Title: Message
Peter,
Apologies for the confusion, I did type BTP when I meant
Choreology. But I think the argument is the same, whether the mapping is
done at the endpoint or at the coordinator is an endless debate because both are
valid solutions.
The
multi-protocol issue is handled by a coordinator-coordinator protocol, which is
the advantage of doing the mapping at the coordinator. You can have
multiple coordinators get involved in bridging otherwise incompatible
protocols.
Sometimes I have the feeling that what Choreology is saying would be
right, if only everyone would agree with you. But that sounds like a
tautology, and therefore Choreology should be open to debate -unless of course
you firmly believe you will convince everyone that you are
right!
Eric
-----Original Message-----
From: Furniss, Peter
[mailto:]
Sent: Thursday, December 11,
2003 4:47 PM
To:
Subject:
[ws-caf] deriving from RE: [ws-caf] Patent Issue(s)
Eric,
I
don't follow your description in the second paragraph below. Choreology (not
BTP) are arguing that the protocols are essentially the same, and therefore
there is no need to have multiple ones - mapping between them is rather easy
and will only be necessary if future legacy has multiple ways of expressing
the same thing. If the coordinator does the bridging, it can only be
valid if the protocols are equivalent. If the semantics are so different that
the same community was forced to use different protocols, how can they be
bridged ?
Or
are you saying that a WS-CAF coordinator (for one transaction) might be
simultaneously handling multiple protocols from the ws-txm family, using one
p'col to some of the participants, a different one to others ? In which
case, it's just a question of whether that distinction is made by different
messages or different qualifications of the same messages - different verbs or
different adverbs.
Peter
-----Original Message-----
From: Newcomer, Eric
[mailto:]
Sent: 11 December 2003
18:47
To: Green, Alastair J.; Mark Little; Guy Pardon; Greg
Pavlik
Cc: Robin Cover; ;
Subject: RE: [ws-caf] Patent
Issue(s)
Alastiar,
Thanks very much for this analysis and comment on the patent
issue.
With regard to the "confusion" betwen WS=T/WS-TXM and BTP proponents,
I tend to describe it very differently. Either you resolve the
differences among transaction protocols and models at the endpoints (which
BTP proposes to do by requiring them to all implement a common protocol) or
you ask the coordinator to do the bridging at the "hub" or in this case the
distributed hub.
A
good analogy for this debate is the transformation debate, whether it should
be done at the hub, at the endpoints, or both. I would say that the
transformation argument isn't settled (and may never be) because both are
important and "it depends" on what you're doing as to which solution, or
combination of solutions is best. That's what we're essentially saying
in WS-CAF, use the solution that's right for the situation. Don't remove the
"it depends" from the equation since you'll never settle the
argument.
With regard to compensation, I'd refer everyone to the recent
discussion at WS-BPEL. Although I didn't attend and therefore am
basing this only on the presentation, I'd have to say it looks like the
problems with a purely compensation-based approach were highlighted.
What order do you process the compensations? What happens when there's
a failure in the compensation? Do you compensate the
compensation? Does this affect the order? Etc.
Use the right tool for the right job, let's not say there's only one
tool in the toolbox that's capable of doing anything
valuable.
Eric
-----Original Message-----
From: Green, Alastair J.
[mailto:]
Sent: Thursday, December
11, 2003 1:09 PM
To: Mark Little; Newcomer, Eric; Guy Pardon;
Greg Pavlik
Cc: Robin Cover; ;
Subject: RE: [ws-caf] Patent
Issue(s)
Guy,
Thanks for
notifying us all of this patent. I found it a very useful summary of
loosely-coupled transactional issues.
I have to say
that I question the novelty of the claimed “invention”, which seems (at a
very hasty first reading) to boil down to interposition (transaction
bridging) plus a global transaction id.
I believe that
the following publications, products or standards very likely incorporate
the key features of this “invention” and are prior
art:
CCR
OSI/TP
OMG
OTS
Transarc Encina
I suspect the
Arjuna research project dating back to the late
eighties may also incorporate features of relevance to the claim of
invention. Mark would be better qualifed to
comment than I.
