[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Subject: Re: [ebxml-msg] Comments on ebMS_v1.04c
Arvola, Good point. I missed the point about being hung by the missing message. There is also another point. In order to deliver the messages in order, they must be persisted. That also argues that you need reliable messaging to do sequencing. I am looking forward to replies to my question about what ordering is good for. 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 ************************************************************************************* "Arvola Chan" <arvola@tibco.com> on 10/15/2001 11:20:41 AM To: Martin W Sachs/Watson/IBM@IBMUS cc: <ebxml-msg@lists.oasis-open.org>, <ebxml-cppa@lists.oasis-open.org> Subject: Re: [ebxml-msg] Comments on ebMS_v1.04c Marty: Thanks for pointing out my mis-statement about messageOrderSemantics being missing from the CPA. Requesting the receiving MSH to guarantee message order when reliable messaging is not used is problematical. If a receiving MSH receives a message with sequence number i before receiving the prior message with sequence number (i-1), it must wait for the latter message before it can deliver both messages to the application. However, if message (i-1) has not been sent using reliable messaging, it can be lost and may never arrive. As a result, message i may never become deliverable. In other words, guaranteed message ordering also implies that there should not be any sequence number gaps among the delivered messages. Regards, -Arvola -----Original Message----- From: Martin W Sachs <mwsachs@us.ibm.com> To: Arvola Chan <arvola@tibco.com> Cc: ebxml-msg@lists.oasis-open.org <ebxml-msg@lists.oasis-open.org>; ebxml-cppa@lists.oasis-open.org <ebxml-cppa@lists.oasis-open.org> Date: Sunday, October 14, 2001 7:38 PM Subject: Re: [ebxml-msg] Comments on ebMS_v1.04c Arvola, I have a couple of comments on your comments: Page 49, regarding messageOrdering without reliable messaging: It should be possible to deliver the messages in order without reliable messaging. Of course, some may be missing. More below. Page 50: messageOrderSemantics is in the CPA ver. 1.0. It is an attribute of the ReliableMessaging element. New MWS comment: I have never understood the value of messageOrderSemantics, especially since ordering is performed separately for each conversation. At today's state of the art, conversation choreographies that I am aware of have only one outstanding message at any time. The conversation moves to the next state on receipt of the response message. So, the choreography completely determines the order in which messages are received and doesn't need any help from the messaging service. Further, even if the choreography allows two business transactions to be executed concurrently, it is likely that the order in which the two transactions are completed is unpredictable and that the application does not care in which order the messages from the two transactions are received. I suggest deleting message ordering, thereby reducing complexity and execution cycles in the MSH. 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 **************************************************************************** ********* Arvola Chan <arvola@tibco.com> on 10/14/2001 06:54:26 PM To: ebxml-msg@lists.oasis-open.org cc: ebxml-cppa@lists.oasis-open.org Subject: [ebxml-msg] Comments on ebMS_v1.04c David: Please see the attached Word document (View Comments) for the contexts for these comments. Regards, -Arvola ---------------------------------------------------------------------------- ---- Page: 2 Fulfills Page: 2 Pete Wenzel (RosettaNet/SeeBeyond) is missing from this list. Actually, quite a few names found on http://www.oasis-open.org/committees/ebxml-msg/ are missing from this list. Page: 2 E2Open Page: 7 This section needs to be updated to reflect more closely the current round of modularization changes. Also, the use of punctuation after each bulleted item seems inconsistent. Page: 7 Invalid reference Page: 7 Invalid reference Page: 7 Invalid reference Page: 7 There is a section 11 with the title Multi-Hop Module. Appendix C is currently titled Supported Security Services. Page: 17 Chris Ferris has suggested that the namespace value should match the schemaLocation value. The examples need to be updated accordingly if we agree to adopt that convention. Page: 23 Shouldn't the TimeToLive element be present if Reliable Messaging is in use? Page: 23 How can the equivalent of BestEffort be specified? In that case, the semantics is not at-least-once. Page: 25 I find the reorganization of the Security section rather awkward. Shouldn't the section on Security and Management precede the sections on Signature element and on Signature Generation. It does not make sense talk about Signature Generation even before Persistent Digital Signature is introduced as a countermeasure technology. Page: 27 This is insufficient. If intermediaries can change the AckRequested and Via elements, then these elements have to be excluded from the signature. Page: 28 This section needs to be updated to reflect changes in the 1.1 CPP/A spec. Currently, the CollaborationRole element contains a CertificateRef element that is also used for security definition. Page: 30 Should www.ebxml.org be replaced with www.oasis-open.org? Page: 31 This should probably be replaced with an appropriate XPointer. Page: 33 What happens when the incoming message specifies syncReply=true? Shouldn't errors be returned synchronously in this case? Page: 33 I suggest moving this section to the end of Part II Optional Features. Alternatively, the introductory paragraph in this section should be updated to indicate that optional elements from Part II can also be present. Page: 35 Incorrect statement. Page: 35 Should this be replaced with oasis-open.org? Page: 36 I don't see why this is necessary. A DeliveryReceipt is generated by the To Party MSH and should be returned at the latter's earliest convenience. It should always be possible to piggyback the DeliveryReceipt element with the first application level message that is in response to the message in question and therefore be able to share the same RefToMessageId