[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [wsbpel] 12.5.1 Default Compensation Order
This is the only response I got so far. I don't have a problem with this
sentence, but maybe there is a proposal that would provide even more
clarification.
I assume there is no disagreement on the intention, i.e.
- allow parallel compensation where possible (where not constrained by
control dependencies)
- parallel compensation may result in interleaved compensation processing
- processes that *require* interleaved compensation processing are
disallowed
Kind Regards
DK
Dieter König Mail: dieterkoenig@de.ibm.com IBM Deutschland Entwicklung GmbH
Senior Technical Staff Member Tel (office): (+49) 7031-16-3426 Schönaicher Strasse 220
Architect, Business Process Choreographer Fax (office): (+49) 7031-16-4890 71032 Böblingen
Member, Technical Expert Council Tel (home office): (+49) 7032-201464 Germany
Danny van der
Rijn
<dannyv@tibco.com To
> Dieter Koenig1/Germany/IBM@IBMDE
cc
12.05.2006 18:48 wsbpel@lists.oasis-open.org
Subject
Re: [wsbpel] 12.5.1 Default
Compensation Order
I'm not sure I understand the meaning of:
"While this is allowed using
parallel execution of compensations respecting control dependencies"
Dieter Koenig1 wrote:
Satish, Alex and I had a mail exchange about the default compensation
order
and came up with a clarification for section 12.5.1 Default
Compensation
Order, last paragraph, last sentence (starting on line 5150) - it
would
change from:
Strict reverse order of completion applied to compensation of all
scopes
might not be in depth-first order, and could require interleaving
of
nested compensations across peer scopes, which is disallowed by
this
specification.
to:
Strict reverse order of completion applied to compensation of all
scopes
might not be in depth-first order, and could require interleaving
of
nested compensations across peer scopes. While this is allowed
using
parallel execution of compensations respecting control
dependencies,
process definitions that require interleaving of nested
compensations
across peer scopes in order to respect control dependencies are
disallowed by this specification.
If I hear no objection, I will open an editing Action Item for this
change.
Kind Regards
DK
Dieter König Mail:
dieterkoenig@de.ibm.com IBM Deutschland Entwicklung GmbH
Senior Technical Staff Member Tel (office): (+49)
7031-16-3426 Schönaicher Strasse 220
Architect, Business Process Choreographer Fax (office): (+49)
7031-16-4890 71032 Böblingen
Member, Technical Expert Council Tel (home office): (+49)
7032-201464 Germany
Kind Regards
DK
Dieter König Mail:
dieterkoenig@de.ibm.com
IBM Deutschland Entwicklung GmbH
Senior Technical Staff Member Tel (office): (+49)
7031-16-3426 Schönaicher Strasse 220
Architect, Business Process Choreographer Fax (office): (+49)
7031-16-4890 71032 Böblingen
Member, Technical Expert Council Tel (home office): (+49)
7032-201464 Germany
Alex Yiu
<alex.yiu@oracle.
com>
To
Dieter
Koenig1/Germany/IBM@IBMDE
10.05.2006 19:39
cc
Satish Thatte
<satisht@microsoft.com>, Alex
Yiu
<alex.yiu@oracle.com>
Subject
Re: [Fwd: [wsbpel] the last
paragraph of "Default
Compensation
Order" section]
Hi Dieter,
How do you propose to carry out the situation forward?
Since we are done with chapter 12, do we want to open an issue for
this
rewording change?
Thanks!
Regards,
Alex Yiu
Dieter Koenig1 wrote:
>:-) Thanks for your help! IMO, the new text addresses the point more
>clearly, i.e. allowing parallel compensation as long as this does
not
>require interleaved execution.
