Re: Trading Partner Logical Identification based on EDIFA

[Nicholas Kassem <Nick.Kassem@eng.sun.com>]

At 01:33 PM 8/18/2000 -0700, Gordon van Huizen wrote:
>Farrukh,
>
>There was a great deal of confusion about this in San Jose, and I'm not
>sure that we all came away with the same mental picture of what's being
>addressed WRT to windows.
>
>Here are the two levels of windows that I would have been comfortable
>with:
>
>1) A low-level framing mechanism for marshaling data over the wire based
>on a given transport protocol. The rationale for this mechanism is
>efficiently and reliably propagating all traffic over a given link.
>Traffic, here, means a mix of messages (for multiple logical sending and
>receiving parties and multiple conversations, with varying delivery
>guarantees, message priorities, etc.) that happen to be sent between
>service providers on two ends of a physical link. In practice this
>framing is necessary for performance and reliability over underlying
>transport protocols (certainly for TCP/IP). The key here is that the
>windowing is between service providers that own the physical link, not
>between logical parties. Few project team members were comfortable
>addressing this level of window in the spec, although it does affect
>performance, reliable messaging and the ability to have over-the-wire
>interoperability.

Forgive me for re-hashing issues that may have already been sorted out but 
I'm struggling to understand the problem being tackled. My understanding 
was (from way back in Dallas) that the goal of reliable messaging, in the 
ebXML TRP context, was to offer a quality of service over and above what 
e.g. HTTP/SMTP offer on their own, but less than what most MOM products 
typically already offer today. The goal, presumably, is not to duplicate 
what MOM products already do in a proprietary way but rather to unburden 
applications from having to re-invent some basic notion of reliable 
messaging every time. I think it would be good to de-couple the 
fragmentation/re-assembly problem from the reliable messaging problem. IMO, 
the latter requires the definition of ebXML transport level messages (that 
are never seen at the ebXML application level - because that's the whole 
point of this exercise) and their associated choreography. The exchange of 
such messages between peer ebXML transport providers would provide some 
assurance to a message producing application entity (Business Process) as 
to the delivery (or non-delivery) state of the application payload. This 
may be good enough for most web centric business applications and we could 
also define ebXML transport level retry mechanisms, time-out's etc. Those 
applications that need an MQ class of service will continue to use the wide 
range of commercial products out there and will utilize whatever API they 
see fit.
Am I way off the mark ?


>2) An application-level message batching mechanism for application-level
>transaction management, ala JMS, where at the application level you can
>commit or roll-back a group of messages and receive notification of
>time-outs, etc. The scope of a window here is application-based. The
>group agreed that this was out of scope for the reliable messaging spec,
>indicating that the reliable messaging layer to be addressed is not an
>application-level issue.

Agreed.


>What I believe we have in the reliable messaging spec is most definitely
>not #1, and I don't believe it's #2 either. Instead I believe it's some
>middle layer that will need to be mapped in both directions.

Agreed. I think we need to be closer to #1 without going overboard and with 
an emphasis on inter-operability. The challenge is to explicitly define 
what is meant by "reliable".


