Minutes of CPA Negotiation Conference Call Mar. 20, 2002
Marty Sachs
Mar. 2, 2002
Attendees: Marty Sachs, Dale Moberg, Heiko Ludwig, Himagiri Mukkamala, Jean Zheng.
Dale Moberg will host the April conference calls.
Hima is analyzing the swapping of roles in the negotiation process that we previously discussed. He is creating a Negotiation CPA that can express swapping of roles.
We had a preliminary discussion of Kartha’s NDD writeup. Everyone please review it and post comments to the list server.
Heiko pointed out that the writeup discusses the negotiation subteam deciding what is negotiable. Marty replied that anything in the CPP should be negotiable and a particular NDD will state what is negotiable for a particular CPP. Dale added that we will need to specify things such as where enumerations, ranges of numeric values, etc, are needed. Marty stated that we will also have to define how the NDD refers to specific elements and attributes in the CPP (for example, using Xpath or Xlink).
Dale wondered if the BPSS instance would be negotiable. Marty suggested that agreement on one BPSS instance out of several possibilities could be useful but we should not get into negotiating the contents of a BPSS instance.
Dale also wondered if roles are negotiable. For example, there might be business processes where either party could play either role.
Jean reported that she has discussed the patterns with Brian Hayes and those discussions will continue.
Brian’s initial draft has party A always sending offers and Party B always sending counter offers. We will need to be sure that this is sufficiently flexible.
The draft assumes that offer and counter offer have the same structure, i.e. the same schema.
Jean pointed out that we will need some kind of metadata that link a counter offer to the corresponding offer and to support retry after message losses.
We will probably need a unique ID associated with each offer or counter offer. In addition, the counter offer will have to state the ID of the offer it is responding to.
Hima suggested including an ID of the CPA being negotiated. This would be a constant value for the duration of the negotiation process, not necessarily the CPAId in the CPA. Marty pointed out that the conversation is an additional handle for this purpose. As has been discussed on the list server, we should not assume that the message-level conversationId would be available to the application. However, if our design relates one instance of the negotiation process with one conversation, then we can simply define that a single conversation negotiates only one CPA, in which case the ID of the CPA under negotiation is implict in the process instance.
Dale reminded us that we previously decided also to have a lightweight protocol built on basic SOAP. We should put together a list of requirements on this.
Jean said that we still need to decide whether to send a complete draft CPA back and forth during the negotiation process. Doing so (rather than just exchanging the fragments involved in a particular iteration) should reduce the amount of negotiation state information that each party has to manage but it would also requiring each draft CPA by an NDD that identifies what is changed in the draft.
Respectfully submitted,
Marty Sachs
Metadata
Please notify the team of any corrections needed in the minutes.