an AS
to give it a JWT AT. Due to the opaque nature of ATs, it seems natural that
an AS can change the AT profile it uses at its own discretion, but there's
no negotiation or parameter that can force an AS to return a JWT AT, with
the potential for breakages. Do the RS and AS agree on the profile ahead of
time?  Perhaps this is specified in a separate document, but I think the
topic of profile negotiation should be covered in this document, even at a
high level, or reference to more detailed coverage.

A few other notes on the document itself:

   - Section 2.1: Is there precedence of "application/at+jwt" being used in
   the wild that prevents the SHOULD from being a MUST?
   - Figure 1: Nitpick: An `&` that looks like it should prefix `state` is
   on the preceding line.
   - Figure 2: The example `typ` is `at+JWT`, should this be `at+jwt`?
   - Section 4: This is a personal opinion, but I'd recommend moving the
   step concerning `aud` validation to the top of the list, for focus. This is
   because, having personally taken more than a little while to figure out the
   difference between access tokens and ID tokens, I see the risk of cross-JWT
   confusion as very relevant. I think the necessity of the other steps are a
   bit more obvious.

Thanks!

Kind regards,

Seán.

On Tue, 6 Oct 2020 at 22:53, <vittorio.bertocci=[e-mail]>
wrote:

> Hi Andrii,
>
> Thanks for the thoughtful comments! Here’s my 2 c.
>
>
>
> On the proposed language: given that the JWT AT profile is just providing
> more details on the content of an AT, making JWT ATs a proper subset of all
> ATs, readers should have no reason to believe that introspection or any
> other existing spec discussing AT behavior should operate differently. That
> assumption holds for everything across the board, hence it would be a bit
> odd to call out this particular case. On the userinfo case, I would find it
> even more left field to say anything about it.
> If we do reach a consensus on specific revocation practices that apply to
> JWT ATs specifically, and we conclude that they should live in this
> document, I will be happy to add language accordingly.
>
>
>
> On the AJWT: I hear you on the dissonance that JWT AT carries, but I am
> very hesitant to introduce new acronyms to an already crowded/impenetrable
> domain.
> JWT access token might not roll off the tongue, but at this point ‘JWT’ is
> nearly a proprietary eponym and the expression “JWT token” is
> extraordinarily common in literature, a quick google query will give you
> the full measure of the phenomenon, hence I think we’ll be OK with the
> current form.
>
> Cheers,
>
> V.
>
>
>
> *From:* Andrii Deinega <[e-mail]>
> *Sent:* Tuesday, October 6, 2020 2:25 PM
> *To:* [e-mail]; [e-mail]
> *Cc:* Jim Manico <[e-mail]>; Nicolas Mora <[e-mail]>
> *Subject:* Re: [OAUTH-WG] JWT access tokens and the revocation endpoint
>
>
>
> Vittorio and WG,
>
>
>
> Would it be possible to include something like the following to
> https://tools.ietf.org/html/draft-ietf-oauth-access-token-jwt-10
>
>
> In case the authorization server exposes the introspection, revocation,
> and OpenID Connect userinfo endpoints they MUST act in the same way as it
> happens with a regular access token. That allows the AS to change the type
> of an access token on the fly and that won’t lead to interoperate issues.
> Plus, the consumers of these endpoints use them regardless of the type of
> access token.
>
>
>
> The way how the AS can notify RSs that the token revocation event happened
> (if it decides to do so) is completely left to implementers.
>
>
>
> ?
>
>
>
> Another minor editorial thing from me is it would possible to change and
> refer to "JWT access tokens" as AJWT. Thus, this document won't repeat the
> token word twice.
>
>
>
> Regards,
>
> Andrii
>
>
>
> On Tue, Oct 6, 2020 at 2:22 PM <vittorio.bertocci=
> [e-mail]> wrote:
>
> Hey Jim, regarding
> > Every logout event should trigger token revocation
> That isn’t necessarily the case. An access token represents the ability of
> a client to access a given resource; the fact that it requires an
> authentication transaction/session establishment to be issued to the client
> does not mean that the AT lifetime is tied to the lifetime of that session.
> Say that I allow LinkedIn to tweet on my behalf. Once I have done that, it
> doesn’t matter whether I stay logged in their web app or otherwise. Even if
> I log out of the session in which context I got my twitter AT, I can still
> post on LinkedIn from my native LinkedIn app and the corresponding post
> will show up on twitter as well.
> Now, one might choose to *explicitly* tie tokens lifetime to originating
> sessions lifetime, see the discussion on the OpenID Connect group about a
> possible online_access scope for influencing RTs and Ats (in particular, in
> the context of SPAs) but that's additional semantic that isn’t defined
> today.
>
> -----Original Message-----
> From: OAuth <[e-mail]> On Behalf Of Jim Manico
> Sent: Sunday, October 4, 2020 5:17 PM
> To: Nicolas Mora <[e-mail]>
> Cc: [e-mail]
> Subject: Re: [OAUTH-WG] JWT access tokens and the revocation endpoint
>
> > In this model, considering that token revocations don't happen a lot...
>
> Just a brief note, a secure piece of software makes the logout feature
> prominent. Every logout event should trigger token revocation.
>
> I’m mentioning this because a lot of OAuth solutions in the mobile space
> literally ignore the logout event, such as Facebook’s mobile OAuth
> solution.
>
> - Jim
>
> > On Oct 4, 2020, at 6:55 AM, Nicolas Mora <[e-mail]> wrote:
> >
> > Hello,
> >
> >> Le 20-10-04 à 11 h 27, Thomas Broyer a écrit :
> >>
> >>    There might be some kind of pushed events between the AS and the RS
> when
> >>    a JWT AT is revoked, to allow the RS not to introspect a JWT AT at
> all.
> >>    Like this, the RS knows if a JWT AT has been revoked or not.
> >>
> >>
> >> If there are some kind of pushed events between the AS and the RS,
> >> then it could push the revoked (and/or expired) opaque AT too, giving
> >> almost no advantage to JWT ATs.
> >>
> > Not necessarily, let's say the AS informs the RS only of the revoked
> > ATs, when a RS checks an AT, it verifies the signature first, then the
> > claims, then checks if the AT has been revoked by checking its
> > internal list filled by the AS pushed events.
> >
> > In this model, considering that token revocations don't happen a lot,
> > the ratio revoked AT/valid AT is very low, so the advantage of a JWT
> > is important, because it means not so much communication between the
> > AS and the RSs, and a very reliable AT.
> >
> > But this means a communication mechanism that isn't standardized yet.
> >
> > /Nicolas
> >
> > _______________________________________________
> > OAuth mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/oauth
>
> _______________________________________________
> OAuth mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/oauth
>
> _______________________________________________
> OAuth mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/oauth
>
> _______________________________________________
> OAuth mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/oauth
>