>-gvh-
>
>Farrukh Najmi wrote:
> >
> > Gordon,
> >
> > I am still unclear on the second two of three suggested levels of 
> windows. Can
> > someone give a reason for why we need the additional grouping of 
> messages. Is it
> > not enough to group messages that are related to one conversation (ala JMS
> > transaction as suggested)? And why does the transport have any role to 
> play in the
> > grouping of messages? Since these additional levels of "windows" add 
> complexity at
> > the conceptual level I would like to understand the rationale behing them.
> >
> > --
> >
> > Regards,
> > Farrukh
> >
> > Gordon van Huizen wrote:
> >
> > > I know we've had a lot of discussion about this, but it seems to me that
> > > there are three levels of "windows" involved in the current model for
> > > ebXML messaging:
> > >
> > > - Conversation-level windows (more like transacted messages in JMS)
> > > - Reliable messaging windows (at a logical level, above transport
> > > bindings)
> > > - Windows for framing messages carried across a particular transport
> > >
> > > Is this what everyone else is picturing? If so, the Reliable Messaging
> > > window can be pretty simple, and then we have the issue of mapping these
> > > to the other two levels. I want to be clear on this point prior to
> > > making final comments on the reliable messaging spec.
> > >
> > > If the above separation is NOT what we're doing, the reliable message
> > > spec needs some work.
> > >
> > > -gvh-
> > >
> > > Marc Breissinger wrote:
> > > >
> > > > The Message Window (which I believe as been renamed in the latest 
> version of
> > > > the Reliable Messaging spec) is not the same as the 
> conversation.  Messages
> > > > from many conversations can be within the same window.
> > > >
> > > > 
> ==========================================================================
> > > > Marc Breissinger                                   voice (W): 
> 703-460-2504
> > > > Director, Product Strategy - webMethods, Inc.      voice (C): 
> 703-989-7689
> > > > Email:  marcb@webmethods.com                               We're 
> hiring!!!
> > > > Email2: breissim@earthlink.net              URL: 
> http://www.webmethods.com
> > > > 
> ==========================================================================
> > > >
> > > > > -----Original Message-----
> > > > > From: Farrukh Najmi [mailto:Farrukh.Najmi@east.sun.com]
> > > > > Sent: Friday, August 18, 2000 9:54 AM
> > > > > To: ebXML Transport (E-mail)
> > > > > Subject: Re: Trading Partner Logical Identification based on EDIFA
> > > > >
> > > > >
> > > > >
> > > > > The EbXML header also calls this (i.e. Message Set ID) a
> > > > > Conversation ID. I find this
> > > > > term to be the most descriptive because an ebXML based Business
> > > > > Process may be conducting
> > > > > several concurrent conversations with the same trading partner.
> > > > > Since this ID is used to
> > > > > link Messages associated with the same conversation I propose we
> > > > > stick with Conversation
> > > > > ID. This will also reduce confusion since the term is already in
> > > > > use in the ebXML header
> > > > > DTD.
> > > > >
> > > > > On a related subject, I have heard the term Message Window in the
> > > > > Reliable Messaging
> > > > > spec. Again my interpretation is that this is referring to a
> > > > > Conversation. If so we
> > > > > should align the terminology and consistently call this concept a
> > > > > Conversation that is
> > > > > identified by a Conversation ID.
> > > > >
> > > > > Is there a horizontal team that makes sure we get rapid alignment
> > > > > on common terminology
> > > > > across working groups? If so they should resolve this terminology
> > > > > issue quickly.
> > > > >
> > > > > --
> > > > >
> > > > > Regards,
> > > > > Farrukh
> > > > >
> > > > > srh@us.ibm.com wrote:
> > > > >
> > > > > > When we decided in Dallas on the term "Message Set" (my 
> suggestion) the
> > > > > > intention was
> > > > > > merely to give identification within *an undefined context at 
> the time*
> > > > > > to that conversation/unit-of-business. I happen to like the 
> term, as it
> > > > > > avoids the overloaded *transaction*.
> > > > > > There was no further semantic impled, certianly not routing. If 
> we can
> > > > > > seperate what/how issues it may help.
> > > > > >
> > > > > > Scott Hinkelman
> > > > > > Senior Software Engineer, IBM Austin
> > > > > > Emerging Technologies, SWG
> > > > > > 512-823-8097 (TL 793-8097) (Cell: 512-940-0519)
> > > > > > srh@us.ibm.com, Fax: 512-838-1074
> > > > > >
> > > > > > Martin W Sachs/Watson/IBM@IBMUS on 08/16/2000 03:15:12 PM
> > > > > >
> > > > > > To:   David Burdett <david.burdett@commerceone.com>
> > > > > > cc:   "ebXML Transport (E-mail)" <ebxml-transport@lists.ebxml.org>
> > > > > > Subject:  RE: Re[2]: Trading Partner Logical Identification
> > > > > based on EDIFA
> > > > > >
> > > > > > Dave,
> > > > > >
> > > > > > I haven't been around ebXML long enough to know what the old
> > > > > message set ID
> > > > > > is but let me go into conversation ID a little more.
> > > > > >
> > > > > > A conversation (in tpaML terms) is the two way set of messages
> > > > > comprising a
> > > > > > single "unit of business".  It's sort of like a session as I 
> think I
