[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Subject: RE: New work and loose ends
Dale, my replies embedded below. 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 ************************************************************************************* Dale Moberg <dmoberg@cyclonecommerce.com> on 07/20/2001 12:13:30 PM To: Martin W Sachs/Watson/IBM@IBMUS, Karsten Riemer <Karsten.Riemer@sun.com>, ebxml-cppa@lists.oasis-open.org cc: Stefano POGLIANI <stefano.pogliani@sun.com>, chris.ferris@east.sun.com, "Riemer, Karsten" <Karsten.Riemer@east.sun.com> Subject: RE: New work and loose ends I think of this issue as the Conversation Correlation Handle problem. In other words, how does the MSH know what conversationId to stick on a response when it just receives a response payload from an app? MWS: The MSH doesn't have to keep track of what conversationId to stick on a response. This is a piece of information that should come down from the application with the reply message payload. Of course that isn't spelled out anywhere because the Messaging Service spec doesn't include an abstract service interface for the MSH. The conversationId in a reply is, of course, the same as the conversationId in the message it is replying to but there is no reason for the MSH to keep that kind of long term state information when the upper levels have that information. MWS: If there is a concern about Party A understanding a conversationId supplied by Party B, one answer is that PartyA establishes a conversation generates a GUID as the conversationId, maps it to whatever local handle it uses, and sends the GUID in the message. PartyB generates a mapping of that conversationId to whatever local handle it is using. The conversationId created by PartyA is the one the appears in all the messages of the conversation. This approach matches the current MSH spec, which has only one conversationId and requires it to be unique to the two parties. In general "notification from app" and "submission to app" components of a BSI are important for MSH to "app" integration, but I am not sure what assistance the classical CPA and CPP provides in easing this burden. MWS: I am not sure whether the above paragraph is specific to conversations or is a general statement about signalling between the MSH and the app. I view that as a MSH service interface matter, not a CPA matter. As to conversations in the CPA, today, the CPA doesn't deal with conversations in a normative way (except for concurrentConversations), nor should it. The BPSS or any other collaboration protocol will have either an implicit or explicit abstract definition of what goes into a conversation and it is up to the application implementation to define the conversation boundaries and pass down begin- and end-conversation signals to the middleware. Of course, we may run into conversation semantics issues that do require help in the CPA but so far we haven't. The CCH problem is possibly "flavored" by the value of the concurrency allowed (I am forgetting the tag name here), MWS: The concurrentConversations attribute of the ConversationConstraints element. This is meant only as an implementation assist to prevent one party from overloading the other. MWS: I'm not sure what the CCH problem is. conversationId is the handle and its value is stated (in the TRP spec) as an implementation matter. I suppose that an implementation could restrict concurrentConversations to limit the number of conversation handles but that seems unlikely in this era of Gbyte main memories. but that is the main impact what we have represented has on this problem, that I recall right this second. I think that the question of what we as a group would do about formulating a BSI contract should definitely be a distinct item. It would be taking on something new, but certainly in line with promoting "easier, better, faster, interoperable" b2b integration methods. MWS: The term "BSI contract" is a little too abstract for me. I believe that what we (CPPA+BP+MSG) need to define is a software interface between the legacy application bridge and the top of the BSI and an abstract interface between the MSH and the rest of the BSI. Is that what you mean by "contract"? I also believe that these interfaces should not be APIs but should describe function and leave the middleware implementers to define the specific APIs. I understand the counter-argument that a normative definition of the interface between the bridge and the BSI will enable application code to run on everyone's BSI but I don't think that that argument applies to the BSI-to-MSH interface. I do not expect MSH implementations to be sold separately from the whole middleware system. It is certainly worth raising as a new potential work item either by you or Karsten or both. -----Original Message----- From: Martin W Sachs Sent: Fri 7/20/2001 8:11 AM To: Karsten Riemer; ebxml-cppa@lists.oasis-open.org; Dale Moberg Cc: Stefano POGLIANI; chris.ferris@east.sun.com; Riemer, Karsten Subject: RE: New work and loose ends Karsten, Yes we should have at least an initial discussion next week. I think it fits into one of the "loose ends and new function" topics that are already on the list to be discussed. If not, bring it up wherever it seems to fit in. Dale, do you think we should fit this in as a separate agenda item? 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 ************************************************************************ ************* Karsten Riemer <Karsten.Riemer@sun.com> on 07/20/2001 09:43:54 AM Please respond to Karsten Riemer <Karsten.Riemer@sun.com> To: Stefano POGLIANI <stefano.pogliani@sun.com> cc: Martin W Sachs/Watson/IBM@IBMUS, chris.ferris@east.sun.com, "Riemer, Karsten" <Karsten.Riemer@east.sun.com> Subject: RE: New work and loose ends Stefano, I will let Marty reply about the TRP level, since I am not aware of what state a transport client is expected to keep. I am also copying Chris Ferris so he can comment, too, from a TRP point of view. As to BSI, we were having discussions at yesterday's BPSS meeting, and several people, including Bob Haugen were asking for you, now that we seem to be ready to adress BSI issues. We will put a 'proposal' for a BSI workgroup to the UN ADHOC group to be formed possibly next week, but I really think BSI work should happen either in OASIS (maybe inside CPP team), or at the very least in a joint OASIS/UN group, not in a UN-only group. Marty, can we discuss this in Phoenix next week? -karsten >Marty, Karsten, > > just another little topic, perhaps trivial or perhaps not applicable. >For this reason I address it ONLY to you two before sending to the whole >list. >If you think it is valid, feel free to express it in plain english and >distribute. > >I have been thinking since a while to a possible source of discrepancy >between the TRP and the CPA. > > - The TRP, because of the QoS issues, may be proned to "keep a state" > about the exchanged messages. > The focus of TRP is, though, on THE "message", independently > on the fact that this message is actually part of a choreography > > - The CPA gives origin to the BSI. The BSI actually manages the > runtime, i.e. the BSI manages the choreography. > In my opinion, it is likely that the mgmt of the choreography > requires the maintenance of a "state" at choreography-level > (something, at least, to be able to understand at which > point in the long-living collaboration the process is in) > >Now, keeping these two "state mgmt" stuff separate risks to be source >of confusion and inconsistency. > >Am I missing something? > >Best regards > >/stefano > > >> -----Original Message----- >> From: Martin W Sachs [mailto:mwsachs@us.ibm.com] >> Sent: 19 July 2001 20:37 >> To: ebxml-cppa@lists.oasis-open.org >> Subject: New work and loose ends >> >> >> I have attached two documents that discuss potential maintenance and new >> work items that I will review next week. >> >> Regards, >> Marty >> >> (See attached file: CPA-CPP-changes.pdf)List of work items as of >> the end of >> the Vienna ebXML meeting. This includes additional discussion of >> some items >> in the >> CPPA.new.work document below >> >> (See attached file: CPPA.new.work.pdf)Summary of all proposed work and >> loose ends as of today. I will be using this as slides next week. >> >> ****************************************************************** >> ******************* >> >> 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 >> ****************************************************************** >> ******************* >>
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Powered by eList eXpress LLC