Re: [Gen-art] Gen-ART LC review of draft-ietf-abfab-aaa-saml-12

[Stephen Farrell <[e-mail]>]

Hiya,

On 10/12/15 07:32, Alejandro Pérez Méndez wrote:
> Dear Roni and chairs of the ABFAB WG,
> 
> thank you for the revision. Please, see my responses inline (specially
> the one related to point #2)
> 
> El 03/12/15 a las 15:31, Roni Even escribió:
>>
>> I am the assigned Gen-ART reviewer for this draft. For background on
>> Gen-ART, please see the FAQ at
>> <http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
>>
>> Please resolve these comments along with any other Last Call comments
>> you may receive.
>>
>> Document: draft-ietf-abfab-aaa-saml-12
>>
>> Reviewer: Roni Even
>>
>> Review Date:2015–12-3
>>
>> IETF LC End Date: 2015–12-4
>>
>> IESG Telechat date:
>>
>> Summary: This draft is almost ready for publication as an
>> Informational RFC.
>>
>> Major issues:
>>
>> Minor issues:
>>
>> 1.Why is the RADIUSNasIpAddress a string and not as specified in for
>> example in RFC2865
>>
> 
> The RADIUSNasIpAddress is a SAML metadata element, thus it has to comply
> with existing SAML types. The string type allows to encode the "display"
> value of these RADIUS attributes (e.g. "192.168.1.1", or "::1"). Note
> that current text specifies that the element contains an acceptable
> value for RADIUS NAS-IP-Address or RADIUS NAS-IPv6-Address attributes,
> so no arbitrary values are accepted nonetheless.
> 
>> 2.In general I was wondering why this is an Informational document. It
>> defines procedures and has normative language.
>>
> 
> That sounds like kind of an unfortunate bug. For some reason, it changed
> from Standards Track to Informational between versions -00 and -01.
> However, we want it standards-track with a normative downreference to
> radsec. Can it be done at this moment or does it require a more complex
> process?

Hmm. The shepherd write-up says informational is correct. If the WG
chairs want to, we can re-spin the IETF LC. But this has been so
long in the process and has slowly so I'd prefer to not do that
unless someone really cares, and it makes a difference.

For now, I've kept this on the Dec17 IESG telechat as informational
but if needed we can push it into the new year.


> 
>> 3.In the IANA consideration in section 11.1, as far as I understand
>> the IANA attribute type registry you need to ask for values for TBD1
>> and TBD2 from the unassigned space (and not the reserved space)
>>
> 
> I agree. I cannot find where we state otherwise, though. Could you point
> the specific text where we say it?
> 
>> 4.In step 2 of figure 7 (section 7.2) the text says “In step 2, the
>> Relying Party may optionally issue a <samlp:AuthnRequest> message to
>> be delivered to the Identity Provider using the SAML-Protocol RADIUS
>> attribute.”  My reading is that the rest of the steps are when this
>> message is sent, since it is  “may” what happens if the message is not
>> sent?
>>
> 
> If the <AuthnRequest> is not sent, the procedure follows the
> "unsolicited response" (explained in 7.4.4), where the IdP deliveres a
> <samlp:Assertion> element. I agree that step 4 needs to include this
> clarification. The new text should read as:
> 
> 
> 4.  Identity Provider issues <samlp:Response> to Relying Party
>        (Section 7.3.4).  In step 4, the Identity Provider issues a
>        <samlp:Response> message to the Relying Party using the SAML
>        RADIUS binding.  The response either indicates an error or
>        includes a SAML Authentication Statement in exactly one SAML
>        Assertion. If the RP did not send an <samlp:AuthnRequest>, the IdP
> 
> issues an unsolicited <samlp:Assertion>, as described in section 7.4.4.

I'll add the above as an RFC editor note.

Cheers,
S.

> 
> 
>> Nits/editorial comments:
>>
>>  1. In section 1 please expand ABFAB
>>
> 
> Ok
> 
>> 1.
>>
>>
>>  2. In section 7.2, the text says “To implement this scenario, a
>>     profile of the SAML Authentication   Request protocol is used in
>>     conjunction with the SAML RADIUS binding  defined in Section 4.” I
>>     think that the language should be more normative maybe it should
>>     say  “To implement this scenario, this profile of the SAML
>>     Authentication   Request protocol MUST Be (or SHOULD if there are
>>     other options) used in conjunction with the SAML RADIUS binding
>>     defined in Section 4.”
>>
> 
> Agree. I think "MUST be" is the one to be used.
> 
> Best regards,
> Alejandro
>