RE: [ws-rx] Gil's proposal for i021

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

Umit,

Both message- and operation-level subject
requires that the RMS have 
knowledge of the WSDL, for starters,
and that it parse each message to match
it to the WSDL message/operation to
determine whether, or not, to apply RM.

While some future, sooper-dooper all-knowing,
all-singing-and-dancing
policy engine of the future might be
capable of this, I am no longer 
of the opinion that we need to be overly
concerned about this level of 
granularity at present.

I would much prefer that we get the
simple semantic of on/off at the
endpoint level addressed, as that is
far more likely to be consistent
with implementations that deploy over
the next two years or so and will
likely satisfy a significant majority
of use cases.

It troubles me deeply that we are expending
so much of the TC's energies 
around this and related issues. Not
only that, but it seems to me that 
as soon as we hit a policy-related issue,
we go round and round in circles and
rarely come to any consensus agreement
on closure. Why am I not surprised?

Let me turn the question around (again)
and ask the members of the TC, how many
of the implementations people are working
on (and one would hope
that there are implementations in development
as we speak :-) are prepared to 
be able to address message- and/or operation-level
policy in the application of 
RM?

Cheers,

Christopher Ferris
STSM, Emerging e-business Industry Architecture
email: chrisfer@us.ibm.com
blog: http://www.ibm.com/developerworks/blogs/dw_blog.jspa?blog=440
phone: +1 508 377 9295

"Yalcinalp, Umit" <umit.yalcinalp@sap.com>
wrote on 02/02/2006 09:16:36 PM:

