Re: [OAUTH-WG] Pete Resnick's No Objection on draft-ietf-oauth-assertions-17: (with COMMENT)

[Brian Campbell <[e-mail]>]

Thanks for your review and feedback, Pete. Replies are inline below...

On Tue, Oct 14, 2014 at 2:42 PM, Pete Resnick <[e-mail]>
wrote:

> Pete Resnick has entered the following ballot position for
> draft-ietf-oauth-assertions-17: No Objection
>
> When responding, please keep the subject line intact and reply to all
> email addresses included in the To and CC lines. (Feel free to cut this
> introductory paragraph, however.)
>
>
> Please refer to http://www.ietf.org/iesg/statement/discuss-criteria.html
> for more information about IESG DISCUSS and COMMENT positions.
>
>
> The document, along with other ballot positions, can be found here:
> http://datatracker.ietf.org/doc/draft-ietf-oauth-assertions/
>
>
>
> ----------------------------------------------------------------------
> COMMENT:
> ----------------------------------------------------------------------
>
> 3 -
>
>    Assertions used in the protocol exchanges defined by this
>    specification MUST always be protected against tampering using a
>    digital signature or a keyed message digest applied by the issuer.
>
> Why is that? Aren't you using assertions over a protected channel (as
> required by the spec) and therefore not need to sign the assertions?
> Indeed, why would a self-issued Bearer Assertion need to be signed at
> all? Does that even make sense?
>
>
Yes, assertions are sent over a protected channel, which does provide
integrity protection for the transport form client to AS and also gives
server authentication. But it doesn't provide client authentication, which
is kind of the point of the Client Authentication part of this draft. And
for authorization the signing or MACing is what authenticates the issuer of
the assertion - sometimes the issuer is the client but often the issuer
will be a 3rd party system.

I do agree with you in one specific case that, if the client is trusted to
be the assertion issuer and the client is properly authenticated, then an
unsigned assertion could be reasonably used as an authorization grant. But
it's a fairly rare and very specific case. And one that can be accommodated
in other ways. So it's not worth introducing the complexity and potential
security problems that having the signature be option would bring.




> 4.1 -
>
>    grant_type
>       REQUIRED.  The format of the assertion as defined by the
>       authorization server.  The value MUST be an absolute URI.
>
> That MUST is unnecessary. It's just definitional from 6749, 4.5 (which,
> as it happens, doesn't use 2119 language for this). s/MUST/will
>

Makes sense.


>
>    assertion
>       REQUIRED.  The assertion being used as an authorization grant.
>       Specific serialization of the assertion is defined by profile
>       documents.  The serialization MUST be encoded for transport within
>       HTTP forms.  It is RECOMMENDED that base64url be used.
>
> The MUST seems weird here. Are you saying that the profile could not
> possibly have a serialization for an assertion that did not require
> further encoding? But the RECOMMENDED seems downright wrong: Either an
> implementer *needs* to know the encoding independently of the profile,
> and therefore this needs to be a MUST, or the profile is going to
> describe the encoding, which could be base64url or could be something
> else, and the implementation will do whatever the profile says. If you
> really want to say something here, I suggest replacing the last two
> sentences with:
>
>       If the assertion is going to be transported within HTTP forms, the
>       profile document needs to describe what (if any) encoding will be
>       needed to do so. The base64url encoding is widely implemented and
>       therefore suggested.
>
>
In reading it again, I agree with you that it's weird and not appropriate
here. In fact the JWT profile itself does not require any further encoding.

Barring any objection, I suggest that the last two sentences just be
removed. So it would just be,

  "assertion
      REQUIRED.  The assertion being used as an authorization grant.
      Specific serialization of the assertion is defined by profile
      documents."



>     scope
>     [...]
>                                                    As such, the
>       requested scope MUST be equal or lesser than the scope originally
>       granted to the authorized accessor.
>
> s/MUST/will (unless you explain whether it's the server or the client
> that's supposed to be obeying that MUST, and for what reason)
>

They are both supposed to obey it - the client shouldn't ask for more and
the server will reject the request, if it does.

Is "will" more appropriate than "MUST" here? Or maybe a non 2119 "should"?


>
>       If the scope parameter and/or
>       value are omitted, the scope MUST be treated as equal to the scope
>       originally granted to the authorized accessor.  The Authorization
>       Server MUST limit the scope of the issued access token to be equal
>       or lesser than the scope originally granted to the authorized
>       accessor.
>
> In the first sentence, is this MUST for the server? (That is, shouldn't
> it be, "If the scope parameter and/or value are omitted, the server MUST
> use the value of the scope originally granted to the authorized
> accessor."?) And anyway, don't these two sentences contradict 6749, which
> says:
>
>    The authorization server MAY fully or partially ignore the scope
>    requested by the client, based on the authorization server policy or
>    the resource owner's instructions.
>    [...]
>    If the client omits the scope parameter when requesting
>    authorization, the authorization server MUST either process the
>    request using a pre-defined default value or fail the request
>    indicating an invalid scope.
>
>
Yes, I think you're correct that there is a contradiction with 6749. And,
honestly, I'm surprised at the text there as I read it again. I don't think
that was the intent.

I'd propose that the sentence, " If the scope parameter and/or
      value are omitted, the scope MUST be treated as equal to the scope
      originally granted to the authorized accessor." be removed.



Here and throughout: s/non-normative example/example (As far as I know,
> there are no other kinds in IETF documents.)
>

I thought I picked that language up from some other draft or RFC but I'm
now not sure where it came from and can't easily find other examples of the
same thing.  So I am happy to remove the "non-normative" throughout, if it
is already understood and/or not customary to say so.



>
> 4.1.1 - s/MUST construct/constructs
>

OK


>
> 4.2, client_assertion_type and client_assertion: See comments from 4.1
> regarding grant_type and assertion.
>

Agree and same answer.


>
> 4.2.1 - s/MUST construct/constructs
>

OK


>
> 5.2 -
>
> s/MUST identify/identifies
>

OK


>
> For clarification:
> OLD
>       Assertions that do
>       not identify the Authorization Server as an intended audience MUST
>       be rejected.
> NEW
>       The Authorization Server MUST reject any assertion that does not
>       contain the its own identity as the intended audience.
> END
>

OK, I agree that reads better.