I think that the BP TC has discussion of
WSBPEL relations or referencing for version 3 of BPSS. So maybe some aspects of
your proposed agenda are already penciled in on the calendar after version 2 is
finished.
From: Sacha Schlegel
Sent: Friday, February 04, 2005
9:55 AM
To: ; ebXML BP; ebxml-msg@lists.oasis-open.org
Subject: AW: [ebxml-bp] Closing
the gap between MSI and BSI and move on
So MSI part is as I understand.
The BSI part becomes a bit vague. Its abstract, its virtual, its
logical, all well and good but at one point it has to become specific.
So what we agree is that whatever the BSI is, it will be responsible
for BPSS signals.
So what is a specific BSI?
o J2EE applications, busses, simple backend application, series of web
services, currently �ts just about anything ...
I can argue contrary that you do not want to give the responsibilities
of the BPSS signals to all and everything being part of the BSI. This is the
opposite to rip out existing applications or internal data formats.
1) A BSI is anything spread among anything at one party
2) A BSI (or BSH) is a specific component (a ebBP business process
engine) which takes on the responsibity of the BPSS signals. A component that
is build for ebBP�s.
With the second case we could close the gap between BSI and MSI.
But with the second case we still have not integrated with the actual
backend applications, services. With this approach we would at least, have
a proper ebXML chain consisting of BSI and MSI.
Following private business processes from too far away it seems to me
that private business process engines (such as WSBPEIL) are discussed today as
a solution for private business processes.
If we were at this stage we would still have the gap between the BSI
and WSBPEL. That would be the next task.
To come back to the BPSS signals. If a WSBPEL instance knows to communicate
with a BSI and vice verca, a WSBPEL instance would know about BPSS signals,
such as Acceptance Acknowledgements.
So, having a specific BSI (with an interface between BSI and MSI) and
an interface between BSI and WSBPEL for example is ONE way to get where we
want.
PS: Of course MSH implementations can still be connected straight to
their backend applications or service or abstract-virtual BSI ;) as
is probably done today.
Von:
Gesendet: Fr 04.02.2005 08:31
An: Sacha Schlegel; ebXML BP;
ebxml-msg@lists.oasis-open.org
Betreff: RE: [ebxml-bp] Closing
the gap between MSI and BSI and move on
Hi Sacha,
The BSI concept mainly comes from the ebXML architecture document and the BPSS
working area.
The MSI concept was, I think, one that has been worked on from time to time in
Messaging and was basically an API for communication between the software
component(s) functioning in the MSH role (at a given node) and the software
(other middleware or business applications) that needed to communicate with a
MSH at that node. [An interface between "levels" of the stack at a
node.] Submitting and being notified of message payloads (along with
selected metadata concerning that payload), being notified of exception
conditions, and so on were the envisioned functions of the MSI API. Matt and
others even submitted materials for an API definition and we found that OAG (or
was it OMG?) had drawn up some IDL for an API in this area. Messages on this
are probably still in the Messaging list archive. [If you go back several
years, you will find other contributions. It is a perennial topic.]
I think that the Messaging TC decided to work on version 3.0 as a higher
priority than the API for MSI. That doesn't prevent someone from submitting a
draft for TC consideration eventually, though Ian would need to check the
process for us in Messaging. I agree with you that someday this needs to get
done. At the moment each MSH integration is a customized exercise depending on
the deployment environment (EAI or not; if so, what style, etc. etc.)
The BSI as discussed in BPSS and the early Architecture document is IMO an
interface at a specific level of the protocol stack, but an interface between
the same stack level but at different nodes. It is a virtual (or logical)
interfacing whose actual realization is accomplished by using the supporting
layers of the stack to carry out its protocol. Abstractly the BSI is
accomplished whenever ebXML message exchange carryies out (via its lower
protocol levels) a BusinessTransaction (in the BPSS sense). BSI in BPSS
involves the "signals" (both positive and for exceptions) that
achieve state alignment (including clear failure states).
So BSI is somewhat defined. What more do you think needs to be done for it?
It is true that BSI will be realized by some assignment of responsibilities to
software as deployed in some environment. I am not certain that anyone has
proposed standardization of its implementation by even saying which roles of
software components carry out which BSI related tasks. There are just a whole
lot of existing combinations in businesses for these deployment environments,
and overall the consensus view has been that standardizing these largely
internal configurations would be too controversial to achieve consensus. [A
standard which said that businesses had to rip out existing applications or
internal data formats or message busses would be probably doomed to sit without
any market traction, with at most one large vender supporting it, which would be
the vender whose software or platform or language or development IDE followed
the standard!] So I am a little dubious that we should try to standardize this
area without receiving clear broadly based requirements and commitments from
the end user community. Absent their interest, a basis for consensus in
standardization among ISVs is most unlikely.