> > > > > > pointed out in an earlier post about the messaging 
> specification.  So I
> > > > > > think it is equivalent to what you called a message set ID in
> > > > > your posting.
> > > > > > I did point out elsewhere that each party should be allowed to
> > > > > separately
> > > > > > identify the conversation and suggested that the existing
> > > > > conversation ID
> > > > > > should have separate subelements for the two parties.  The
> > > > > reason is that
> > > > > > for routing purposes, the conversation ID should be a rapid 
> locator. I
> > > > > > suggested that each partner should be allowed to put in
> > > > > whatever it needs
> > > > > > for its half of the conversation ID to enable rapid 
> locating.  We do use
> > > > > > the conversation ID for routing within each party in our IBM 
> Research
> > > > > > prototype run-time. I'm not sure that the URI qualifies as a
> > > > > rapid locator
> > > > > > though of course using it shouldn't be precluded.
> > > > > >
> > > > > > The TPA is an identifier of the TPA document itself.  The TPA
> > > > > ID identifies
> > > > > > the application and pair of parties while the conversation ID
> > > > > identifies a
> > > > > > specific conversation.  I did err in suggesting that both the 
> TPA ID and
> > > > > > the conversation ID are needed for routing.  If the 
> conversation ID is
> > > > > > unique within one party's messaging system, then conversation 
> ID is the
> > > > > > routing quantity and the TPA ID could be part of the state
> > > > > information for
> > > > > > the conversation rather than being used for routing.  For the 
> greatest
> > > > > > implementation flexibility, however, I strongly suggest that 
> the TPA  ID
> > > > > > also be allowed to be a rapid locator and that each party be 
> allowed to
> > > > > > choose its own value for the TPA ID.  Clearly, the local TPA
> > > > > IDs could be
> > > > > > associated with a global repository ID for the TPA but it's the
> > > > > two local
> > > > > > values that should be in the message header to give the best 
> routing
> > > > > > implementation flexibility.
> > > > > >
> > > > > > Regards,
> > > > > > Marty
> > > > > >
> > > > > >
> > > > > ******************************************************************
> > > > > *******************
> > > > > >
> > > > > > IBM T. J. Watson Research Center
> > > > > > P. O. B. 704
> > > > > > Yorktown Hts, NY 10598
> > > > > > 914-784-7287;  IBM tie line 863-7287
> > > > > > Notes address:  Martin W Sachs/Watson/IBM
> > > > > > Internet address:  mwsachs @ us.ibm.com
> > > > > >
> > > > > ******************************************************************
> > > > > *******************
> > > > > >
> > > > > > David Burdett <david.burdett@commerceone.com> on 08/16/2000 
> 02:27:30 PM
> > > > > >
> > > > > > To:   Martin W Sachs/Watson/IBM@IBMUS
> > > > > > cc:   "ebXML Transport (E-mail)" <ebxml-transport@lists.ebxml.org>
> > > > > > Subject:  RE: Re[2]: Trading Partner Logical Identification
> > > > > based on EDIFA
> > > > > >
> > > > > > Marty
> > > > > >
> > > > > > Somewhere in this text you say ...
> > > > > >
> > > > > > >>>Again, it is the combination of registry ID, partner ID, TPA 
> ID, and
> > > > > > conversation ID that does the routing<<<
> > > > > >
> > > > > > I'm under the impression that conversation ID is equivalent to
> > > > > what the TRP
> > > > > > used to call the Message Set Id which was defined as a unique 
> identifier
> > > > > > for
> > > > > > the set of messages exchanged between two parties that support
> > > > > the instance
> > > > > > of an execution of a service.
> > > > > >
> > > > > > On the other hand the current spec defines conversationId as a 
> URI which
> > > > > > identifies the conversation instance of the Trading Partner
> > > > > Agreement which
> > > > > > governs the processing of the message and which holds the state 
> of the
> > > > > > conversation between the two Parties
> > > > > >
> > > > > > I don't think these things are quite the same. Can you clarify the
> > > > > > distinction between a TPA Id and a Conversation Id. As if a
> > > > > Conversation Id
> > > > > > means the same as the old Message Set Id I don't think it 
> should be used
> > > > > > for
> > > > > > routing.
> > > > > >
> > > > > > David
> > > > > >
> > > > > > -----Original Message-----
> > > > > > From: mwsachs@us.ibm.com [mailto:mwsachs@us.ibm.com]
> > > > > > Sent: Tuesday, August 15, 2000 3:24 PM
> > > > > > To: David RR Webber
> > > > > > Cc: Mark NOBLES; ebXML-Transport@lists.ebxml.org
> > > > > > Subject: Re: Re[2]: Trading Partner Logical Identification based on
> > > > > > EDIFA
> > > > > >
> > > > > > Quite a flurry of postings!  I just got caught up on it.  I
> > > > > sense a great
> > > > > > deal of violent agreement on the identifier subject.  I agree 
> with Mark
> > > > > > Nobles' last posting and Dave Webber's agreement with it.  Here
> > > > > are a few
> > > > > > summary points:
> > > > > >
> > > > > > <SNIP>
> > > > >
>begin:vcard
>n:Van Huizen;Gordon
>tel;work:510-848-1988
>x-mozilla-html:TRUE
>url:http://www.sonicmq.com
>org:Progress Software;XML and Internet Technology
>adr:;;14 Oak Park;Bedford;MA;01730;
>version:2.1
>email;internet:gvh@progress.com
>title:Director, Product Management
>fn:Gordon Van Huizen
>end:vcard