Re: Are nested assertions part of the policy vocabulary?

Hi

What confuses me is that it appears to be some inconsistency in the definition of what the empty nested Policy means in the scope of ws-addressing (see [1]), <ws-addressing><Policy/></ws-addressing> and the fact that this nested <Policy> does not intersect with a more qualified nested Policy such as <ws-addressing><Policy><AnonymousResponses/></Policy></ws-addressing>.

According to [1] If the provider uses <Policy/> then it means this provider will work with consumers using either anonymous or non-anonymous WSA qualifications. And yet, the requester saying, by using <ws-addressing><Policy><AnonymousResponses/></Policy></ws-addressing>, that it wishes a provider to support  <AnonymousResponses/> will fail to intersect with the provider using <Policy/> which says that all types of responses are supported...

I think what this means is using an all inclusive <Policy> alternative alone on the server is just not safe as it will cause compliant clients (say those wishing a provider to support <AnonymousResponses/>) to break...I'd even say that the empty nested ws-adressing <Policy> should be prohibited...Just have two nested policies, one allowing anonymous responses, another one allowing non-anonymous ones...That way a provider supporting all types of responses can list two alternatives and it will match all clients....


Thanks, Sergey

[1] http://lists.w3.org/Archives/Public/public-ws-policy/2007Apr/att-0094/WSAddrPolicyAlgerntiveGprime.htm
  ----- Original Message ----- 
  From: Bob Freund 
  To: Daniel Roth ; Ashok Malhotra ; Maryann Hondo ; Anthony Nadalin 
  Cc: public-ws-addressing@w3.org ; public-ws-addressing-request@w3.org ; public-ws-policy@w3.org ; public-ws-policy-request@w3.org 
  Sent: Tuesday, May 08, 2007 12:22 AM
  Subject: RE: Are nested assertions part of the policy vocabulary?


  +1

  From a plain reading of the WS-Policy intersection algorithm, these policies indeed are not compatible per the WS-Policy 1.5 framework CR spec.

  -bob

   

  From: public-ws-addressing-request@w3.org [mailto:public-ws-addressing-request@w3.org] On Behalf Of Daniel Roth
  Sent: Tuesday, May 01, 2007 4:52 PM
  To: Ashok Malhotra; Maryann Hondo; Anthony Nadalin
  Cc: public-ws-addressing@w3.org; public-ws-addressing-request@w3.org; public-ws-policy@w3.org; public-ws-policy-request@w3.org
  Subject: RE: Are nested assertions part of the policy vocabulary?

   

  Hi Ashok, 

   

  These two policies do not intersect and we believe this is verified in the test cases.  If Policy 2 is the policy for a requester then this intersection result may at first seem incorrect, so let me explain: 

   

  It is incumbent on the Addressing authors to specify the semantics of the assertions.  The Addressing assertion expresses a requirement that WS-Addressing be used to exchange messages without qualifications.  The nested addressing assertions (which indicate additional characteristics of  the base WS-Addressing assertion)  qualify this semantic to say that either only non-anonymous responses are supported or that only anonymous responses are supported.  In the WS-Addressing protocol it's the requester's message that requests a specific kind of response - anonymous, non-anonymous, or maybe even a mixture of the two.   

   

  The thing to recognize is that if Policy 2 is a requester policy then it is incomplete in that it is not acknowledging that the base assertion also reflects support for anonymous responses.  The requester determines what response type should be used.  So, a client that needs non-anonymous responses will also work with a service that supports all of addressing.   The client's policy should reflect that  it is compatible with an endpoint that supports all of addressing by adding a second alternative.  This can be easily done using the Optional attribute as is shown in section 3.1.6 in the WS-Addressing Metadata spec: 

                <Policy><Addressing ><Policy><AnonymousResponses wsp:Optional="true" > </Policy></Addressing ></Policy> 

   

  Note that if Policy 2 is a provider policy and Policy 1 is the requester policy - where the requester wants unqualified support for addressing, but the provider only supports a specific response type - then there is no issue.  These policies should not intersect and they don't. 

   

  We hope this helps. 

   

  Daniel Roth

   

  From: public-ws-policy-request@w3.org [mailto:public-ws-policy-request@w3.org] On Behalf Of Ashok Malhotra
  Sent: Thursday, April 19, 2007 12:33 PM
  To: Maryann Hondo; Anthony Nadalin
  Cc: public-ws-addressing@w3.org; public-ws-addressing-request@w3.org; public-ws-policy@w3.org; public-ws-policy-request@w3.org
  Subject: RE: Are nested assertions part of the policy vocabulary?

   

  Hi Maryann:

  Perhaps I misunderstood.  Let me rephrase my comments as questions.

   

  Since Policy 1 

  <Policy>
  <ws-addressing>
  <Policy/>
  </ws-addressing>
  </Policy> 
  was intended to mean that ALL options ( anonymous, non-anonymous) are supported. 

   

  And Policy 2

  <Policy>
  <ws-addressing>
  <Policy>
  <ws-addressing: Anonymous> 
  </Policy>
  </ws-addressing>
  </Policy> 
  was intended to mean that  ONLY anonymous was supported. 