I also believe
that the innards of many transactional products, and the utilization
patterns of databases to achieve high concurrency through application
locking which are employed by thousands of end-users, will also betray
many similar key features.
The cluster of
problems that make application- (service-) level coordination difficult
today for the assemblers of business processes which utilize interacting
systems includes many issues which are just practical (integration with
heterogeneous environments, usability etc). Much of the novelty
of Business Transaction Management (app-level coordination) is
context-conceptual.
One aspect is
getting over the equation of first use with all possible future uses.
For example: How
can a protocol which is associated with ACID possibly possess an interface
that is capable of being used for compensations? Of course, the XA control
structure can be so used. Then, treat XA function names as message names
and you have a distributed “interface” with the same capability … and so
on, through all the permutations of two-phase. XAResource used for JCA is a useful
case.
The second aspect
is failure to abstract sufficiently.
People often
confuse the form or the specific context with the essence. The
difficulties that the proponents of WS-T and WS-TXM have with the notion
that all the proposed interoperable outcome protocols (BTP, WS-T BA, WS-T
AT, WS-TXM Acid, LRA and BP) are in fact versions of one underlying “Ur”
protocol (optimizable two-phase outcome with
qualifiers), in my view has its root in this type of failure.
(Parenthetically, there is every reason to have contractual [or
contract-referring] information exchanged in contexts attached to
application exchanges: but that need does not transit to the coordination
protocol).
Concretely, the
biggest conceptual problem (and advance) of BTM standardization, is seeing
compensation as one, special case of negative finalization of a high-level
business transaction, coupled with the need to always have a positive
finalizer (which in that one, special case will
“just” drop its persistent
memory of the business transaction). Put another way: there is a
higher-level abstraction (business transaction), which stands clear above
the database transactions which are now driven by the business
transaction.
Alastair
Green
CTO
Choreology
Ltd
Choreology
Cohesions -- Business Transaction Management software for application
coordination
-----Original
Message-----
From: Mark
Little [mailto:]
Sent: 11 December 2003
03:25
To: Newcomer, Eric;
Guy Pardon; Greg Pavlik
Cc: Robin Cover;
;
Subject: Re: [ws-caf] Patent
Issue(s)
Apologies for any backlog of
email responses, but I've been flying to XML 2003 today. Just to second
Greg's email - LRA is based on Sagas and similar to Open Nested
transactions.
Mark.
----
Mark Little,
Chief
Architect, Transactions,
Arjuna Technologies
Ltd.
www.arjuna.com
----- Original
Message -----
From: Newcomer, Eric
To: Guy Pardon ;
Greg Pavlik
Cc: Mark
Little ; Robin Cover ;
;
Sent:
Wednesday, December 10, 2003 7:58 PM
Subject: RE:
[ws-caf] Patent Issue(s)
On
the surface, it does not seem to confict since it requires a
specific implementation of Java classes and operations that create
the transaction system, whereas the WS-CAF approach is to wrap or
abstract the transaction systems (i.e. not to create
one).
-----Original
Message-----
From:
Guy Pardon [mailto:]
Sent: Wednesday, December 10,
2003 1:32 PM
To: Greg
Pavlik
Cc: Mark
Little; Robin Cover; Newcomer, Eric; ;
Subject: Re: [ws-caf] Patent
Issue(s)
Hi,
FYI, here is the patent application's
description:
http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=%2Fnetahtml%2FPTO%2Fsearch-bool.html&r=1&f=G&l=50&co1=AND&d=PG01&s1=pardon.IN.&OS=IN/pardon&RS=IN/pardon
Guy
On woensdag, dec 10, 2003, at 15:07
Europe/Brussels, Greg Pavlik wrote:
Well, the LRA model
as I read it seems to be a relatively close mapping of sagas, which
were as far as I know first described in 1987 by Garcia-Molina. I
think it would be useful if someone took the action to provide a more
detailed breakdown of what exactly is at issue for the group, because
it's not obvious to me after a cursory review of the CheeTah papers.
Guy Pardon wrote:
Hi everyone,
The VLDB paper is
attached. It describes composite systems in the CheeTah prototype
(which served as the provisional application for the US patent).
