OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

ebxml-bp message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: Re: [ebxml-bp] State Alignment and Web Services


Title: Message
John,
 
Picking up from here:
I think we should specify the signals, and then allow binding to various implementations of those signals (for instance RM can provide receipt ack, and acceptance can be part of response)
 
That is what I was suggesting to JJ - that a prototype model would give
us some context here - since you can potentially use the signals as part
of the BT definition - in lieu of say "confirmBOD" - and then its
perfectly clear what the business function of the signal is.
 
If I have some signals examples - I can add a swim lane to the model to
allow people to see those definitions - and then in the BT part of the model
UI - you can drag and drop a response that uses signals and then use
the pulldown list of available signals to select the one you want.
 
Should be quite simple.  I can colour code the signal action to differentiate
it from a regular transaction exchange.
 
Underneath the hood in the model - we can write all the correct
bits (hopefully!) in the BPSS instance to make this work for the user....
 
I'm just speculating here - and presumably there's some QoS settings
in the exchange profile (aka document-envelope) that may apply
here so that the responder knows that they have to do the signals....
 
Any - I'm interested in creating a second sample model that uses
signals...
 
Thanks, DW
 
----- Original Message -----
Sent: Thursday, June 10, 2004 11:24 AM
Subject: RE: [ebxml-bp] State Alignment and Web Services

I'll take issue with just the RM point.  Reliable Messaging as it is normally discussed is implemented at and encapsulated within the transport layer, while business action state confirmation can be implemented at several levels (transport, message, process).
 
The physical receipt ack can be either come as a byproduct of the transport layer, or be explicitly driven by the messaging layer.  The important part is that the receipt ack is signed for non-repudiation.
 
The acceptance ack can be part of the response.  Often acceptance is contractually linked to the signed receipt ack.  (e.g. contract explicitly states "I agree to process any request within 4 days")
 
The messaging layer can provide receipt signals, but if it does not then they must be explicitly provided.
 
This point is important because of the large number of transports in use, and because of the extra communication required between messaging and process layers in order to take business action based upon the state of the collaborative exchange -> as JJ says the state alignment must be verifiable.  One of our key objectives is to keep interdependencies between layers "requirements based" instead of "reference based" to allow legacy process to participate, and to ease evolution of each layer.
 
In early RosettaNet the discussion of "what if the receipt ack / acceptance ack are dropped" led to explicit timeouts and the NotificationOfFailure transaction, if you start a transaction and it does not complete within the timeout, it is failed.
 
Reliable messaging can reduce the number of these failures, however the process layer should still retain the ability to recognize the failure and take action.
 
I think we should specify the signals, and then allow binding to various implementations of those signals (for instance RM can provide receipt ack, and acceptance can be part of response)
 
John
-----Original Message-----
From: Jean-Jacques Dubray [mailto:jeanjadu@Attachmate.com]
Sent: Wednesday, June 09, 2004 9:58 PM
To: Jacques Durand; Monica J. Martin; ebXML BP
Subject: RE: [ebxml-bp] State Alignment and Web Services

Yes, this is precisely my point. In order to ensure state alignment and offer non-repudiation.

 

1)       RM is mandatory

2)       You need the receipt Ack for non repudiation

3)       You need the acceptance Ack for confirmation that the message was process-able

 

You can only achieve that with signals and not messages. This is because signals have a fixed structure and are handled by the BPSS infrastructure so no one can say “I did not understand the signal”. Once you got a signal, you can choose to ignore it, but you can deny that you got it or could not understand it. Signals avoid “acking the acks”.

 

Without such a scheme in place you can never achieve the appropriate level of state alignment in all cases.

 

To correct what I have said in my previous email, Systinet and Apache just released a few days ago support for WS-RM (but I am not sure if it is a published spec or not).

 

JJ-

 


From: Jacques Durand [mailto:JDurand@us.fujitsu.com]
Sent: Wednesday, June 09, 2004 8:33 PM
To: Jean-Jacques Dubray; Monica J. Martin; ebXML BP
Subject: RE: [ebxml-bp] State Alignment and Web Services

 

This state alignment appears to be above the messaging layer
So RM alone - whether in ebXML or WS - would only help, but not suffice:
RM would (1) help deliver the message in spite of trouble, (2) but cannot even
ensure that: only guarantee that your credit card comp is notified
in case the message could not be delivered after all.
And if the message is delivered to the end-point, does not mean it has
been read indeed.

So if what the credit card company wants (or should have wanted) is a kind of non-repudiation from *you* (not just from your messaging end-point), I think this is precisely where BPSS bus transactions help.

(not sure this would require BTP though).

Jacques

-----Original Message-----
From: Jean-Jacques Dubray [mailto:jeanjadu@attachmate.com]
Sent: Wednesday, June 09, 2004 5:11 PM
To: Monica J. Martin; ebXML BP
Subject: [ebxml-bp] State Alignment and Web Services

 

Yesterday something not so funny happened to me. I talked to my credit
card company because two payments on my credit card did not happen. I
had set up recurring payments and things went without a glitch for a
couple of years.