  Should Policy 1 match Policy 2 in the intersection algorithm?

  All the best, Ashok 


------------------------------------------------------------------------------

  From: public-ws-addressing-request@w3.org [mailto:public-ws-addressing-request@w3.org] On Behalf Of Maryann Hondo
  Sent: Thursday, April 19, 2007 5:55 AM
  To: Ashok Malhotra; Anthony Nadalin
  Cc: Ashok Malhotra; public-ws-addressing@w3.org; public-ws-addressing-request@w3.org; public-ws-policy@w3.org; public-ws-policy-request@w3.org
  Subject: RE: Are nested assertions part of the policy vocabulary?

   


  Ashok, 
  I would like to clarify my comments. 

  I was trying to say, that the WS-Addressing group seemed to be trying to use nested assertions to indicate a "constraint". 
  My understanding of the semantics are the following: 
  <Policy>
  <ws-addressing>
  <Policy/>
  </ws-addressing>
  </Policy> 
  was intended to mean that ALL options ( anonymous, non-anonymous) are supported. 

  and 
  <Policy>
  <ws-addressing>
  <Policy>
  <ws-addressing: Anonymous> 
  </Policy>
  </ws-addressing>
  </Policy> 
  was intended to mean that  ONLY anonymous was supported. 

  This to me, ths meant that the "intent" of the base assertion was being "constrained" by the presence off a nested assertion 
  and that was ok if the working group understood the semantics they were expressing ( i.e. the "absence" of a nested assertion 
  means "no constraints" or "all options are supported") 

   and I thought the language of the ws-policy spec allowed this interpretation since  a nested assertion could be seen to be 
  qualifying the base assertion with a constraint rather than a capability. 

  Authors MAY define that an assertion contains a policy expression (as defined in 4. Policy Expression) as one of its [children]. Nested policy expression(s) are used by authors to further qualify one or more specific aspects of the original assertion. 


  The spec already says the following so I don't think alternative 1 really adds anything, unless I'm missing something, like Tony, I need more of an explanation of what you are suggesting you want the intersection to do: 

  Because the set of behaviors indicated by a policy alternative depends on the domain-specific semantics of the collected assertions, determining whether two policy alternatives are compatible generally involves domain-specific processing. 

  Maryann 

        Anthony Nadalin/Austin/IBM@IBMUS 
        Sent by: public-ws-addressing-request@w3.org 

        04/19/2007 03:41 AM 
       To
             "Ashok Malhotra" <ashok.malhotra@oracle.com> 
             
              cc
             "public-ws-addressing@w3.org" <public-ws-addressing@w3.org>, "public-ws-policy@w3.org" <public-ws-policy@w3.org>, public-ws-policy-request@w3.org 
             
              Subject
             RE: Are nested assertions part of the policy vocabulary?
             

         

               
              
             

       




  #1 " dependent on the semantics of the parent assertion." not sure what this would mean can you give some guidance here ?
  #2 is real dangerous as you have no idea what you are matching on, one day it could be XYZ and another day it could be ABC.

  Anthony Nadalin | Work 512.838.0085 | Cell 512.289.4122
  "Ashok Malhotra" <ashok.malhotra@oracle.com>

        "Ashok Malhotra" <ashok.malhotra@oracle.com> 
        Sent by: public-ws-policy-request@w3.org 

        04/16/2007 04:23 PM 
        




              To
             
              "public-ws-policy@w3.org" <public-ws-policy@w3.org>, "public-ws-addressing@w3.org" <public-ws-addressing@w3.org> 
             



              cc
             
             



              Subject
             
              RE: Are nested assertions part of the policy vocabulary?
             

         


             
             

       



  I'm at the OASIS Symposium and have had extensive discussions with the WS-Addressing folks re. the problems they are having in using WS-Policy to express their requirements.

  As I see it, the sticky usecase is where the provider wants to say this it supports WS-Addressing in all its manifestations and the requester specifies that it supports a particular variation of WS-Addressing. These two policies must match. Thus, the provider says:

  <Policy>
  <ws-addressing>
  <Policy/>
  </ws-addressing>
  </Policy>

  And the requester says:

  <Policy>
  <ws-addressing>
  <Policy>
  <ws-addressing-specific-assertions> 
  </Policy>
  </ws-addressing>
  </Policy>

  These two policies must match in the intersection algorithm. The text that prevents them from matching says:

  "If either assertion contains a nested policy expression, the two assertions are compatible if they both have a nested policy expression and the alternative in the nested policy expression of one is compatible with the alternative in the nested policy expression of the other."

  In the note below (which Glen +1ed), Maryann suggests that a Policy with just the <ws-addressing> assertion is expressing a constraint which can be met in several ways - at least that's how I read her note. She does not, however, suggest concrete wording. Here are a couple of suggestions: 
  1. Change the quoted text above to say that matching of nested policy assertions is dependent on the semantics of the parent assertion. This way, WS-Addressing could define its own semantics for matching and solving their usecase.
  2. Bob Freund suggested a wildcard assertion that could be included within a nested Policy that would match any other nested policy. 