Concerning WS-CAF,
I think there may be some overlapping with the LRA model described in
the TXM spec.
It was not my
intention to raise panic and 'warn' anybody, rather just to point out
that the patent exists (which is, I believe, required on behalf of TC
members).
Best,
Guy
On dinsdag, dec 9,
2003, at 20:27 Europe/Brussels, Mark Little wrote:
Robin, thanks. I
think a description from Guy about how he thinks any patent
relevant to this
paper may conflict with WS-CAF is in order though. Many
companies have been
doing composite applications/transactions for years so
I'd like to know
what aspect of CheeTah may either impact WS-CAF now or in
the future. This is
certainly the first I've heard of CheeTah so from an
Arjuna perspective
it certainly didn't influence WS-CAF, but maybe Guy's
email is intended
to just warn the TC in its future directions?
Mark.
----- Original
Message -----
From: "Robin
Cover" <>
To: "Newcomer,
Eric" <>
Cc: "Guy
Pardon" <>; <>
Sent: Tuesday,
December 09, 2003 7:08 PM
Subject: RE:
[ws-caf] Patent Issue(s)
Perhaps:
CheeTah: a
Lightweight Transaction Server for
Plug-and-Play
Internet Data Management
-------------------------------------------
Guy Pardon, Gustavo
Alonso.
-------------------------------------------
Abstract:
The ability to
maintain transactional interaction
in a distributed
system has proven to be a key feature
in information
systems. Unfortunately, as technology moves
towards more
distribution and decentralization, it becomes
increasingly
difficult to use existing transactional
tools. In fact,
current solutions are entirely unsuitable
for what we call
composite systems. Composite systems
can be
characterized as a collection of distributed,
autonomous
components, linked in an arbitrary configuration.
In this paper, we
describe CheeTah, a Java based set of
tools for building
composite components capable of
interacting
transactionally in arbitrary, dynamically
changing
configurations. We describe the technology provided,
how designers would
use it to build composite transactional
systems, and
examine in detail the performance of the
resulting solution.
Among the results we have achieved,
the performance and
the simplicity of use are of particular
interest.
http://www.inf.ethz.ch/department/IS/iks/publications/pa2000.html
http://www.inf.ethz.ch/department/IS/iks/publications/files/pa2000.pdf
-----------------------------------------------------
Robin Cover
XML Cover Pages
WWW: http://xml.coverpages.org
Newsletter:
http://xml.coverpages.org/newsletter.html
On Tue, 9 Dec 2003,
Newcomer, Eric wrote:
Guy,
To second Mark's
email, can you please also send additional information
about the patent,
such as the VLDB paper you mention.
Thanks,
Eric
-----Original
Message-----
From: Guy Pardon
[mailto:]
Sent: Tuesday,
December 09, 2003 9:53 AM
To:
Subject: [ws-caf]
Patent Issue(s)
Hi all,
As a TC member it
is my duty to report that there exists a US patent (US
patent application
09/682,969 -- finally approved about two months ago)
concerning
composite systems and an additional European application being in
process. This
patent is probably relevant to the CAF specs (and most likely
to BTP as well).
Basically, the
patent is an extension of the Composite Systems VLDB
paper we (Gustavo
Alonso and I) had at VLDB2000.
Best,
Guy
Dr. Guy Pardon
( )
Atomikos: Your
Partner for Reliable eBusiness Coordination
http://www.atomikos.com/
The information in
this email is confidential and only meant for the
addressee(s). The
content of this email is informal and will not be legally
binding for
Atomikos.
Dr. Guy Pardon
( )
Atomikos: Your
Partner for Reliable eBusiness Coordination
http://www.atomikos.com/
The information in
this email is confidential and only meant for the addressee(s). The
content of this email is informal and will not be legally binding for
Atomikos.
Dr. Guy Pardon (
)
Atomikos: Your
Partner for Reliable eBusiness Coordination
http://www.atomikos.com/
The information
in this email is confidential and only meant for the addressee(s). The
content of this email is informal and will not be legally binding for
Atomikos.