in MessageData. If the DeliveryReceipt is sent on its own, again the RefToMessageId in MessageData should suffice. Page: 36 This contradicts lines 1397 and 1398. Page: 36 Duplicate elimination may be a more meaningful phrase in this context. Page: 37 Messages that are sent in response to messages arriving on a reliable but synchronous delivery channel also need to be kept in persistent storage! Page: 37 Is the presence of a Via element mandatory when there are two instances of the AckRequested element? Page: 37 This suggests that the From Party MSH can request the To Party MSH to return an end-to-end acknowledgment, thereby rendering the DeliveryReceipt element redundant. I would recommend getting rid of the DeliveryReceipt element. Page: 38 Should this be replaced with oasis-open.org? Page: 38 An. Page: 38 Included in the same message with. Page: 38 What is this attribute? Page: 38 Shouldn't this be the From Party? Page: 38 Should this be "used by one Message Service Handler to indicate to another Message Service Handler that it ?" instead? Page: 39 Again, I question whether this is necessary. See my earlier comments regarding the use of a RefToMessageId attribute within the DeliveryReceipt element. Page: 39 Is the presence of a Via element mandatory when there are two instances of the Acknowledgment element? Page: 39 Incorrectly cut and pasted. Page: 39 Corresponding to the ? Page: 39 Isn't it true that an Acknowledgment element an an AckRequested element cannot be present in the same ebXML message if the latter is not generated by the application layer? Page: 40 Please note that Retries, RetryInterval, PersistDuration are only in the CPA, not in the message header. Page: 40 Spelling. Page: 40 Spelling. Page: 40 Spelling. Page: 40 Spelling. Page: 40 Is there such an element? There is a duplicateElimination attribute in the QualityOfServiceInfo element. Page: 40 See comment above. Page: 40 Has received? Page: 40 There are two kinds of retries: retry due to non receipt of Acknowledgement from the nextMSH, and retry due to non receipt of Acknowledgment from the ToPartyMSH. The CPA only deals with the latter type of retry. Page: 40 See above comment on Retries. Page: 41 This statement is imprecise. TimeToLive is of type dateTime whereas the product of Retries and RetryInterval is of type duration. Page: 41 This should be of type duration. Page: 41 The timestamp for a reliably sent message (found in the message header), plus its PersistDuration (found in the CPA), must be greater than its TimeToLive (found in the message header). Page: 41 All of the figures in this section illustrate the single-hop case. It should be made clear whether the discussions apply both to the single-hop and multi-hop cases. Page: 42 This statement is very confusing. Why should it be either or? It seems that if the sender is aware of the use of an intermediary, then it should specify two AckRequested elements. The phrase "as determined by the AckRequested parameter" is rather meaningless. Page: 42 Are we waiting for the Ack from the nextMSH or the ToPartyMSH? Page: 42 Again, are we talking about an Ack from the nextMSH or an Ack from the ToPartyMSH? Page: 42 Isn't it the case that only the ToPartyMSH is supposed to perform duplicate elimination? Page: 42 Intermediate MSH's should not have to do this. Page: 42 If the CPA indicates that an application response is included, then it must be the case that the processing of the earlier message is not yet complete. In this case, the MSH must wait for the response to the earlier message and then return it synchronously. Page: 43 I still think that either all intermediaries, or none of them, should exhibit store and forward (with retry) behavior. Page: 43 Earlier discussions in this draft indicate that an Acknowledgment message must contain an Acknowledgment element. See line 1443. Page: 43 I think we should allow for the possibility of having the Acknowledgment message returned synchronously without being piggybacked on the business response. Page: 43 Should this be oasis-open.org? Page: 43 This works only for the ToPartyMSH. Need to explain how an intermediary should populate the From andTo elements when it needs to generate an intermediate Acknowledgment because sync reply is not in use. Page: 44 Need to account for the fact that there are two kinds of retries and two sets of retry parameters. Page: 44 Strike out this phrase. Page: 44 I suggest renaming this as "first response message" Page: 45 First response message. Page: 45 The From Party MSH many have no knowledge of an intermediate MSH. Why should it trust a Delivery Failure Notification from an untrusted intermediate MSH? Page: 46 Should this be oasis-open.org? Page: 49 Should this be oasis-open.org? Page: 49 Should this be oasis-open.org? Page: 49 Use. Page: 49 This is quite different from the 1.0 spec which states that the SequenceNumber element must not be present unless deliverySemantics is OnceAndOnlyOnce and messageOrderSemantics is Guaranteed. Message ordering cannot be guaranteed unless messages with sequence numbers are guaranteed not to be lost! Page: 50 I don't think messageOrderSemantics is currently specified in the CPA. We may have to provide feedback to the CPP/A TC. Page: 51 Does this rule apply only to the ToPartyMSH or does it apply to the nextMSH as well? Page: 51 Should the AckRequested element that is addressed to the nextMSH be placed under the Via element? Page: 52 Should this be www.oasis-open.org? Page: 52 Don't need to capitalize? Page: 52 Forwarding the message to ? Page: 52 The sending MSH may not be aware of the use intermediate MSH's and therefore not include a Via element. When the first MSH forwards the message to the second MSH, it will have to construct a Via element that contains a TraceHeaderList with Two TraceHeaders! Page: 55 The sending MSH may not be aware if any intermediary is used in the message path. Therefore, the decision to exclude the Via element from the signature must not rely on knowledge of the presence or absence of intermediaries. Page: 56 The locations of these schemas have to be updated.
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Powered by eList eXpress LLC