RE: [ws-rx] [WS-RX] Issue i0024 - take 2

[Christopher B Ferris <chrisfer@us.ibm.com>]

Ashok,

I'm not convinced that we have a dilemma
at all. IMO, the authors of a policy assertion
have the prerogative to define its semantics
as they see fit. I don't see a requirement
that we need to have a WS-Policy semantic
decoration that indicates whether an 
assertion is informational or operational.

Cheers,

Christopher Ferris
STSM, Emerging e-business Industry Architecture
email: chrisfer@us.ibm.com
blog: http://webpages.charter.net/chrisfer/blog.html
phone: +1 508 377 9295

Ashok Malhotra <ashok.malhotra@oracle.com> wrote
on 10/10/2005 07:16:44 PM:

> Correcting typo:
>  
> As I see it, we are on the horns of a dilemma,
 our charter says we 
> must use WS-Policy as is, but
> WS-Policy does not let us define the semantics
of the RM (AND the 
> QoS) assertion.  We need a 
> creative solution to get around this!
>  
> All the best, Ashok 
>  
> 
> From: Ashok Malhotra [mailto:ashok.malhotra@oracle.com] 
> Sent: Monday, October 10, 2005 3:54 PM
> To: Patil, Sanjay; Yalcinalp, Umit; Christopher B Ferris
> Cc: ws-rx@lists.oasis-open.org
> Subject: RE: [ws-rx] [WS-RX] Issue i0024 - take 2

> Hi Sanjay:
> I'm not disagreeing with you but let me explain
how we got to this situation.
>  
> At the f2f someone suggested that issue 24 could
be resolved in the 
> manner we had resolved 
> another issue, I think issue 9, by appealing
to the wsp:
> Usage='Observed' in WS-Policy.
> This seemed right to me and I agreed to suggest
wording.  When I 
> wrote the wording and
> checked the spec I realized that the wsp:Usage='Observed'
had 
> somewhat different semantics than
> what we had discussed and, further, wsp:Usage
had been removed from 
> the latest version of WS-Policy.
>  
> So now, WS-Policy does not have a mechanism to
distinguish between 
> assertions that impact 
> message content, such as encryption, and assertions
that do not 
> impact message content but are
> useful as properties of the service such as Auditing,
Privacy Policy and RM.
>  
> As I see it, we are on the horns of a dilemma,
 our charter says we 
> must use WS-Policy as is but
> WS-Policy does not let us define the semantics
of the RM (at the 
> QoS) assertion.  We need a 
> creative solution to get around this!
>  
> All the best, Ashok 
>  
> 
> From: Patil, Sanjay [mailto:sanjay.patil@sap.com] 
> Sent: Monday, October 10, 2005 2:32 PM
> To: ashok.malhotra@oracle.com; Yalcinalp, Umit; Christopher B Ferris
> Cc: ws-rx@lists.oasis-open.org
> Subject: RE: [ws-rx] [WS-RX] Issue i0024 - take 2

>  
> This discussion seems to be getting into generic
policy framework 
> areas and should be carried elsewhere, IMHO.
>  
> As far as the issue i024 is concerned, my recollection
is that at 
> the Redmond F2F [1] the TC agreed to resolve this issue by adding
> *clarification text* to the specs about the meaning of the term 
> "observed" (assuming this term is used for describing the
DA 
> assertions, etc.). An AI was opened [2] specifically to track the
> needed clarification.
>  
> I am concerned that we may now be expanding the
scope of this issue 
> by suggesting normative changes (which is more than *clarification
> text*) that not just affect the WS-RX specifications but they also
> presuppose certain feature enhancements (such as wsp:Informational)
> to other specifications that are outside the scope of this TC.
>  
> May I suggest to limit the scope of this AI to
what the TC agreed 
> with, that is, add : clarification text on meaning of “observed”
in this spec.
>  
> Thanks,
> Sanjay
>  
> [1] http://www.oasis-open.org/apps/org/workgroup/ws-rx/download.
> php/14693/MinutesWSRXF2f-0905.htm#_Toc115510920
> [2] http://www.oasis-open.org/apps/org/workgroup/ws-
> rx/members/action_item.php?action_item_id=1048
>  
>  
>  
> 
> From: Ashok Malhotra [mailto:ashok.malhotra@oracle.com] 
> Sent: Monday, Oct 10, 2005 13:34 PM
> To: Yalcinalp, Umit; Christopher B Ferris
> Cc: ws-rx@lists.oasis-open.org
> Subject: RE: [ws-rx] [WS-RX] Issue i0024 - take 2

