A couple of comments below. 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
*************************************************************************************
Dan Weinreb <dlw@exceloncorp.com> on 11/14/2001 11:59:23 PM
Please respond to Dan Weinreb <dlw@exceloncorp.com>
To: ebxml-msg@lists.oasis-open.org
cc:
Subject: [ebxml-msg] Proposed "idempotent" parameter and attribute and
CPA element
Here's the proposal that I said I'd provide:
This proposal changes the "duplicateElimination" attribute to the
"idempotent" attribute. Note that the sense is reversed; that is,
loosely speaking, idempotent being false means the same thing as
duplicateElimination being true.
Very unfortunately, the current CPA spec has an element called
"idempotency", which is TRUE if the message IS NOT idempotent, and
FALSE if the message IS idempotent. (The CPA spec uses the phrase
"idempotency check" to mean "duplicate check".) We propose that the
CPA group change the name of the element to "idempotent", with the
sense that the value is true if and only if the message is idempotent.
MWS: I agree that the element in the CPA is mis-named. However, the
intent was for TRUE to mean that duplicate checking is required. In
this sense the element is redundant with the deliverySemantics attribute
and
could have been eliminated if nothing else changed.
MWS: I'm getting the sense that the proposal is tantamount to doing
reliable messaging with optional duplicate elimination. That may also be
OK but is not onceAndOnlyOnce. So if the deliverySemantics attribute is
still alive, some new value is needed for reliable messaging without
duplicate elimination. In that case, the idempotency attribute is still
redundant and could be eliminated.
MWS: All in all, the amount of restructuring taking place in the area
of reliable messaging impresses me as being well beyond the scope of a
maintenance release and should be deferred to version 2.
The following text is proposed to replace sections 3.1 and 7.4.1; in
accordance with earlier decisions of the MS TC, the
QualityOfServiceInfo element of MessageHeader is removed, and the
"idempotent" attribute is "promoted", so to speak, to the
MessageHeader element.
This text assumes that the CPA definition adopts our suggestion about
the use of the "perMethod" value. If the CPA does not adopt this
suggestion, the rules at the end of 7.4.1 must be changed accordingly.
3.1. idempotent attribute
Valid values for idempotent are:
true - the To Party MSH MAY assume that the message has idempotent
semantics
false - the To Party MSH MUST NOT assume that the message has idempotent
semantics
MWS: This definition sounds like the idempotent attribute is completely
divorced from the reliable messaging protocol and applies equally well to
bestEffort.
Is that what is intended?
The definition of "idempotent", and the interaction of this attribute with
the "idempotent" element from the CPA, are described in section 7.4.1.
7.4.1 idempotent
A message is said to be "idempotent" if processing the message more
than once has the same effect as processing the message once.
If the value of the "idempotent" parameter for a message is "true",
the To Party MSH MAY safely assume that the message has idempotent
semantics. If a duplicate idempotent message is received, i.e. if the
same idempotent message is received more than once, the To Party MSH
MAY deliver the message to the application more than once. (However,
the To Party MSH is free to perform duplicate elimination even knowing
that the message has idempotent semantics. This might be wise if the
cost of re-processing the message exceeds the cost of duplicate
elimination.)
MWS: I agree with this definition of idempotent. As noted above, I
am concerned about how it relates to reliable messaging and about the
amount of redefinition of reliable messaging taking place in a
maintenance release.
If the value of the "idempotent" parameter is "false", the To Party
MSH MUST assume that the message might not be idempotent. If a
duplicate non-idempotent message is received, the To Party MSH MUST
not deliver the message to the application more than once. That is,
the To Party MSH must eliminate duplicates.
The value of the "idempotent" parameter is controlled by the
"idempotent" element of the CPA together with the "idempotent"
attribute of the MessageHeader, as follows:
-- If the value of the "idempotent" element of the CPA is "true" or
"false":
then the value of the "idempotent" parameter is "true" or "false"
respectively. Furthermore, if the "idempotent" element is present in
the MessageHeader, it must agree with the parameter value, or else the
To party MSH reports an error with errorCode set to Inconsistent and
severity set to Error.
-- If the value of the "idempotent" element of the CPA is "perMessage":
If the "idempotent" attribute of the MessageHeader is "true", the
idempotent parameter is "true", otherwise the idempotent parameter
is "false".
----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>