>
>Kind Regards
>DK
>
> Dieter König Mail:
dieterkoenig@de.ibm.com
IBM Deutschland Entwicklung GmbH
>
> Senior Technical Staff Member Tel (office): (+49)
7031-16-3426 Schönaicher Strasse 220
>
> Architect, Business Process Choreographer Fax (office): (+49)
7031-16-4890 71032 Böblingen
>
> Member, Technical Expert Council Tel (home office):
(+49)
7032-201464 Germany
>
>
>
>
>
>
>
> "Satish Thatte"
> <satisht@microsof
> t.com>
To
> "Alex Yiu"
<alex.yiu@oracle.com>
> 10.05.2006 00:30
cc
> Dieter
Koenig1/Germany/IBM@IBMDE
>
Subject
> RE: [Fwd: [wsbpel] the last
> paragraph of "Default
Compensation
> Order" section]
>
>
>
>
>
>
>
>
>
>
>See? I can still be semi-useful J
>
>
>From: Alex Yiu [mailto:alex.yiu@oracle.com]
>Sent: Tuesday, May 09, 2006 2:26 PM
>To: Satish Thatte
>Cc: Dieter Koenig1; Alex Yiu
>Subject: Re: [Fwd: [wsbpel] the last paragraph of "Default
Compensation
>Order" section]
>
>
>Satish,
>
>I am happy with your rewording. :-)
>
>Regards,
>Alex Yiu
>
>
>Satish Thatte wrote:
>How about rewording
>
>Strict reverse order
>5151 of completion applied to compensation of all scopes might not
be in
>depth-first order, and
>5152 could require interleaving of nested compensations across peer
scopes,
>which is
>5153 disallowed by this specification.
>
>As
>
>Strict reverse order
>5151 of completion applied to compensation of all scopes might not
be in
>depth-first order, and
>5152 could require interleaving of nested compensations across peer
scopes.
>While this is allowed using parallel execution of compensations
respecting
>control dependencies, process definitions that require interleaving
of
>nested compensations across peer scopes in order to respect control
>dependencies are disallowed by this specification.
>
>Is this what you are trying to say? If so it is not coming across
with
the
>current wording.
>
>See below for another comment
>
>
>From: Alex Yiu [mailto:alex.yiu@oracle.com]
>Sent: Tuesday, May 09, 2006 11:28 AM
>To: Satish Thatte
>Cc: Dieter Koenig1; Alex Yiu
>Subject: Re: [Fwd: [wsbpel] the last paragraph of "Default
Compensation
>Order" section]
>
>
>I thought the "interleaving" term was refering to the cyclic
dependencies
>among peer scopes?
>[Satish Thatte] In my mind you either have a stack or a tree during
default
>comp-stack being single-compensation-at-a-time. I think we have
made
stack
>possible by disallowing dependencies that would force interleaving,
i.e.,
>tree-like default compensation unrolling. But we allow interleaving
for
>parallelism.
>
>Strict reverse time order may require interleaving due to
parallelism in
>forward work. We allow parallelism in reverse work thus we allow
strict
>reverse time order. This is what "The default compensation order
mandated
>by the rules here is consistent with strict reverse order of
completion"
>means to me.
>
>
>I tend to think parallelism in compensation would be interpreted as
>"interleaving" only in (round-robin) work scheduler of BPEL
processor
>viewpoint.
>
>Anyhow, I am more than happy to accept one more version of rewording
of
>paragraph, as long as people (including me) can interprete the new
wording
>correctly. :-)
>
>Regards,
>Alex Yiu
>
>
>Satish Thatte wrote:
>Actually I am confused because the use of allowed parallelism in
default
>compensation may cause interleaving don't you think?
>
>
>
>From: Alex Yiu [mailto:alex.yiu@oracle.com]
>Sent: Monday, May 08, 2006 4:18 PM
>To: Satish Thatte
>Cc: Dieter Koenig1; Alex Yiu
>Subject: Re: [Fwd: [wsbpel] the last paragraph of "Default
Compensation
>Order" section]
>
>
>Hi Satish,
>
>Here is the PDF generated based on the version checked in last
friday.
