Forwarded message 1
- From: www-ws-request@w3.org <www-ws-request@w3.org>
Sent: 16 May 2003 18:53
Hi, Just catching up with some maile thread about the process model as defined in DAML-S is interesting and unfortunately there is not enough information on the DAML-S site that describes how the composition process will be handled. I would like to expand a bit on the middle part of the reply by Massimo to get a clearer picture. In particular, I am referring to this part: >Also, during composition of web services clients need to know details on how the web service >performs its tasks so they can integrate different services. For example, before spending >some money a client may want to make sure that there is money in an account. This requires >careful composition of the process model of the selling Web service (which withdraw the money >from the account) with the process model of the Web service of the Bank that holds the account >(which tells if there is enough money). I am assuming that a matchmaker will return to the user a list of selling and banking services in reply to the user's query. If we are to assume an autonomous composition of web services it is of utmost importance that either the user or especially his agent have access to the process model of each service so that the workflow of composed services will be defined. In the case of the user agent, some form of reasoning in the agent will have to identify that composition of web services is required and a process starts with the aid of a composition engine to handle such integration. This I would imagine will be based on two major issues, the IOPEs of each service and also the control constructs defined in the process model. What is still an enigma and also a bit far fetched in MHO, presuming that the above argument is valid, is the reasoning ability of the agent required to collaborate with the composition engine (again I am assuming that such an engine will be available for example as a web service itself) to perform the tasks as described above, in the correct order of execution. I would imagine that the reasoning part should be very powerful to perform such types of inferencing. Even if the user performs this workflow definition manually, I would imagine that he would have to be knowledgeable about the area of web services and all that comes with it. My assumptions also include that such a scenario will require a solid communication framework between agent and engine, such that the services are composed in their exact order and execution state. The result of such a composition could be a new process model that is generated as a result of this composition and possibly cached for future use. Does the argument above make sense? Am I missing something? I would appreciate any feedback Regards, Charlie ------------------------------------------------- Charlie Abela Research Student, Dept. of Computer Science and AI University of Malta, MSD06. Malta Web: http://alphatech.mainpage.net -----Original Message----- All email is scanned by Keyworld against known Viruses. This service is offered to all Keyworld subscribers and hosted domains and does not carry any warranty. You are advised to protect your PC with updated antivirus software at all times.
Sent: 16 May 2003 18:53
Sudhir, There are many reasons why a Web services may want to display its process model. The first one is that its clients can derive the interaction protocol from the Process Model. For example a book selling Web service like Amazon allows you to browse and to reserve the book (amazon's web service does not allow to buy right now). so you could have a process model that looks like sequence(browse-atomic-process, reserve-atomic-process). The client at this point knows that first it has to deal with the browsing task, then with the reserve task. Acting differently from what the process model require will lead to errors. But there are other reasons. I understand that supply chain management depends on the providers to provide information on the state of processing of the different orders. Ultimately this requires the providers to display parts of their process model. Also, during composition of web services clients need to know details on how the web service performs its tasks so they can integrate different serices. For example, before spending some money a client may want to make sure that there is money in an account. This requires careful composition of the process model of the selling Web service (which withdraw the money from the account) with the process model of the Web service of the Bank that holds the account (which tells if there is enough money). Ultimately, the DAML-S Process Model allows implementors of Web services to display as much of the Web service they want to display. If they do not want to display anything, they use only one atomic process. Minimally, they have to display enough information to allow their clients to derive the interaction protocol, and in a way this is what e-shops do already (amazon or orbitz ask you quite a few questions before letting you to buy something) everything else it is up to their business model and what they want to achieve. I hope that this helped, --- Massimo Sudhir Agarwal writes: > > Dear all, > > i currently can not understand the purpose of the ProcessModel completely. I > understand why an AtomicProcess is needed. But, why does ComplexProcess > exist? Isn't it enough to have only AtomicProcess? Why should a web service > provider show how his services works? On the other hand, im not sure that a > web service requester is interested in knowing all that (if-then-else, while, > split, fork etc.) stuff as long as the service does what he wants. Even if > someone really wants to know that, what can he do with that knowledge? Does > it help him in any way? > > could someone help me? > > Thanks and regards > > Sudhir >
Received on Tuesday, 3 June 2003 13:16:36 UTC