Fwd: Expressiveness of CDL

Begin forwarded message:

> From: Steve Ross-Talbot <steve@pi4tech.com>
> Date: 20 June 2006 23:21:04 BST
> To: "Paul Bouche (HPI)" <paul.bouche@hpi.uni-potsdam.de>
> Cc: Robin Milner <Robin.Milner@cl.cam.ac.uk>, Nobuko Yoshida 
> <yoshida@doc.ic.ac.uk>, Kohei Honda <kohei@dcs.qmul.ac.uk>, Marco 
> Carbone <carbonem@dcs.qmul.ac.uk>
> Subject: Re: Expressiveness of CDL
>
> Dear Paul,
>
> I scan read your paper with considerable interest. I think your 
> approach is interesting but perhaps not accurate in answering the 
> notion of "based on". I think one needs to be very clear about what 
> this means. In your paper you suggest that to be "based on" a mapping 
> should exist that is essentially bi-directional. This is not 
> necessarily wrong but it is not what we have ever meant. What drove us 
> down the path of formalism was a desire to have some notion of 
> behavioral types that could be used to determine liveness and possibly 
> bi-simulation. You will find these as part of the requirements 
> document that was published a long time ago.
>
> To fulfill our requirements, goals and so on, we do not need the level 
> of equivalence that you suggest is needed in you paper. Session types 
> were added to pi-calculus as part of a higher order calculus. As I am 
> sure you know higher order calculii are "based on" pi but are not 
> necessarily as expressive as pi. In fact the reduction in 
> expressiveness turns out to be important because there are good 
> patterns in pi and equally bad ones.
>
> The global calculus that has been developed by our invited experts is 
> akin but not quite the same as these higher order calculii. After all 
> CDL has no explicit send or receive, it has an atomic concept of 
> interaction in which sends and receives are effectively pre-matched. 
> The end point calculus which has a direct mapping to the global 
> calculus, is very similar to pi-calculus with session types. All of 
> this work will be published soon. In fact it is awaiting the correct 
> W3C rendering prior to publication so it should only be a matter of 
> weeks.
>
> On the issue of names. Names are not directly supported in as much as 
> there is no "new" operator in CDL. Then again you will be hard pressed 
> to find one in any language that makes similar claim (although not 
> impossibly). CDL does have a strong naming concept. When you scratch 
> deeper into what a name may mean with respect to a description of 
> global interaction you will find names are the basis of correlation. 
> They appear as identities on channel type definitions and could be 
> said to be the basis of a mapping to names in pi. Indeed they 
> represent a very important part of CDL because they do "name" an 
> interaction thread through a conversation in CDL and this in turn 
> enables us to understand causality which in turn helps us ask 
> questions of liveness against a specific description. All of this and 
> more will appear in the work by our invited experts.
>
> I am very glad you have shared your paper with us and encourage our 
> invited experts to comment further.
>
> best of luck
>
> Steve Ross-Talbot
>
>
> On 20 Jun 2006, at 22:30, Paul Bouche (HPI) wrote:
>
>> <WS-CDL.equals(Pi4WS)_paper.pdf>
>

Received on Tuesday, 20 June 2006 22:23:42 UTC