Re: [xacml] Issue: Ambiguity in defn of Request wrt Attributes Category

Hi Erik,

    

    That sounds good to me. I believe that is consistent

    w the process we used for XACML 2.0, so now that

    3.0 is in OS state, we should start the errata process:

    agenda item for next mtg.

    

        Thanks,

        Rich

    

    
On 3/4/2013 6:38 AM, Erik Rissanen
      wrote:

    

    
      
      
Hi,

        

        I would say that if you are not supporting multiple decisions,
        then it's an error. It might be the case that it's not
        explicitly mentioned in the spec. We should add this to an
        errata item or an implementer guide.

        

        Best regards,

        Erik

        

        On 2013-03-04 01:00, rich levinson wrote:

      

      
        
        To TC:

        

        The "issue" here is that it does not appear to me that it
        specifies

        anywhere that a single PDP authorization decision is based on

        a set of <Attributes> element each with a different
        Category and

        that there can be at most one <Attributes> element of a
        specific

        Category involved in a single decision.

        

        I have been working under the assumption that this is true for

        quite a while now, but recently I have tried to find the
        supporting

        documentation to unambiguously validate that claim, and I have

        been unable to do so, leaving me with the conclusion that either

        the claim is wrong and that more than one <Attributes>
        element

        of the same Category can be involved in a single decision (which

        I believe to be false), or that the documentation should be
        improved

        somewhere to emphasis this fundamental point.

        

        The description in the 3.0 spec says:

        "<Attributes> [One to Many]

          Specifies information about attributes of the request context
          by

          listing a sequence of <Attribute> elements associated
          with an

          attribute category.

          One or more <Attributes> elements are allowed.

          Different <Attributes> elements with different
          categories are used

          to represent information about the subject, resource, action,

          environment or other categories of the access request."

        
        While the text above might lead one to believe that one might

        want to consider only including one <Attributes> element

        per Category, it says nothing definitive one way or the other.

        

        I am well aware that the multi-decision profile uses multiple

        <Attributes> elements of the same Category as the basis
        for

        generating a cross-product to generate "Individual Decision

        Requests", where in section 2.3.3 it states:

        "For each combination of repeated <Attributes>
          elements,

          one Individual Decision Request SHALL be created.

          This Individual Request SHALL be identical to the original

          request context with one exception:

          only one <Attributes> element

             of each repeated category

             SHALL be present."

          
        
        Again, this text is useful wrt providing the ContextHandler w
        the

        info it needs to generate multiple decision requests, however,

        it does not imply that this is a requirement from the PDP side.

        

        For example, what would happen if someone submitted a

        Request with two <Attributes> elements, each for a
        different

        Target resource, but both with the same Category, namely:

        "urn:oasis:names:tc:xacml:3.0:attribute-category:resource"?

        
        Would this be a request for which the user was asking for access

        to both resources at the same time? Or would this be an error

        if the multi-profile was not supported?

        

        I don't believe questions like this can be answered w the info
        in

        the specs as they currently exist. (unless, of course, I have
        missed

        something, which is entirely possible :) )

        

            Thanks,

            Rich