David,
Some comments below, as "MWS".
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
*************************************************************************************
David Fischer <david@drummondgroup.com> on 11/01/2001 10:21:37 AM
To: Dan Weinreb <dlw@exceloncorp.com>, ebxml-msg@lists.oasis-open.org
cc: ebxml-cppa@lists.oasis-open.org
Subject: RE: [ebxml-msg] Attributes specified in both the message and
the CPA
Hi Dan,
I think you have made an assumption which is at the core of this problem.
You
have assumed that there is a pre-agreement. Many in this group agree with
you
but some don't.
Take the case, for example, of someone requesting a stock quote. This may
be a
one-time transaction, and a rather simple one at that. Why go through all
the
trouble of negotiating a CPA?
MWS: You still have to get both ends' middleware to agree on how to
exchange
the message. What you need is to be able to negotiate and deploy the CPA
so rapidly that it can be thrown away after one use. Some people call it
"dynamic eCommerce". Automated negotiation is a step in that direction.
Take the case, for example, of someone purchasing a sweater from L.L.Bean.
This
is a B2C transaction and again, may be one time. Why bother with a CPA
negotiation which takes longer than the transaction?
MWS: My comment above applies. However let's face it. B2C works quite
well
today using browsers. What we are really after is dynamic B2B eCommerce.
When we agreed not to require a CPA, we agreed not to require it. When we
ask
the question, "how does this work without a CPA?" we have to assume there
is no
CPA and no "virtual" CPA. Does this mean we have no configuration? Of
course
not, it only means there is no negotiation between the parties.
MWS: Excuse me. You may call it "negotiation" or not but the two parties
still have to understand each other's IT requirements or the message won't
flow.
What then is
required in the message headers to accomplish this task? Many things, like
persistDruation, retries, retryInterval, etc. can be given default or
"bootstrap" values.
MWS: Yes, you could throw the entire equivalent of the CPA into the
message
headers but you thus guarantee that two different implementations are
required,
one with and one without CPAs.
Even the CPAId can be given a default -- like a credit card
number -- or the sender can pick one and the receiver continues to use it.
So
we come down to -- what can't be given defaults? These are going to be
transmission parameters such as syncReply, and parameters which may change
per-message, like AckRequested. These are the things we need to maintain
in
both the MessageHeader and the CPA (my personal opinion is that
syncReplyMode
should be removed from CPA).
MWS: OK, I think we more or less agree in these last two paragraphs.
Someone
else can comment on whether it makes implementation
sense to dynamically change between sync and nonsync
(or ack requested and no ack requested) for the same action. It
makes perfect sense to define some actions as sync and others non-sync.
Different messaging characteristics can be accounted for by defining
different
actions (i.e. different business transaction activities) in the BPSS spec.
That's
why the CPA provides for specifying a different delivery channel for each
action.
We also MUST allow ebXML-MS to work with a CPA -- which will probably be
the
most common case. To do this, we agreed NOT to override CPA values with
values
in the MessageHeader.
Regards,
David Fischer
Drummond Group.