W3C home > Mailing lists > Public > public-ws-policy@w3.org > March 2007

Re: [NEW ISSUE] v.next: Intersection mode is neither defaulted nor specified [Issue 4292]

From: Sergey Beryozkin <sergey.beryozkin@iona.com>
Date: Thu, 15 Mar 2007 09:57:40 -0000
Message-ID: <019c01c766e8$5e9705b0$c301020a@sberyoz>
To: "Charlton Barreto" <barreto@adobe.com>, <public-ws-policy@w3.org>
Hi Charlton

>I would add that strict consumers would also fail without necessarily being able to report the precise >reason they failed. 

The reason stict consumers will fail is because they have encountered the unrecognized assertion, good tools will show to a user at design time that the intersection failed because the unrecognized assertion <foo/> was encountered or log a descriptive message at runtime, when they work in the automatic mode.
Even better tools might proceed further at design time, something Asir elaborated upon in this email :

http://lists.w3.org/Archives/Public/public-ws-policy/2006Nov/0101.html, specifically

""Requestors have the option to chose one or more modes for policy intersection (strict | lax | strict, delegate-to-user | lax, 
delegate-to-user | strict, lax, delegate-to-user | ...)""

And :
"A provider may provide requestors with out-of-band information to influence the choice of mode for policy intersection.""

Where the outbound information may be something like : this ignorable assertion is about the well-known quality of service, here the certificates from the ignorabletrust.com vouching that if you the strict client decide to ignore this assertion then you won't be adversly affected, etc... Perhaps this outbound process can be formalized in v.next once we learn more about it all, as Asir suggested earlier too.

Cheers, Sergey 

  ----- Original Message ----- 
  From: Charlton Barreto 
  To: Sergey Beryozkin ; public-ws-policy@w3.org 
  Sent: Wednesday, March 14, 2007 4:55 PM
  Subject: RE: [NEW ISSUE] v.next: Intersection mode is neither defaulted nor specified [Issue 4292] 


  Hi Sergey,

   

  The consumer can dictate the intersection mode but a provider may require that a particular mode be applied in order to consumer its messages. Since lax mode addresses compatibility for non-ignorable policy assertions, is it correct to say that to have a provider declare its assertions as ignorable is equivalent to declaring them as lax?

   

  Possibly the question to address is this: Is it appropriate for a provider to declare assertions ignorable when it really intends for lax mode to be applied? Since lax mode addresses compatibility testing for non-ignorable assertions, one can fathom the scenario where a provider intends for consumers to use the policy in this mode (test the non-ignorable assertions). 

   

  Although the end result in terms of processing may be the same, it would be useful to understand why the failure occurs. Your statement that strict consumers will unnecessarily fail if they see unrecognized assertions which are safe to be ignored definitely applies. I would add that strict consumers would also fail without necessarily being able to report the precise reason they failed. 

   

  I see the main issue being the enablement of the consumer to report why failures occur, which is a refinement of what we currently have in the framework. 

   

  -Charlton.

   


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

  From: Sergey Beryozkin [mailto:sergey.beryozkin@iona.com] 
  Sent: 14 March 2007 09:30
  To: Charlton Barreto; public-ws-policy@w3.org
  Subject: Re: [NEW ISSUE] v.next: Intersection mode is neither defaulted nor specified [Issue 4292] 

   

  Hi

   

  This is tagged as a v.next issue but I'd like to ask a clarification.

   

  > provider requires one mode

   

  Isn't it a consumer which dictates in what mode the intersection mode ? Provider are not working in the lax/strict mode, unless they may be (direct/indirect) consumers themselves, right ?

   

  For example, we've just have the issue closed where the primer advises the ignorable assertion author to be aware that the impact of the ignorable attribute at the discretion of the consumer. By introducing a wsp:ignorable assertion the author actually wishes/requires the lax mode be used for the intersection purposes but the consumer is free to ignore this wish/request and just use a strict mode.

   

  I just don't understand what a strict consumer would do it if saw this proposed mode statement ? It seems to me it would just duplicate what the provider is already trying to say by using "ignorable assertions".

   

  I propose to tackle the problem from the different angle. I believe the actual problem is that the fact that consumers can choose to be strict limits the exposure of policies which contain the ignorable assertions. 

  Or the problem can be presented like this : strict consumers will unnecessarily fail if they see unrecognized assertions which can actually are safe to be ignored.

   

  Strict mode is about failing when seeing unrecognized assertions, irrespectively of how they're marked.

  I propose for the framework to let assertion authors to assert somehow, through the security signatures, etc, that the given ignorable assertion is safe to ignore and update the rules for the strict mode to let such unrecognoized ignorable assertions be actually ignored in the automatic mode. 

