[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [wsbpel] Issue 115 - RE: [wsbpel] appendix C revision
Dear Paco,
We strongly agree that Appendix C, in some form, has great value in
tersely (and pictorially) summarizing the interactions surrounding
compensatory scopes. We think one would lose a great deal more than
extra text to review if it were excised altogether. Issues 6, 135 and
142 are all usefully illuminated by this. We have also found Appendix C
useful in discussing, designing, interfacing and implementing
distributed coordination of process-driven services, and would recommend
its retention. Vendors who desire to create federated or fully
distributed BPEL implementations will benefit, we believe, from this
helpful and concise view on the content of the normative spec.
The changes we made (beyond removing the term "WS-BA") boil down to
a) clarifying the meaning of "local" vs "distributed" in this context
b) changing the names of messages/states to correspond to their ordinary
names in BPEL
c) aligning the diagram with the normative text of the specification as
it has developed during the TC's work, and with the original text of the
Appendix (with which it was previously at some variance). The original
diagram derived from WS-BA (2002) was inadequate, and also does not
reflect the current WS-BA draft.
On your two specific points:
1. Either we go the whole hog, and (as you suggest) specify all the
messages required for a distributed coordination protocol, which would
permit implementation of a BPEL process engine as a federation of
independently existing processing units, which neither share state nor
can be assumed to be failure-coupled. Or, we take the route of the
original Appendix C, and omit messages which are needed to resynchronize
in the event of failures. We chose to stick as closely as possible to
the original Appendix C. We have no problem with going the whole hog, if
that makes things clearer or more complete.
2. On exit handling: the key point is that a nested scope can still
unilaterally exit, either rethrowing its fault as it does so, or
containing the fault, and reporting that the inner scope has completed
erroneously. These two paths to the Ended state now travel through a new
state, Faulting, which was introduced in order to elucidate the textual
point (bullet 2 of the original appendix/ bullet 3 of our revision) that
either route to Ended must begin with the raising of an internal fault,
breaking the inner scope out of its Active state. In other words, we
separated Active into Running and Faulting, to make the genesis of these
transitions clear. There is no loss of content or functionality, only
greater precision and alignment with the specification as a whole, and
the appendix in particular.
Yours,
Alastair and Peter
-----Original Message-----
From: Francisco Curbera [mailto:curbera@us.ibm.com]
Sent: 14 September 2004 17:59
To: Furniss, Peter
Cc: wsbpel@lists.oasis-open.org
Subject: RE: [wsbpel] Issue 115 - RE: [wsbpel] appendix C revision
Hi Peter,
Here is IBM's feedback on your proposed resolution.
We agree that Appendix C provides important value in helping clarify
the
interaction between BPEL and distributed protocols. That was the
original
motivation for the appendix; in its current form (modulo references to
the
WS-BA spec) it does this job appropriately and that is why we agreed
that
it should not be removed. Our major concern with the proposed changes is
that the new text only deals with localized behavior and does not help
anymore in interoperability scenarios. In that case we think it would be
better to leave it out of the document.
More specifically, these are the two issues that concern us:
1. Missing fault and compensation-fault acknowledgements: The state
diagrams are meant to accommodate any underlying infrastructure and
therefore we believe that every transition requires some form of
protocol-level acknowledgement to assure the partner has processed the
signal.
2. Exit Handling: We need to allow a nested scope to unilaterally leave
the
workscope.
Regards,
Paco
"Furniss, Peter"
<Peter.Furniss@chor To: "Satish
Thatte" <satisht@microsoft.com>, <wsbpel@lists.oasis-open.org>
eology.com> cc:
Subject: RE: [wsbpel]
Issue 115 - RE: [wsbpel] appendix C revision
08/25/2004 02:04 PM
Sounds reasonable - "Success" should be changed to "Succeeded" by the
same
count.
Peter
-----Original Message-----
From: Satish Thatte [mailto:satisht@microsoft.com]
Sent: 25 August 2004 18:15
To: Furniss, Peter; wsbpel@lists.oasis-open.org
Subject: RE: [wsbpel] Issue 115 - RE: [wsbpel] appendix C revision
Peter,
It would be helpful to follow the original convention of ending all
signals
from a nested scope with 'ed - by this convention "Fault" would be
"faulted". Thus all these signals look informative as opposed to
imperative.
Satish
From: Furniss, Peter [mailto:Peter.Furniss@choreology.com]
Sent: Wednesday, August 18, 2004 7:19 AM
To: wsbpel@lists.oasis-open.org
Subject: [wsbpel] Issue 115 - RE: [wsbpel] appendix C revision
Forgot to set the title so my own scripts will link the thread. Please
reply to this thread, not my original one.
Peter
-----Original Message-----
From: Furniss, Peter
Sent: 18 August 2004 14:57
To: wsbpel@lists.oasis-open.org
Subject: [wsbpel] appendix C revision
At last, the proposed text for appendix C from Alastair and myself.
Thanks also to Tony Fletcher for comments.
The bit that gave us pause was the introduction - the difference
between a
notionally monolithic BPEL implementation and a general distributed
case
becomes questionable if in fact the BPEL implementation is federated -
especially when, e.g., different flows are running in separate
processes
that could fail independently.
Peter
------------------------------------------
Peter Furniss
Chief Scientist, Choreology Ltd
web: http://www.choreology.com
email: peter.furniss@choreology.com
phone: +44 870 739 0066
mobile: +44 7951 536168
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]