RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call

[Martin W Sachs <mwsachs@us.ibm.com>]

Todd,

Some replies below.

Regards,
Marty

*************************************************************************************

Martin W. Sachs
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
*************************************************************************************



Todd Boyle <tboyle@rosehill.net> on 11/21/2001 03:40:56 PM

To:    Martin W Sachs/Watson/IBM@IBMUS
cc:    ebtwg-bps@lists.ebtwg.org, ebxml-cppa@lists.oasis-open.org
Subject:    RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call



1.  Is BPS a reasonable shorthand for a BP Specification instance?

MWS:  To me there isn't enough hamming distance from BPSS.  However
the first thing needed is to pin down a precise definition of
"BPSS Instance" including maybe another term.

In other words the BPSS is a schema which allows the creation
of business process specifications which are themselve schemas?

MWS:  The BPSS is a specification, including an XML schema document,
for creating business process specifications.  A business process
specification is an XML instance document that conforms to the single
XML schema in the BPSS specivication. In my view, the use of "Schema"
in the title "Business Process Specification
Schema" causes confusion because the specification produces an XML instance
document, not an XML schema document.

2.  Could you give me, just briefly, is it an objective that UUID's are
globally unique rather than just unique within a particular RegRep,
and if so what is the mechanism to ensure ?   I apologize for failing
to study the history of this,

MWS: The uniqueness of the UUIDs of "BPSS instances" goes beyond a single
RegRep. The concatenation CPAId.UUID has to guarantee that a message
arriving
in a system will be directed to the proper software entry point for that
CPA.
Given the variety of system structures in the world, its a hard to be more
precise.  HOwever if the UUID is globally unique throughout the known
universe,
that should suffice.

Thank you,
Todd

At 10:16 AM 11/21/01, Martin W Sachs wrote:

>Dale,
>
>The definition of "BPSS Instance" in your posting ("chunk of XKL") is to
>loose to be useful.  The real question is the definition of what that UUID
>is attached to.  If it's just "a chunk of XML", then the implication to me
>is that if several instances of an application are deployed with different
>CPAs, but all referring to the same BPSS instance document, then the UUID
>is the same for all of them.  That is actually OK as long as the UUID is
>always qualified with the CPAId.
>
>I would like to suggest that a more useful definition of "BPSS Instance"
is
>based on the deployment rather than the chunk of XML, I.e. the BPSS
>instance document deployed for a specific instance of an application.  I
>would then expect a unique UUID to be associated with each deployed
>instance of that (or any) BPSS instance document.
>
>I suspect that either of the above approaches (CPAId.UUID or UUID for each
>deployed instance) would work as long as everything is clearly spelled out
>in the specification.
>
>Regards,
>Marty
>
>
*************************************************************************************

>
>Martin W. Sachs
>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
>
*************************************************************************************