By talking to them I realized that they had sent me an email in April
telling me that they were discontinuing on-line statements and in this
email supposedly they were asking me to sign up for on-line payment one
more time.

As a result I ended with a bill of over $100 of finance charges. Of
course they promptly reverted these charges when I explain that I did
not think it was a good way to do business (it is close to a scam if you
are me).

So here is precisely what happens when state alignment is not
guaranteed. They sent a message in the hope that I would understand it
perfectly. They did not expect me to send an ack, nor did I got the
information that I HAD to send an ack (signaling an important message
for instance).

SMTP has some kind of Reliable Messaging capability. Their message would
have bounced back if my email address was not valid anymore, but SMTP
cannot tell them that I actually read that email.

The conclusion of this story is that misalignment of state for
commitments is a really really really bad idea, it is very costly to
rewind (I spend more than an hour on the phone to solve this). Imagine,
I send you a PO request for a widget, and you send me an ack but I never
receive it or I misinterpret it. I go off and buy the widget from
another supplier. Now I receive two widgets. How do we sort that out. Is
it preferable to avoid being in this situation?

Web Services does not have yet an agreed upon WS-RM spec (which is just
one part of the problem), I don't know products that support it (e.g.
Microsoft just shipped WSE 2.0 without it and probably will have to wait
until Indigo comes out, I heard a comment that this would be really hard
to support in WSE).

By contract, ebXML and BPSS supports state alignment (aka business
transaction) today which rely on ebXML RM and ebXML BTP.

JJ-



-----Original Message-----
From: Monica J. Martin [mailto:Monica.Martin@Sun.COM]
Sent: Tuesday, June 08, 2004 6:53 AM
To: ebXML BP
Subject: [ebxml-bp] [ebBP] 6/8/2004: WI-12 WSDL Support - Preparation
for Quorate Vote

Discussion|OASIS.ebBP.WI12-WSDL Support;
Topic|;
Point|Preparation for v2.0 Vote Opening 14 June 2004;
Attachment|http://www.oasis-open.org/archives/ebxml-bp/200406/msg00053.h
tml;
Attachment|http://www.oasis-open.org/apps/org/workgroup/ebxml-bp/downloa
d.php/7102/ebbp-mtgminutes-mm1-060704.txt;

mm1@
Yesterday, we had a productive discussion with many good ideas about how

to iteratively approach the Operation 'thingy'. As discussed, we've had
two detailed proposals from Anders Tell and JJ Dubray, both of which I
believe can be supported. .  Kenji has also provided some insight that
MEP may typically exist below the business transaction constructs (and
could be related to DocumentEnvelope). So, we have identified three
important criteria for this capability:

    * Support the guiding principles from other sources such as
      UNCITRAL, other UN legal documents and the ebXML eCommerce
      Patterns (v1.0) [3]. We do anticipate we will be adding more
      support in a later version. So, this is our first step to lay the
      groundwork. Ensure that dispatch-reach requirements are understood
      and considered here.
    * Provide the capability to support an abstract web service
      reference (Operation 'thingy') [1]. Ensure we can support
      monitoring and existing capabilities given this new function -
      statuses, conditions, transitions, roles, etc.
    * Ensure the technical specification clearly identifies that there
      are at least two distinct use case areas for these services:
      Provide a basis a business agreement and support of an overall
      exchange. With the latter, we should ensure it is clearly defined
      and usage is controlled appropriately. As currently understood,
      ebBP constructs or web services will be used but in separate
      collaborations.
          o Specification should indicate that the web service should
            not used to initiate or fulfill/discharge a business
            commitment [2].
          o Use web services to provide state alignment where parties
            can't use a robust capability such as those defined in
            existing business transaction patterns [2].
          o Provide constraints in the form of requirements for CPPA for
            v2.1 errata (which will support web services as well).
          o For v3.0, create a technical note that addresses how use of
            web services is accomplished and with more research how that
            occurs in the context of a constrained business transaction
            pattern. Further definition defined by the team. In the
            interim, contact IV&I to see if they could be the test case
            implementation for this capability.
                + Extend or continue to develop the functionality to
                  meet the core requirements identified in v2.0 and
                  others that are identified in the interim.

***The vote will open 14 June 2004 (8 a.m. EDT) and close 21 June 2004
(11 p.m. PDT).***

[1] This will be further refined given 7 June 2004 teleconference,
submission by Nagahashi, and collaboration with Tell, Dubray, Yunker and

Moberg.
[2] Yunker, 7 June 2004

Summary sent 7 June
2004:http://www.oasis-open.org/archives/ebxml-bp/200406/msg00053.html
Meeting minutes 7 June 2004:
http://www.oasis-open.org/apps/org/workgroup/ebxml-bp/download.php/7102/
ebbp-mtgminutes-mm1-060704.txt

***KENJI, GIVEN YESTERDAY'S CALL AND THE INPUTS THUS FAR REGARDING YOUR
SUGGESTION ON THE DOCUMENT ENVELOPE, YOUR INPUT IS REQUIRED. THANKS!***
@mm1

 



[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]