>  
> 
> > -----Original Message-----
> > From: Marc Goodner [mailto:mgoodner@microsoft.com] 
> > Sent: Thursday, Feb 02, 2006 12:49 PM
> > To: Anish Karmarkar
> > Cc: Gilbert Pilz; Patil, Sanjay; ws-rx@lists.oasis-open.org
> > Subject: RE: [ws-rx] Gil's proposal for i021
> > 
> > I don't understand how you reached your interpretation of the
text I
> > quoted.
> > 
> > Message level subject goes to far IMO. I understand we're 
> > talking about
> > reliable messaging, but why must it match the flexibility of
WSDL in
> > terms of how it is applied? I can't envision any practical application
> > to tweek it by individual messages on the endpoint.
> > 
> > So going to message level seems like it would open a lot of other
> > potential problems in terms of conditions we would need to 
> > clarify. What
> > is the benefit of making it that flexible? Who are we 
> > helping? What are
> > we enabling? I just don't see it.
> > 
> 
> Let me turn the question around. 
> 
> -- why do you think the message level subject vs operation subject
is
> harmful in solving particular case? 
> 
> -- Can you quantify the problems that you see with message policy
> subject? Please be specific. 
> 
> I am not advocating free for all Message policy subject, but it seems
to
> me that we need to do BOTH of the following to solve this problem:
> 
> (a) Allow message level subject for attachment (Granularity) 
> 
> (b) Allow the attachment to appear ONLY within the context of inbound
> messages in a binding/operation/message in a WSDL document in the
wsrmp
> text (Scoping to inbound messages in WSDL). 
> 
> This would solve i021 and it will be consistent with WS-Policy
> Attachment.  Further this does not get into the problem of what
happens
> to outbound messages. Operation level subject unfortunately covers
both
> inbound and outbound messages in an exchange and I am very weary of
> that. 
> 
> IMO, stating that simply the assertion applies to inbound messages
> without using the message Policy Subject would be a mistake. However,
> message Policy subject on its own is an overkill too. Therefore, we
need
> to combine both requirements and state them in conjunction with each
> other. 
> 
> > Marc Goodner
> > Technical Diplomat
> > Microsoft Corporation
> > Tel: (425) 703-1903
> > Blog: http://spaces.msn.com/mrgoodner/ 
> > 
> 
> I don't see this particular option as one of the proposals and I would
> like to put it on the table.  
> 
> Thanks. 
> 
> --umit
> 
> > 
> > -----Original Message-----
> > From: Anish Karmarkar [mailto:Anish.Karmarkar@oracle.com] 
> > Sent: Thursday, February 02, 2006 12:05 PM
> > To: Marc Goodner
> > Cc: Gilbert Pilz; Patil, Sanjay; ws-rx@lists.oasis-open.org
> > Subject: Re: [ws-rx] Gil's proposal for i021
> > 
> > Marc Goodner wrote:
> > > Just expanding on the subject level, from section 4.1.2
on Endpoint 
> > > Policy Subject in WS-PolicyAttachment [1]:
> > > 
> > >  
> > > 
> > > "An Endpoint Policy Subject applies to behaviours 
> > associated with an 
> > > entire endpoint of
> > > 
> > > the service, irrespective of any message exchange made."
> > >
> > 
> > Not a policy expert, but my interpretation is that I don't think
that 
> > precludes us from defining assertions that apply to only in 
> > messages or 
> > out-message. WS-PolicyAttachment framework defines a 
> > framework, but the 
> > assertion definer defines the semantics of the assertion. I.e.,
it 
> > should be ok to say that my BAZ assertion means that only in-bound
> > messages are reliable -- and that applies to the entire endpoint
> > (irrespective of the message exchanges or operations).
> > 
> > Having said that. I don't think Gil's proposal goes far 
> > enough. Yes, it 
> > works quite well for the most common case (one-way, 
> > request-response). 
> > But when we go to WSDL 2.0, this looks rather short-sighted.
> > It seems to
> > 
> > me that the policy assertion subject should be 'message' -- which
> > provides the granularity that provides the most flexibility.
> > After all 
> > we are talking about reliable messaging.
> > 
> > My $.02
> > 
> > -Anish
> > --
> > 
> > >  
> > > 
> > > To me that seems to say we can't declare that it doesn't
apply to 
> > > inbound or outbound messages that are part of the endpoint.
> > If that is
> > 
> > > what you want you would use operation level subject. I still
think 
> > > endpoint subject is what you would almost always want, but
> > we've been 
> > > talking about this long enough that I can see a case for
operation.
> > > 
> > >  
> > > 
> > > 1 http://specs.xmlsoap.org/ws/2004/09/policy/ws-policyattachment.pdf
> > > 
> > >  
> > > 
> > > Marc Goodner
> > > 
> > > Technical Diplomat
> > > 
> > > Microsoft Corporation
> > > 
> > > Tel: (425) 703-1903
> > > 
> > > Blog: http://spaces.msn.com/mrgoodner/
> > > 
> > >
> > --------------------------------------------------------------
> > ----------
> > > 
> > > *From:* Gilbert Pilz [mailto:Gilbert.Pilz@bea.com]
> > > *Sent:* Wednesday, February 01, 2006 2:35 PM
> > > *To:* Marc Goodner; Patil, Sanjay; ws-rx@lists.oasis-open.org
> > > *Subject:* RE: [ws-rx] Gil's proposal for i021
> > > 
> > >  
> > > 
> > > Comments in line . . .
> > > 
> > >  
> > > 
> > > - gp
> > > 
> > >      
> > > 
> > >
> > --------------------------------------------------------------
> > ----------
> > > 