   

  Cheers, Sergey

    ----- Original Message ----- 

    From: Charlton Barreto 

    To: public-ws-policy@w3.org 

    Sent: Wednesday, March 14, 2007 2:07 PM

    Subject: [NEW ISSUE] v.next: Intersection mode is neither defaulted nor specified [Issue 4292] 

     

    See: http://www.w3.org/Bugs/Public/show_bug.cgi?id=4292 

     

    Title: Intersection mode is neither defaulted nor specified [v.next]

    Target: v.next of Framework specification (http://dev.w3.org/cvsweb/~checkout~/2006/ws/policy/ws-policy-framework.html?content-type=text/html;%20charset=utf-8)
    25 January 2007

     

    Description: A potential problem exists with intersection mode selection. Intersection mode is not defaulted in the Framework specification, and mode indication is not specified. This leaves Framework open to implementations where a provider requires one mode and a client may or may not interpret it, and may or may not support that same mode. From the implementation perspective, a consumer will not be able to talk to a provider and not be able to report why. 

    At present, a consumer can only detect that a service request has failed when intersection cannot be performed. Intersection rules as specified in Framework do not provide for a standard way of reflecting why intersection could not be performed - intersection can fail for reasons other than the policies incompatibility as driven by mode (e.g. a policy not being complete when intersection is attempted). Intersection may also fail due to poorly written policies which would not be compatible regardless of mode.

     

    Consider:

    Scenario 1:

                Entity A: Service policy requires that any intersections apply lax mode

                Entity B: Service policy requires that any intersections apply strict mode

     

    In the current Framework spec

                1. Entity B attempts to use Entity A's service. Entity A's required lax mode conflicts with Entity B's required strict mode and Entity B's request will fail.

                2. Entity B reports that the request failed, but cannot report why.

     

    For example, a provider of a service which handles LOB processing can provide a number of QoS which are exposed as policies. A consumer can effectively navigate which QoS (service interface) to use through its requirements. However, a consumer of an LOB processing service may be required to always apply either its own assertions or always accept the provider's base (e.g. always accepting the provider's base assertions as a way of certifying/validating it's interaction with the service). 

     

    Scenario 2:

                Entity A: Service policy requires that any intersections apply strict mode

                Entity B: Service policy accepts strict intersection but, but the entity doesn't implement strict

     

    In the current Framework spec

                1. Entity B attempts to use Entity A's service, but has not implemented handling of strict mode intersection and this fails in its request.

                2. Entity B reports that the request failed, but cannot report why. 

     

    For example, a provider has a service which, while ignorable, prioritises its policies (strict) - for example, an internet banking service which always requires its security and reliability policies to the exclusion of a consumer's requirements. However, certain consumers of this service type may be required to use their own security policies - they may not have implemented the capacity to use the provider's service with the provider's required security policy (thus supporting only lax). 

     

    Either scenario highlights the fact that strict and lax mode, as ignorable is currently specified, can be used exclusively by either consumer or provider. As such, WS-Policy needs to address cases where they are used as such.

                

    Proposal: Specify the indication of intersection mode in the Framework spec and specify that the default intersection mode is strict.

     

    Spec text to add to Framework:

                The wsp:IntersectionMode element indicates the mode required by a policy assertion for intersection. The default intersection mode is strict. 

     

                <wsp:IntersectionMode>

       <anyElementQName>

          standard-elements

       </anyElementQName>

    </wsp:IntersectionMode>

     

    The schema outline for this element is as follows:

     

                <wsp:Policy ... > 

      ...

      <xsd:element ref="IntersectionMode" />

      ...

                  <xsd:element name="IntersectionMode" type="tIntersectionMode" />

      <xsd:complexType name="tIntersectionMode">

        <xsd:sequence>

          <xsd:any namespace="##other" processContents="lax" />

        </xsd:sequence>

      </xsd:complexType>

      ... 

    </wsp:Policy>

     

                The following describes the Element Information Item defined in the schema outline above:

     

                /wsp:Policy/wsp:IntersectionMode

     

                This attribute is of type wsp:tIntersectionMode. If the actual value (See XML Schema Part 1 [XML Schema Structures]) is "strict", the policy's supported intersection mode is strict. If the actual value is "lax", the policy's supported intersection mode is "lax". If the actual values are "strict" and "lax", the policy's supported intersection modes are "strict" and "lax". Omitting this attribute is semantically equivalent to including it with a value of "strict".

     

    Note that the proposed wsp:IntersectionMode provides for extensibility. As additional intersection modes are defined and implemented, this element can be used to specify their use.        

                                        

    In the first scenario above, if Entity B could reliably consume the intersection mode emitted by Entity A in its policy, it could interpret that A requires lax; in the second, if Entity B could reliably consume the intersection mode emitted by Entity A in its policy, it could interpret that A requires strict. In either case, with the changes in this proposal, Entity B can:

                1. Not fail and perform some other logic to manager the situation

                2. Fail and report the actual cause of failure

     

    Best,

     

    -Charlton.

     

     


               Charlton Barreto
                Senior Computer Scientist
                Adobe Systems Incorporated
                345 Park Avenue, MS E15
                San Jose, CA 95110-2704 USA
                408.536.4496 p,   415.692.5396 m
                charltonb@adobe.com 
               

         

     

     




image001.gif
(image/gif attachment: image001.gif)

Received on Thursday, 15 March 2007 09:56:15 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:20:48 GMT