>
>
>
>"Dale Moberg" <dmoberg@cyclonecommerce.com> on 11/21/2001 12:52:56 PM
>
>To:    Martin W Sachs/Watson/IBM@IBMUS
>cc:    <ebtwg-bps@lists.ebtwg.org>, <ebxml-cppa@lists.oasis-open.org>
>Subject:    RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call
>
>
>
>Just a quick response. By "BPSS instance"
>I just mean a chunk of XML that conforms
>to the BPSS schema, er, DTD.
>
>Maybe there is some more elaborate
>definition but I don't recall seeing
>it anywhere.
>
>As far as the remedies go, I think we
>are in basic agreement on what is
>required. I would like BPSS to
>provide some requirements or at
>least recommendations on how
>BP designers should group, and
>with it, some guidance on where
>named subservice packages might be
>found.
>
>Still, however, real users will
>want to pick and choose based
>on the realities of their backend
>systems. So some approach like 3
>or 4 is likely to be needed in practice.
>
>Dale
>
>-----Original Message-----
>From: Martin W Sachs [mailto:mwsachs@us.ibm.com]
>Sent: Wednesday, November 21, 2001 10:42 AM
>To: Dale Moberg
>Cc: ebtwg-bps@lists.ebtwg.org; ebxml-cppa@lists.oasis-open.org
>Subject: RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call
>
>
>
>A few questions/comments below.
>
>Regards,
>Marty
>
>************************************************************************
>*************
>
>Martin W. Sachs
>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
>************************************************************************
>*************
>
>
>
>"Dale Moberg" <dmoberg@cyclonecommerce.com> on 11/21/2001 11:44:09 AM
>
>To:    Martin W Sachs/Watson/IBM@IBMUS
>cc:    <ebtwg-bps@lists.ebtwg.org>, <ebxml-cppa@lists.oasis-open.org>
>Subject:    RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call
>
>
>
>Marty,
>
>This issue was mentioned on the call, and I believe
>we did agree that the BPSS instance would always
>have a UUID that would uniquely identity and distinguish
>it from other BPSS instances.
>
>David Smiley summarized this as:
>
>"Service = BPSS Unique Process Identifier (UUID)
>Request to make UUID a URN"
>
>MWS:  What is a BPSS instance?  Suppose I have three separate
>"applications" that use the same BPSS instance document. By
>this I mean that there is literally one BPSS instance document
>that is deployed separately for each application?  Is that
>one BPSS instance with one UUID or three BPSS instances with
>three distinct UUIDs?  If it's 3 separate BPSS instances with
>distinct UUIDs, then I agree that everything is fine.
>
>MWS:  The term "BPSS Instance" needs to be defined, probably
>in the BPSS specification.
>
>If a short name is used within an agreement, its
>uniqueness would depend upon being
>scoped by the CPAId, as you and Hima note.
>
>Some issues that still have not been resolved
>concerning standardization of naming conventions
>flow from  the following: 1. A BPSS instance is not
>guaranteed to have any "functional" unity (that is,
>it can consist of unrelated packages of functionality),
>2. The CPA currently uses the ServiceBinding
>to specify similar treatment for all leaf actions
>under a BPSS instance (when BPSS used), and 3. CPAs
>may wish to make agreements only over a subset of
>leaf actions within a BPSS instance.
>
>Suppose 2 companies want to engage in OrderManagement,
>but do not wish to engage in POCancel actions or
>POChange actions. [This is not a far-fetched example,
>by the way.] If the ServiceBinding embraces all actions
>in the instance, they will in effect agree to bindings
>for actions that they do not support across the mythical
>BSI interface.
>
>These situations may have the following possible remedies
>(and I am sure there are others as well!):
>
>1. Create ways to specify a portion
>of a BPSS instance as the
>targeted package of leaf actions.
>That reopens themyriad "name" attributes
>from which the service name
>could come from. May be feasible though.
>2. Create ways to have Override specify
>a null binding for
>actions not agreed to.
>(Not wild about
>this one either
>as it seems to introduce clutter that
>the binding-up-actions as a group)
>3. Apply Tony W's XSLT filter ideas to
>the BPSS instance to extract
>the right ones. If we use this technique,
>then we still
>should document how the service
>naming convention works.
>
>MWS:  There is one more remedy, which is
>to create a BPSS instance document that
>defines only the functions that will be
>used for a particular interaction. This
>is probably just a restatement of (3) and
>is my preference.
>
>MWS: I agree that we MUST document the service
>naming convention for any supported
>approach.
>
>
>
>
>-----Original Message-----
>From: Martin W Sachs [mailto:mwsachs@us.ibm.com]
>Sent: Wednesday, November 21, 2001 7:34 AM
>To: Dale Moberg
>Cc: Dan Weinreb; dsmiley@mercator.com; ebtwg-bps@lists.ebtwg.org;
>ebxml-cppa@lists.oasis-open.org
>Subject: RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call
>
>
>
>Dale,
>
>It isn't clear to me from previous discussions that when more than one
>application behind the same MSH uses the same BPSS instance document,
>the
>different "BPSS instances" will have unique service names.  Do they all
>have the same UUID or unique UUIDs in this case?
>If the service names are the same in that case, it's OK as long as they
>are
>qualified with the CPAId.
>
>Regards,
>Marty
>
>************************************************************************
>*************
>
>Martin W. Sachs
>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
>************************************************************************
>*************
>
>
>
>Dale Moberg <dmoberg@cyclonecommerce.com> on 11/20/2001 06:44:17 PM
>
>To:    Dan Weinreb <dlw@exceloncorp.com>, dsmiley@mercator.com
>cc:    ebtwg-bps@lists.ebtwg.org, ebxml-cppa@lists.oasis-open.org
>Subject:    RE: [ebxml-cppa] Re: Notes from BPSS/CPPA conference call
>
>
>
>Dan Weinreb asked:
>
>"It seems to me that this would mean that every message sent under the
>control of a particular BPSS would have the same Service value.  That
>is, all the messages sent by a given business process would all have
>the same value for Service.  Is that what was intended?"
>
>
>Yes, I think that is the current proposal under consideration.
>
>This convention would be coupled with the
>constraint that all distinct actions
>within the BPSS instance would have
>distinct names (both the short version
>names and the long Xpath names
>would be distinct within a "service")
>
>So an OrderManagement service would
>have PurchaseOrder ,  PurchaseOrderChange,
>PurchaseOrderCancel, PurchaseOrderAcknowledge
>actions, for example.
>
>
>
>----------------------------------------------------------------
>To subscribe or unsubscribe from this elist use the subscription
>manager: <http://lists.ebtwg.org/ob/adm.pl>
>
>
>
>
>
>
>
>
>
>
>----------------------------------------------------------------
>To subscribe or unsubscribe from this elist use the subscription
>manager: <http://lists.ebtwg.org/ob/adm.pl>