> > >     *From:* Marc Goodner [mailto:mgoodner@microsoft.com]
> > >     *Sent:* Wednesday, February 01, 2006 1:10
PM
> > >     *To:* Patil, Sanjay; Gilbert Pilz; ws-rx@lists.oasis-open.org
> > >     *Subject:* RE: [ws-rx] Gil's proposal for
i021
> > > 
> > >     Two immediate concerns I have here.
> > > 
> > >      
> > > 
> > >     One, I don't think it is right to say the
AS and AD have to use
> > RM.
> > >     It is the RMS and RMD that use RM so I think
the 
> > original text is
> > >     correct. I also don't see how this relates
to the multiple
> > endpoints
> > >     via one RMS/RMD as you say below. In that
case isn't 
> > the original
> > >     text still more accurate? Still either way
this is probably
> > >     splitting hairs, it doesn't change that the
assertion 
> > is there or
> > >     what it means. 
> > > 
> > >      
> > > 
> > >     The RMS and RMD don't **use** RM they **implement**
RM. They are
> > the
> > >     things that implement the protocol described
in the WS-RM spec.
> > When
> > >     you are talking about describing policy in
something 
> > like WSDL you
> > >     are indicating if/how you will make use of
the 
> > facilities that the
> > >     RMS and RMD provide.
> > > 
> > >      
> > > 
> > >     Two, I don't think even with two different
assertions we can
> > define
> > >     them to apply just to inbound or outbound
messages. I'm fairly
> > >     certain that WSDL 1.1 and WS-PolicyAttachment
prevent that
> > scoping. 
> > > 
> > >      
> > > 
> > >     I'm not sure I understand why this would be
the case. Can you
> > expand? 
> > > 
> > >      
> > > 
> > >     Marc Goodner
> > > 
> > >     Technical Diplomat
> > > 
> > >     Microsoft Corporation
> > > 
> > >     Tel: (425) 703-1903
> > > 
> > >     Blog: http://spaces.msn.com/mrgoodner/
> > > 
> > >
> > --------------------------------------------------------------
> > ----------
> > > 
> > >     *From:* Patil, Sanjay [mailto:sanjay.patil@sap.com]
> > >     *Sent:* Wednesday, February 01, 2006 7:27
AM
> > >     *To:* Gilbert Pilz; ws-rx@lists.oasis-open.org
> > >     *Subject:* [ws-rx] Gil's proposal for i021
> > > 
> > >      
> > > 
> > >      
> > > 
> > >     Thanks to Gil for making a concrete proposal.
Hopefully 
> > this leads
> > >     to a discussion on the list.
> > > 
> > >      
> > > 
> > >     I have changed the subject line to indicate
the issue number. I
> > >     think folks have better memory of issue numbers
than AI 
> > numbers. 
> > > 
> > >      
> > > 
> > >     Thanks,
> > > 
> > >     Sanjay
> > > 
> > >          
> > > 
> > >
> > --------------------------------------------------------------
> > ----------
> > > 
> > >         *From:* Gilbert Pilz [mailto:Gilbert.Pilz@bea.com]
> > >         *Sent:* Tuesday, Jan 31, 2006
23:01 PM
> > >         *To:* ws-rx@lists.oasis-open.org
> > >         *Subject:* [ws-rx] RE: Action
Item #0078
> > > 
> > >         In light of the (pending) resolutions
to i086 and i087 it
> > seemed
> > >         simpler to present my proposal
in the form of 
> > complete drafts
> > of
> > >         the WS-RM Policy specification.
Attached are two PDF
> > documents.
> > >         One is a clean version of what
the WS-RM Policy spec would
> > look
> > >         like with my proposed changes.
The other contains 
> > change bars
> > >         between my proposal and the
current editors draft of WS-RM
> > >         Policy (cd-02 with Marc's clean
ups applied).
> > > 
> > >          
> > > 
> > >         You will note that my proposal
includes the proposed
> > resolutions
> > >         to i086 and i087. There was
no simple way to 
> > present my ideas
> > >         without doing this.
> > > 
> > >          
> > > 
> > >         You may also note that I have
changed line 93 from:
> > > 
> > >          
> > > 
> > >         "The RM policy assertion
indicates that the RM Source and RM
> > >         Destination MUST use WS-ReliableMessaging
[WS-RM
> > >         <outbind://153/#WSRM>]
to ensure reliable delivery of
> > messages."
> > > 
> > >          
> > > 
> > >         to:
> > > 
> > >          
> > > 
> > >         "In general a RM policy
assertion indicates that the
> > Application
> > >         Source and Application Destination
MUST use
> > WS-ReliableMessaging
> > >         [WS-RM <outbind://153/#WSRM>]
to ensure reliable delivery of
> > >         messages".
> > > 
> > >          
> > > 
> > >         I did this because I think that
policy assertions 
> > have nothing
> > >         to do with sequences or the
entities that maintain 
> > them (i.e.
> > >         the RMS and RMD) except that
they indicate that some,
> > >         unspecified sequence may or
must be used to ensure the
> > delivery
> > >         of inbound or outbound messages.
I think this confusion over
> > >         endpoints (and the policies
attached to those endpoints) and
> > >         sequences lay at the heart of
our difficulties with the idea
> > of
> > >         multiple endpoints with different
policies sharing the same
> > >         sequence. This change may be
considered by some to be the
> > >         resolution to a separate issue.
If anyone has any objections
> > >         I'll back it out and go through
the process of raising a
> > >         separate issue and making a
separate proposal to address it.
> > > 
> > >          
> > > 
> > >         - gp
> > > 
> >