  All the best, Ashok 

   


------------------------------------------------------------------------------


  From: Maryann Hondo [mailto:mhondo@us.ibm.com] 
  Sent: Wednesday, April 04, 2007 7:37 AM
  To: Glen Daniels
  Cc: Ashok Malhotra; Monica J. Martin; public-ws-policy@w3.org; public-ws-policy-request@w3.org
  Subject: RE: Are nested assertions part of the policy vocabulary?


  Glen, 
  I think the problem is that the assertions are really trying to express a constraint .....and should be something 
  like "nonAnonymousONLY". so the absence, is not the absence of support but rather the absence of the constraint. 

  And in this case I think the " no constraints" is sufficient for your use case 
  The client has no constraints on what the provider will do. 
  That should intersect with all the provider options. 

  I hope we can talk this through on the call. 
  Maryann 

        "Glen Daniels" <gdaniels@progress.com> 
        Sent by: public-ws-policy-request@w3.org 

        04/04/2007 09:59 AM 
        

              To
             "Monica J. Martin" <Monica.Martin@Sun.COM>, "Ashok Malhotra" <ashok.malhotra@oracle.com> 
             
              cc
             <public-ws-policy@w3.org> 
             
              Subject
             RE: Are nested assertions part of the policy vocabulary?
             

         

               
              
             

       





  Hi Monica:

  I'm a little confused here. Are you and MaryAnn indeed saying that
  selecting the first alternative in Ashok's (and indeed WS-Addressing's)
  example means that neither anonymous nor non-anonymous responses are
  allowed? That certainly isn't the goal of the policy, and indeed this
  interpretation would seem to disallow ANY kind of response.

  How would you write a consumer policy which was meant to successfully
  intersect with endpoint policies which either a) express nothing about
  anonymous responses, b) express a requirement for anonymous responses,
  or c) express a requirement for non-anonymous responses?

  --Glen

  > -----Original Message-----
  > From: public-ws-policy-request@w3.org 
  > [mailto:public-ws-policy-request@w3.org] On Behalf Of Monica J. Martin
  > Sent: Tuesday, April 03, 2007 5:30 PM
  > To: Ashok Malhotra
  > Cc: public-ws-policy@w3.org
  > Subject: Re: Are nested assertions part of the policy vocabulary?
  > 
  > 
  > 
  > hondo: Ashok,
  > My response is yes.
  > Maryann
  > 
  > >>mm1: Ashok, agree with MaryAnn on question one answer - this point 
  > has been made that the nested assertions are part of the policy 
  > vocabulary. Yet, an important point associated with this surrounds 
  > whether or not the guiding conformance [1] requires support for those 
  > response types - that provides substance on your second 
  > question and its 
  > disposition.. [2]
  > 
  > We also state in Section 3.2 Framework before the statement you cite:
  > 
  > An alternative with zero assertions indicates no behaviors. An
  > alternative with one or more assertions indicates 
  > behaviors implied
  > by those, and only those assertions.
  > 
  > Remember: (no position just stating the action-result), we augmented 
  > this text in 
  > http://www.w3.org/Bugs/Public/show_bug.cgi?id=3602 Issue 3602.
  > 
  > [1] WS-A specification(s) referenced
  > [2] Related to empty and the base assumptions of WS-Addressing.
  > 
  > >Ashok Malhotra wrote: Section 3.2 of Framework says "When an 
  > assertion whose type is part of the policy's vocabulary is 
  > not included in a policy alternative, the policy alternative 
  > without the assertion type indicates that the assertion will 
  > not be applied in the context of the attached policy 
  > subject." Are nested assertions included in the policy's 
  > vocabulary?
  > >
  > >Consider the following example:
  > >
  > > <wsp:ExactlyOne>
  > > <wsp:All>
  > > <wsam:Addressing> <-- supports all response 
  > types --> Alternative 1
  > > <wsp:Policy> 
  > > </wsp:Policy>
  > > </wsam:Addressing>
  > > </wsp:All>
  > > <wsp:All>
  > > <wsam:Addressing> <-- requires Anonymous 
  > responses --> Alternative 2
  > > <wsp:Policy>
  > > <AnonymousResponses />
  > > </wsp:Policy>
  > > </wsam:Addressing>
  > > </wsp:All>
  > > <wsp:All>
  > > <wsam:Addressing> <- requires nonAnonymous 
  > responses --> Alternative 3
  > > <wsp:Policy>
  > > <NonAnonymousResponses />
  > > </wsp:Policy>
  > > </wsam:Addressing>
  > > </wsp:All>
  > > </wsp:ExactlyOne>
  > ></wsp:Policy>
  > >
  > >If Alternative 1 is selected, does this mean that neither 
  > Anonymous responses nor NonAnonymous responses are allowed as 
  > both are part of the policy vocabulary but not included in 
  > the alternative.
  > >
  > >All the best, Ashok
  > >
  > > 
  > >
  > 
  > 
  > 
  > 

Received on Tuesday, 8 May 2007 09:24:21 UTC