- From: Ryusuke Masuoka <rmasuoka@fla.fujitsu.com>
- Date: Wed, 6 Nov 2002 18:15:27 -0500
- To: "'Sheila McIlraith'" <sam@KSL.Stanford.EDU>, "'David Martin'" <martin@ai.sri.com>
- Cc: "Bijan Parsia" <bparsia@email.unc.edu>, "Evren Sirin" <evren@cs.umd.edu>, "Michael H. Grove" <mhgrove@fla.fujitsu.com>, <www-ws@w3.org>
David and Sheila, It was very nice for me to be able to talk to you at DAML PI meeting. I have summarized my comments on DAML-S 0.7 and attached at the end. Some were pointed out at the meeting and some were not. Since we are using DAML-S as our crucial element in our project, its success really matters us. Regards, Ryu ----- Ryusuke Masuoka, Ph.D. Senior Researcher Fujitsu Laboratories of America, Inc. 8400 Baltimore Avenue, Suite 302 College Park, MD 20740-2496, U.S.A. Tel: +1-301-486-0398 (Main) +1-301-486-1564 (Direct) Cell: +1-301-237-3309 Fax: +1-301-441-9676 Email: rmasuoka@fla.fujitsu.com URL: http://www.flacp.fujitsulabs.com/~rmasuoka Latitude: N 38.9935 Longitude: W 76.9330 ------------------------------ Comments on "DAML-S: Semantic Markup for Web Services" ------------------------------ - Grounding: I think putting grounding information in WSDL is not a very good design decision (6.2) It should be in DAML-S. (It is not a "grounding" but a "ceiling"!) I understand this WSDL grounding is a very first shot and an example, but in regard of its importance to us, please let me make comments on it. It also depends on a specific feature (abstract type) of WSDL. This approach will not be applicable for other invocation methods, such as UPnP, CORBA, RMI. (Okay, there is a possibility that you can use WSDL for all of those technologies. But I think it is unlikely to require WSDL for the users of DAML-S to make a call into UPnP. I believe it should be made it possible to call directly UPnP from DAML-S description.) Things usually go from more abstract to more concrete. Because when a more concrete thing (ex. WSDL) was created, they had no knowledge of a more abstract thing (ex. DAML-S) to come. Usually WSDL's are automatically created by the system (sometimes dynamically), and it would be difficult to modify them. There will be many WSDL/SOAP, UPnP, CORBA, RMI services before DAML-S will be deployed. I strongly believe it is better off if there are ways to describe groundings to those services implementations in DAML-S. - Grounding example to existing Web services: I believe giving a semantic markup for an existing Web service (ex. Amazon.com, MS .NET My Services) and show how it can be invoked with semantics objects (ex. credit card and book instances) will elucidate how DAML-S can be used. - Why do IOPE (input, output, precondition, effect) appear both in Profile and Process? It makes very difficult to implement systems if they appear in both places and if "DAML-S does not dictate any constraint between Profiles and Process Models, so the two descriptions may be inconsistent without affecting the validity of the DAML expression" (Isn't it an implementer's nightmare?) Is this because of the limitation of expressiveness of DAML? Is there going to be some way to provide the constraints in natural language? When we have DAML Rule, do we have some constraints on those? I want that information to appear in one place and only once in DAML-S description. There has been given some rationale in 4.1, but it can be solved by registering a semantic service description without parts not to be advertised. - Is it possible to describe "Macros" in DAML-S? Not just composite services, which consist of specific services. - In 4.2.4, the definitions of precondition and effect: Shouldn't their values be "thing"? - Why geographicRadius and qualityGuarantee are deprecated (4.3) I understand that they are moved to serviceParameters for specific kinds of services. But I still want to see them in DAML-S. For pervasive computing, "geographicRadius" is of crucial importance. "qualityGuarantee" is also important factor for the user to decide whether execute the service. I believe there should be serviceParameters for cost and authentication. They are also very important factors for the user to decide whether execute the service. - 5.1 -> CompositeProcess -> Unordered I believe "Z = (Unordered A B)" should be "Z = (Unordered X Y)" ------------------------------------------------------------
Received on Wednesday, 6 November 2002 18:19:03 UTC