- From: Andrew Layman <andrewl@microsoft.com>
- Date: Mon, 8 Apr 2002 12:34:22 -0700
- To: "David Martin" <martin@AI.SRI.COM>, "Anne Thomas Manes" <anne@manes.net>
- Cc: "Timothy N. Jones" <tim@crossweave.com>, <thomi@di.uoa.gr>, <www-ws@w3.org>
Whether a Web service is invoked using a programmatic API is immaterial to its being a Web service. It is much more important that the messages effecting the invocation have a definite, language-neutral and platform-neutral format. That is, Web services are defined in terms of their protocols, not the APIs used for invocation, which may vary from platform to platform or language to language. -----Original Message----- From: David Martin [mailto:martin@AI.SRI.COM] Sent: Monday, April 08, 2002 11:38 AM To: Anne Thomas Manes Cc: Timothy N. Jones; thomi@di.uoa.gr; www-ws@w3.org Subject: Re: potential users of web services Anne Thomas Manes wrote: > IMHO, a web service is a component that is invoked using a programmatic API, ... There seems to be a consensus on the idea that a web service is invoked using a programmatic API, and I certainly agree that's a central element of what we mean by "web service". But it's worth noting that there is are interesting issues to consider, even with this relatively well-defined notion of programmatic API. I'm referring to the idea that a service can have both an offline component and an online component, and we need to consider the full range of variability here. Some services can take place predominantly online, with a minor offline component - For example, one could have an elaborate online buying process, including such things as getting ratings on potential sellers, arranging financing, etc., where the only offline component might be shipping the item purchased. One can also imagine services that happen completely online, except that they require just a moment or two of a human's attention to give final approval to the transaction. And so forth. some can take place predominantly offline, with a minor online component - For example, one might order a pizza online. In this case, of course, the pizza preparation and delivery (and possibly even the payment, if you pay cash) take place offline. Also, what about the case where one has a software agent to locate a pizza restaurant that delivers, but then one calls up that restaurant and places the order by telephone? Does that count as a "web service", given that it was found using a programmatic API? (I suppose it depends on whether this programmatic API is provided by the pizza restaurant or by a third party directory service, but I'm not sure.) and one can imagine a wide range of in-between values for this "online/offline ratio". Although I don't think it's really necessary to arrive at a precise definition of "web service", still, it's probably useful to think about which services on this online/offline spectrum should be regarded as "Web services". Should it be any service with a programmatic API for an online component (even if that component represents only a minor part of the service), or must the service take place completely online to be regarded as a "Web service", or should there be some sense in which the service is "invoked" online (not just "found" online), or what? In our work on DAML-S, we are tending towards the view that includes the broadest range of services - but I can't claim that we have a clear-cut definition of this, or ever will have one :-). David Martin SRI International Anne Thomas Manes wrote: > IMHO, a web service is a component that is invoked using a programmatic API, > hence it is not a user-driven application. If a web service determines that > it needs more input, then either it returns an error or it contains some > logic that allows it to obtain more input. It can obtain that input by > initiating an interaction with a user (which sounds to me like it's still > service-driven, not user-driven), or it might queue a request up on a > workflow process, or it might call some other service or program that knows > how to obtain the input. > > I'm not saying that you can't build a UI-driven application with web > services. What I'm saying is that web services themselves aren't UI-driven. > > I believe that it's important to distinguish between a web service and an > application that has been constructed using web services. Let me illustrate > with an analogy. A database is an example of a service. You communicate with > it using a programmatic API (SQL). The database isn't a UI application, but > you can certainly build UI applications that utilize the database service. > > Anne > > > -----Original Message----- > > From: Timothy N. Jones [mailto:tim@crossweave.com] > > Sent: Thursday, April 04, 2002 2:50 PM > > To: Anne Thomas Manes > > Cc: thomi@di.uoa.gr; www-ws@w3.org; www-ws-arch@w3.org > > Subject: RE: potential users of web services > > > > > > > > I would have to object to the separation of UI and web service described > > below on the grounds that it appears incompatible with the notion of > > interactive web services. For example, it seems reasonable that at some > > point along a web service call chain a service could dynamically determine > > that it needs additional information from the user to complete its task. > > Unless UIs are able to themselves be web services I think the WS > > architecture will not be able to support user-driven applications, which > > would be an unfortunate limitation. > > > > Regards, > > Tim Jones > > CrossWeave, Inc. > > > > > Thomi, > > > Ask five people the definition of "web service" and you'll get six > > answers. > > > I generally describe a web service as a service that > > communicates over the > > > web. A service is a component that exposes a programmatic interface. The > > > service interface must be described; and the service implementation must > > be > > > discoverable. > > > When you relate this abstract definition to current > > technologies, you can > > > implement a web service by creating a service that exposes a SOAP > > interface, > > > which is described by WSDL, and which is registered in UDDI. But I > > wouldn't > > > want to use current technologies to *define* the basic concept. I also > > don't > > > think that it's essential to use any of these technologies to > > create a web > > > service. I can certainly create a web service using XML-RPC or > > RosettaNet > > or > > > a host of other technologies. > > > That said, I would concur that web services are intended to be > > consumed by > > > applications rather than humans. But keep in mind that a user > > interface is > > > an application. If I wanted to arrange food for 500 people for two weeks > > in > > > Dubai, I would use a catering application, which in turn uses > > web services > > > to find caterers that can provide services in Dubai. The UI > > isn't the web > > > service. The UI uses web services to accomplish its work. Hence an ASP > > page > > > or HTML form aren't web services, they are an interface to web services. > > > Best regards, > > > Anne Thomas Manes > > > CTO, Systinet > > > > -----Original Message----- > > > > From: www-ws-request@w3.org [mailto:www-ws-request@w3.org]On > > > > Behalf Of Thomi Pilioura > > > > Sent: Thursday, April 04, 2002 9:09 AM > > > > To: www-ws@w3.org > > > > Subject: potential users of web services > > > > > > > > > > > > Hi all, > > > > > > > > I'm little confused about the notion of the term "web services". > > > > When I'm reading papers related to UDDI,WSDL,SOAP they present > > > > web services > > > > as a new age of distributed computing and as such they are only useful > > to > > > > developers (who are trying to build web applicattions) and not to the > > > > end-users. But when I'm reading papers related to DAML-S the idea I'm > > > > getting for web services is different. They are also useful to > > > > end users as > > > > it shown by DAML-S motivating scenarios: > > > > > > > > Web service discovery > > > > Find me a shipping service that transports goods to Dubai. > > > > > > > > Web service invocation > > > > Buy me 500 lbs. powdered milk from www.acmemoo.com > > > > > > > > Web service selection, composition and interoperation > > > > Arrange food for 500 people for 2 weeks in Dubai. > > > > > > > > Web service execution monitoring > > > > Has the powdered milk been ordered and paid for yet? > > > > > > > > There are also numerous papers that use the term service (and not "web > > > > service") and are talking about UDDI, WSDL and DAML-S. What's the > > > > difference > > > > between "web service" and "service" if both of them work over > > > > Internet? For > > > > example, a search engine (such as google) is a service, but when it is > > > > described in WSDL, published in UDDI and can be invoked using > > > > SOAP becomes a > > > > web service? Ia a asp or an HTML form a service or a web service? > > > > > > > > In summary which are the potential users of web services (web service > > > > providers, developers, end-users)? > > > > > > > > could you please shed some light on this? > > > > regards > > > > Thomi Pilioura > > > > > > > > > > > > > > > > > > > >
Received on Monday, 8 April 2002 15:35:00 UTC