- From: Steve Ross-Talbot <steve@enigmatec.net>
- Date: Thu, 15 Apr 2004 13:12:23 +0100
- To: "Titi Roman" <dumitru.roman@deri.ie>
- Cc: "Assaf Arkin" <arkin@intalio.com>, "WS-Choreography List" <public-ws-chor@w3.org>
Titi, WSCI is dead. It is not used commercially but it was an input and a valid input to the W3C Web Services WG. WS-CDL describes the external observable behaviour of multi-participants from a global model perspective. WS-CDL is based around pi-calculus. I cannot give you proofs but if it is any comfort this is why Robin Milner, Kohei Honda and Nobuko Yoshida joined our WG as invited experts. Such a proof is complex and takes time. I think you will find that few if any commercial standards or products that claim heritage from pi can prove their heritage in any formal way. WSCI cannot and neither can BPML and many other standards and associated products. WS-CDL does not describe the static behaviour only. One only has to think of use cases that require channel passing and so exhibit mobility in the pi-calculus sense, and we have dynamic behaviour. What WS-CDL is not about is on the fly connection of services with not commonly agreed global model. Thus services can join and leave a choreography as and when they need to do so. But they must all conform to the external observable behaviour so defined. This is nothing other than a high level abstract contract that binds the participants based on external observable behaviour. What it does is guarantee conformance, it increases interoperability, it can provide liveness guarantees of so desired all of which are rather useful in the real world and reflect the desires of many vertical standards bodies. We will at some point provide a clearer association to process algebraic formalisms and indeed Nobuko Yoshida has already submitted a type system proposal completely based on some of this work to the WG for inclusion in WS-CDL. WS-CDL has been constructed with all of this in mind and we are in the process of doing the necessary work to meet our objectives, some of which does require some formalism to be shown to be correct and true. Yes WSCI is also a language that describes. Indeed I was wrong. However WSCI, whilst being a good starting point, challenged other parts of the standards world in that it (I think) provided some mechanism for end-point behaviour. This was in direct competition with Abstract BPEL. So WS-CDL does not provide an "intermediate" end-point language. Rather it relies on other work out there and generative techniques to model such behaviour. In short we start where BPEL cannot go. And BPEL stops where we choose to start. Of course in choosing our start point it would be naive to think that BPEL did not influence us in our thinking. It did as did the work on WSCI, the work of Milner, Honda, Yoshida and Kobayshi as well Meredith et al. What we have today is a WS-CDL working draft that meets a set of requirements and goals as described in our requirements document. Further more we have market tested the applicability to what we are doing and our approach an found this to be more than just acceptable. And we look forward to some successes in the not too distant future. I would suggest that you take a good look at the requirements document and this should enable you to see what it is a CDL must be able to do, what it should be able to do and what we might like it to do. Cheers Steve T On 16 Apr 2004, at 09:22, Titi Roman wrote: > > Hi Assaf, > Thanks a lot for clarifying some things...still some questions in-line. > > ----- Original Message ----- > From: Assaf Arkin <arkin@intalio.com> > To: Titi Roman <dumitru.roman@deri.ie> > Cc: Steve Ross-Talbot <steve@enigmatec.net>; WS-Choreography List > <public-ws-chor@w3.org> > Sent: Tuesday, April 13, 2004 11:14 PM > Subject: Re: question: relation between WS-CDL and WSCI > > >>> I am a little bit confused here...In the WSCI 1.0 >>> (http://www.w3.org/TR/wsci/) in section 1.6.2 it says that WSCI "is >>> declarative and cannot, by itself, be executed." (I also haven't >>> heared >>> about amy execution environments for WSCI). I understood that WSCI > possess >>> an operational semantics, did you refer to this when you said that >>> WSCI > was >>> an executable language? >>> >>> >> WSCI was definitely not intended to be executable, but describe the >> observable message exchange between the participants. It did focus on >> a >> level of abstraction that some other proposals would consider > > Which are these other proposals(I am just trying to understand what you > meant by "any language can be viewed as operational compared to a more > abstract language")? > >> "operational", but any language can be viewed as operational compared >> to >> a more abstract language. >> >>> Why do you need a formalism in the choreography context (where can be > seen >>> the usefulness of using such a fomalism for a choreography language)? > Where >>> is it actually needed? >>> >>> >> Formalism allows you to make a lot of proofs about the language >> itself, >> but more interesting, about the specifications you produce using that >> language. In the case of WSCI, formalism becomes more important as a >> way >> to derive conclusions about compositions, and it allows you to >> leverage >> the extensive work done in the field of process calculus. > > Does these mean that WSCI is really dead (as Steve Ross-Talbot pointed > out > in one of his emails)? > > As far as I understood and please correct me if I am wrong, WS-CDL > supports > only "global model"(i.e. the multi-participant view of the overall > message > exchange) - this sounds to me as a static linking between web > services. What > about the dynamic linking between ws from the point of view if > coneversation? What about the view of the overall message exchange as > seen > from one participant? Isn't WS-CDL supposed to support also this view? > > In this context still my previuos questions: > > Where exactly is pi-calculus used in WS-CDL and what scenarios by using > pi-calculus in this context do you predict? > > When you choosd pi-calculus as the underlying fromalizm for WS-CDL, I > suppose you had some reasons for choosing it and not other > formalisms...could you point out shortly these reasons? > > Thanks a lot! > Titi Roman > >
Received on Thursday, 15 April 2004 08:12:43 UTC