> Hi Umit:
> I think we agree that we need to distinguish
between two kinds of 
> assertions: those that impact message content and those that do not
> and merely provide information about the service.  If we agree,
then this
> distinction will need to be explained somehow.
>  
> WS-Policy only discusses assertions that impact
message content.  
> Normalization is done to cast the
> policy in a form that distinguishes between policy
alternatives that
> the client can select from or the client and server can match 
> constraints and capabilities.  Since informational assertions
do not
> participate in client
> selection or client/server policy matching they
do not have to 
> participate in normalization.
>  
> Also, I don't see how the wsp:Optional attribute
would apply to 
> informational assertions.
>  
> All the best, Ashok 
>  
> 
> From: Yalcinalp, Umit [mailto:umit.yalcinalp@sap.com] 
> Sent: Monday, October 10, 2005 11:23 AM
> To: ashok.malhotra@oracle.com; Christopher B Ferris
> Cc: ws-rx@lists.oasis-open.org
> Subject: RE: [ws-rx] [WS-RX] Issue i0024 - take 2

> I am still struggling with why we need an explicit
marker to 
> designate informational policies. Can you elaborate what you have
> written below in terms of why informational policies need to be 
> treated differently within the WS-Policy framework? 
>  
> There are two issues: recognizing sth is informational
and treating 
> it differently in the framework. 
>  
> Lets explore the first. In the WS-Policy framework,
the QName of the
> assertion is crucial to designate the type of the assertion. Thus,
> one could also infer that the RM assertion is informational from the
> QName, thus not rendering the presence of such a marker unnecessary
> for designating sth is informational or not. 
>  
> We are left with the latter problem, whether
the informational 
> policy should be treated differently by the WS-Policy framework. 
> This is where I have a bit of problem as I am having difficulty in
> understanding why "informational" policies should be treated
> differently (whether it would be for normalization, computing 
> alternatives, etc). IMO, after you compute the alternatives, the 
> assertions whether they are informational or not should apply. Even
> if a policy may not affect the message content "explicitly",
this 
> does not mean that it would not be in effect if it is declared for
a
> policy subject. If it is not possible to "quarantee" that
an 
> informational policy is in effect by just inspecting the "message
> content", this does not change the fact which assertion applies
to a
> specific set of alternates, does it? 
>  
> Therefore, I need your help to illustrate why
you regard this kind 
> of policy to be different and should be treated differently within
> the WS-Policy framework. If we don't need it to be treated 
> differently, then we can just rely on the QName trick and do not 
> need this explicit marker...
>  
> Cheers, 
>  
> --umit
>  
> 
> From: Ashok Malhotra [mailto:ashok.malhotra@oracle.com] 
> Sent: Sunday, Oct 09, 2005 2:04 PM
> To: Christopher B Ferris
> Cc: ws-rx@lists.oasis-open.org
> Subject: RE: [ws-rx] [WS-RX] Issue i0024 - take 2

> Yes, it was.  But we need some mechanism
to distinguish between 
> assertions that impact the message
> content, such as encryption, versus assertions
that merely provide 
> information about messages such as
> auditing or reliable messaging.  This is
important because the 
> former require message processing and
> possibly validating that the assertion has been
applied while the 
> latter do not.  Thus, clients and servers
> treat these two kinds of messages quite differently.
> All the best, Ashok 
>  
> 
> From: Christopher B Ferris [mailto:chrisfer@us.ibm.com] 
> Sent: Saturday, October 08, 2005 5:38 PM
> To: ashok.malhotra@oracle.com
> Cc: ws-rx@lists.oasis-open.org
> Subject: Re: [ws-rx] [WS-RX] Issue i0024 - take 2

> 
> Ashok, 
> 
> The wsp:Usage attribute was removed from the WS-Policy spec [1] when
> it was last published in Sept 2004. 
> 
> [1] http://www.ibm.
> com/developerworks/webservices/library/specification/ws-polfram/ 
> 
> Cheers, 
> 
> Christopher Ferris
> STSM, Emerging e-business Industry Architecture
> email: chrisfer@us.ibm.com
> blog: http://webpages.charter.net/chrisfer/blog.html
> phone: +1 508 377 9295 
> 
> Ashok Malhotra <ashok.malhotra@oracle.com> wrote on 10/08/2005
04:27:25 PM:
> 
> > Anish pointed out that the wording I had suggested for i0024
was in 
> > the non-normative example.
> > In the attached file I have added wording to the normative section
> > 2.2 that explains
> > the RM assertion.  The added wording says that the wsp:Usage
> > attribute must be used
> > when the assertion is included in a policy and explains the semantic
> > of this indication.  
> > I've set the value of this attribute to 'Informational' rather
than 
> > 'Observed', as 
> > we discussed, to avoid possible confusion with earlier semantics
> of Observed. 
> >   
> > The new text is in highlighted.
> > 
> > All the best, Ashok
> >  
> > 
> > [attachment "Issue24.pdf" deleted by Christopher B
Ferris/Waltham/IBM]