>
http://www.oasis-open.org/apps/org/workgroup/wsbpel/download.php/18005/wsbpel-specification-draft.rev1.145.pdf
>
>It includes all the latest chapter 12 "scope" changes.
>
>----------------------------------------------------
>5146 An effect of Rule 2 is to permit a depth-first traversal of the
>lexical scope tree for default
>5147 compensation, respecting the control dependency relation among
peer
>scopes as dictated
>5148 by Rule 1. Default compensation order of a scope resulting from
these
>rules is dependent
>5149 only on the compensation of its nested scopes. The default
>compensation order mandated
>5150 by the rules here is consistent with strict reverse order of
>completion. Strict reverse order
>5151 of completion applied to compensation of all scopes might not
be in
>depth-first order, and
>5152 could require interleaving of nested compensations across peer
scopes,
>which is
>5153 disallowed by this specification.
>----------------------------------------------------
>
>I guess the phrase "disallowed by this specification" refers to the
the
>compensation order that require interleaving of nested across peer
scope.
>
>I hope that makes sense to you.
>Please feel free to make more suggestion to re-tune the wordings of
that
>paragraph.
>
>Thanks!
>
>
>Regards,
>Alex Yiu
>
>
>Satish Thatte wrote:
>Hi Dieter,
>
>Good to hear from you. And glad to know that I haven't lost it
completely
>:-)
>
>Given what you say about reversing the edges, what does the "which
is not
>supported by this specification" phrase refer to?
>
>Satish
>
>-----Original Message-----
>From: Dieter Koenig1 [mailto:dieterkoenig@de.ibm.com]
>Sent: Monday, May 08, 2006 5:48 AM
>To: Satish Thatte
>Cc: Alex Yiu
>Subject: RE: [Fwd: [wsbpel] the last paragraph of "Default
Compensation
>Order" section]
>
>Hi Satish, hope everything is going well!
>
>Your response is still correct - we would like to make sure that the
>"reverse the edges" approach, which allows parallel compensation, is
still
>allowed.
>
>Kind Regards
>DK
> Dieter König Mail:
dieterkoenig@de.ibm.com
>IBM Deutschland Entwicklung GmbH
> Senior Technical Staff Member Tel (office): (+49)
>7031-16-3426 Schönaicher Strasse 220
> Architect, Business Process Choreographer Fax (office): (+49)
>7031-16-4890 71032 Böblingen
> Member, Technical Expert Council Tel (home office):
(+49)
>7032-201464 Germany
>
>
>
>
>
>
> "Satish Thatte"
> <satisht@microsof
> t.com>
To
> "Alex Yiu"
<ALEX.YIU@ORACLE.COM>
> 05.05.2006 05:38
cc
> Dieter
Koenig1/Germany/IBM@IBMDE
>
Subject
> RE: [Fwd: [wsbpel] the last
> paragraph of "Default
Compensation
> Order" section]
>
>
>
>
>
>
>
>
>
>
>I have lost a lot of context but I think the main point as I recall
was
the
>following
>
> 1. what are the cases where a strict reverse order is not depth
first?
> In cases of parallelism--in those cases control dependencies
must
> still be restricted
> 2. if we allow default compensation implementations to use
parallelism
> (so long as control dependencies are reversed) then in fact it
is
> possible to perform default comp in reverse order of time but
not
> reverse depth first
>
>Basically depth first is a single threaded reverse comp behavior and
I
>thought we were not prohibiting implementations from being parallel
>
>But then I might be missing something from that or subsequent
discussion
or
>just getting confused J
>
>I thought IBM in particular were keen on the "reversing arrows"
>interpretation.
>
>Dieter?
