[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: auth-response comments



  Darn. I had accomodated (or at least taken into account) all of your comments and written a new draft two months ago, but I was waiting for comments from Mark Wahl before republishing it. Well, anyway, I'm attaching it now.

Rob


"Kurt D. Zeilenga" wrote:
> 
> Rob/Mark,
> 
> Here are some comments, some new, some old... all provided to
> ensure completeness.
> 
> First, please note my general aversion to unsolicited controls.
> IMO, controls should only be sent to clients which are known to
> be able to make use of the information.
> 
> 2. Publishing support for the Authentication Response Control
> 
> s/supportedExtensions/supportedControl/
> 
> 3. Authentication Response
> 
> >The criticality field is not used.
> I would suggest "The criticality of this control SHALL be
> FALSE.  Servers SHOULD not provide the criticality field."
> 
> Not also that the controlType is determined, its the value
> of the field which is TBD.
> 
> You do not specify how AuthResponseValue is to be encoded.
> 
> I do not see the need for authMechanism?  The client knows
> the method and, if applicable, the SASL mechanism used.  However,
> what might be useful is source of credential used to in
> to complete a SASL/EXTERNAL authentication.
> 
> I note you specify the return of a DN and not an authzID.
> You assume that if an userzID is provided (or implied)
> by the client that it must be mapped to a DN and if a
> DN is provided (or implied) that it is not mapped to
> a userzID.  JeffH has made arguments that authzId should
> be the general form of LDAP authorization information.
> In fact, LDAP ACM allows authzID as subjects.  You likely
> should consider changing authDN to authzID.
> 
> 4. Security Considerations
> 
> Please make specific mention that the control is not
> protected by SASL integrity and privacy services
> negotiated by the bind operation it is provided with.
> Due to this, I suggest use of a separate (extended)
> operation instead of a bind control to request/return
> this information.
> 
>         Kurt