>
>
>
>From: Alex Yiu [mailto:ALEX.YIU@ORACLE.COM]
>Sent: Thursday, May 04, 2006 3:27 PM
>To: Satish Thatte
>Cc: Alex Yiu
>Subject: RE: [Fwd: [wsbpel] the last paragraph of "Default
Compensation
>Order" section]
>
>
>Hi Satish,
>
>Dieter has put it a variant of the text that I sent to you. I don't
have
>the exact copy right now.
>
>Yes, we did correct the grammar mistake. :-)
>
>The reason that we tried to reword the paragraph a bit is: we have
>seemingly have two phrases that are seemingly difficult to be jelled
>together:
>"... Rule 2 is to permit a depth-first traversal ..." and
>"default compensation order mandated by the rules here is consistent
with
>strict reverse order of completion, but not in depth-first order
..."
>
>"not supported by this specification" was changed into something
like "not
>allowed by this specification". I was trying to say that when a
process
>definition is accepted, its default compensation order is consistent
with
>the strict reverse order. However, there are some cases of strict
reverse
>order, which is not DF traversal, and which involves interleaving
among
>scopes. Those cases are not supported.
>
>I hope that sounds reasonable to you.
>Thanks!
>
>
>Regards,
>Alex Yiu
>
>
>From Satish Thatte <satisht@microsoft.com>
>Sent Thu 5/4/2006 12:47 PM
>To Alex Yiu <alex.yiu@oracle.com>
>Subject RE: [Fwd: [wsbpel] the last paragraph of "Default
Compensation
>Order" section]
>
>Comment on grammar:
>
>Strict reverse order of completion applied to compensation of all
scopes
>may be not in depth-first order
>
>Should be
>
>Strict reverse order of completion applied to compensation of all
scopes
>may not be in depth-first order
>
>As I recall, the "consistent with" comment was relating to the
openness to
>concurrency of the mandated default order (subject to modeled
control
>dependency not contingent temporal order), i.e., the mandated
default
order
>was a partial order. Thus I don't see how we can say both that it
is
>consistent and "which is not supported by this specification". But
I
don't
>have the full section in front of me right now.
>
>From: Alex Yiu [mailto:alex.yiu@oracle.com]
>Sent: Thursday, May 04, 2006 12:24 AM
>To: Satish Thatte
>Cc: Alex Yiu
>Subject: [Fwd: [wsbpel] the last paragraph of "Default Compensation
Order"
>section]
>
>
>Hi Satish,
>
>Sorry to bother you ... :-)
>
>We are in the middle of paragraph-by-paragraph review of the spec,
in
order
>to wrap the spec up. We are now reviewing the "scope" chapter.
>Particularly, the "default compensation order" section. We have some
>questions about the last paragraph of that section
>
>Since you are one of the transaction related experts in this TC and
the
>original author of the paragraph, I am wondering you think that new
>wordings is a clarified verison of text without losing any real
intent or
>creating any unnecessary implication.
>
>Looking forward to you reply.
>Thanks!
>[ I could rely your response back in a form that you prefer. ]
>
>Regards,
>Alex Yiu
>
>
>-------- Original Message --------
>
>Here is an attempt to smooth this paragraph out:
>----------------------
>An effect of Rule 2 is to permit a depth-first traversal of the
lexical
>scope tree for default
>compensation, respecting the control dependency relation among peer
scopes
>as dictated
>by Rule 1. Since depth-first order implies that such compensation is
only
>dependent on
>the compensation of its nested scopes, default compensation order
can be
>easily defined.
>The default compensation order mandated by the rules here is
consistent
>with strict
>reverse order of completion. Strict reverse order of completion
applied to
>compensation
>of all scopes may be not in depth-first order and require
interleaving of
>nested
>compensations across peer scopes, which is not supported by this
>specification.
>----------------------
>
>(I attached the PDF and DOC version as well).
>Let's see whether this text is accepted tomorrow.
>
>
>Regards,
>Alex Yiu
>
>
>
>
>
>
>
---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail. You may a link to this group and all your TCs
in OASIS
at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]