W3C home > Mailing lists > Public > public-ws-chor@w3.org > July 2006

Fwd: Emailing: CSF for a Process Modelling.doc

From: Steve Ross-Talbot <steve@pi4tech.com>
Date: Tue, 18 Jul 2006 18:47:33 +0100
Message-Id: <6756e3f37b8c54605379f8a863a105f4@pi4tech.com>
To: 'WS-Choreography List' <public-ws-chor@w3.org>

Chaps what sort of UML model will we add to the primer?

Begin forwarded message:

> From: "Frankel, David" <david.frankel@sap.com>
> Date: 18 July 2006 18:21:42 BST
> To: <metamodel@isounifi.com>
> Subject: RE: Emailing: CSF for a Process Modelling.doc
> Reply-To: metamodel@isounifi.com
>
> Matthew,
>  
> Thanks for your comments, which are most enlightening.  In response to 
> your question about where the MOF metamodel for WS-CDL should be done, 
> it is my impression as well that this is not WG4's responsibility.  
> Our responsibility is to point out that such a metamodel is desireable 
> in order to integrate choreographies with MOF-based models and 
> metadata. 
>  
> I have a question about the UML model of WS-CDL that the WS-Chor WG is 
> constructing: Is it purely a class model?  Or does it also have other 
> kinds of models, such as state models, activity models, etc.?  If it's 
> purely a class model, then it should be relatively straightforward to 
> make it a MOF metamodel, since the constructs that make up MOF's 
> metamodeling language are essentially UML class modeling constructs 
> (class, association, constraint, etc.), and one can therefore use 
> generic UML class modeling tools to create MOF metamodels.
>  
> --Dave
>
> From: metamodel-owner@isounifi.com 
> [mailto:metamodel-owner@isounifi.com] On Behalf Of 
> matthew.d.rawlings@jpmchase.com
> Sent: Tuesday, Jul 18, 2006 2:56 AM
> To: metamodel@isounifi.com
> Subject: RE: Emailing: CSF for a Process Modelling.doc
>
>
> Dave - thanks for pointing this out. It is important.
>
> I agree the revised language does not make sense. The original text 
> said that it would be desirable (for the OMG Architecture Group), to 
> extend MOF so it knew about process modeling constructs. The revised 
> text confuses the desired extension of MOF with the requirement for 
> the creation of the metamodel for the process language. These are two 
> distinct things.
>
> So, we can strike out the requirement for extending MOF as this is not 
> a requirement of the process modelling language selection, but instead 
> a goal independent of the language selection. This is something our 
> OMG representatives can convey back to the OMG. The current text can 
> be replaced by David's text marking a MOF metamodel of the process 
> language as desirable. It is moot whether this is a correction or 
> substitution of requirements.
>
> Currently the W3C WS-Chor WG is constructing a UML model of WS-CDL. I 
> believe this is a mistake and it reflects a problem right at the heart 
> of the WS-CDL language, the same problem that I identifed earlier on 
> the mailing list: 
> http://lists.w3.org/Archives/Public/public-ws-chor/2006Jul/0001.html. 
> The model is very welcome as it will make design decisions much more 
> transparent than the current XML Schema. The mistake is treating 
> WS-CDL as if it were at M1, when in fact it is at M2. The problem IMHO 
> at the heart of WS-CDL is that the specification mixes (associates) M1 
> and M2 types (classes and meta-classes). A good example of this was my 
> recent questions about what a WS-CDL Participant is, what a Role is, 
> what a Channel is, and how they relate - Steve reported this prompted 
> lively discussion at the WS-Chor WG. The only way I can demonstrate 
> unambiguously the mix of M1 and M2 is to construct a metamodel/model 
> for WS-CDL, but in the authoring the problems will likely be surfaced 
> and resolved. In other words, if we define this, then the problems in 
> definition will no longer be present, which is not to say they are not 
> there now.
>
> I discussed the mixing of instances and types with Steve Ross-Talbot 
> on Monday 10th. WS-Chor WG had started on a M1 model of WS-CDL, while 
> I was by then convinced WS-CDL was M2. My mistake was to use a UML 
> Profile for WS-CDL, but thanks to Dave Frankel's comments I now 
> realize I had the right layer, but this should instead be a MOF 
> metamodel.
>
> These issues are real and pressing, and not just theoretical concerns. 
> While constructing process models of the OTC and Payments sub-domains 
> I hit many problems deriving from these problems with WS-CDL. For 
> example agreeing whether an instance of Participant is "Tri-Lateral 
> Confirmation Utility" (M1 class) or "DTCC" (M0 object) is critical to 
> modelling process.
>
> The next step is the creation of a MOF metamodel for WS-CDL. My 
> personal view is this is not necessarily a WG4 activity, but the 
> decision is based on WG4's consensus. What is your view?
>
> btw - with reference to our recent discussion about minimizing 
> interleaving between the W3C and OMG stacks of standards, I noted with 
> interest that WS-Chor were building their model in UML and not OWL.
>
>
> Matthew Rawlings
> +44 791 539 7824
>
>
> "Frankel, David" <david.frankel@sap.com>
> Sent by: metamodel-owner@isounifi.com
>
> 17/07/2006 22:41
> Please respond to metamodel
>        
>         To:        <metamodel@isounifi.com>
>         cc:        
>         Subject:        RE: Emailing: CSF for a Process Modelling.doc
>
>
>
> Matthew,
>  
> Thanks for your work on this document. I have an issue regarding CSF 
> 015 (UML Binding), and a concrete suggestion for addressing it:
>  
> The language about MOF does not make sense to me.  What would make 
> sense to me regarding MOF is to say that it's desireable to have a MOF 
> metamodel for the standard's process lanaguage.  A MOF metamodel for a 
> language is expressed in MOF's metamodeling language.  This does not 
> require a MOF metamodel for processes; rather, a metamodel of the 
> process language would *be* a MOF-compliant metamodel for processes. 
>  MOF itself knows nothing about process modeling constructs; it just 
> knows about some very basic constructs for creating metamodels.  
>  
> BTW, the metamodel of the process language would not be *the* MOF 
> metamodel for processes. There is likely to be more than one such 
> metamodel, since there is likely to be more than one process language 
> in the world, and we would eventually like to have a MOF metamodel for 
> each one of them.  Having a MOF metamodel for each language 
> facilitates integration of the languages.  
>  
> My suggestion is to remove the language about MOF from CSF 015 and 
> replace it with a new "desireable" CSF that would read as follows:
>  
> CSF<xx>  MOF Metamodel
> There SHOULD be a MOF metamodel of the standard's process modeling 
> language.
>  
> --Dave
>
> David S. Frankel
> Lead Standards Architect - Model Driven Systems
> NetWeaver Industry Standards
> SAP Labs LLC
> 3410 Hillview Ave, Building E
> Palo Alto, CA 94304
> Phone & Cell +1 530 893-1100
> mailto:David.Frankel@sap.com
> http://www.sap.com
>
>
>  
> --Dave
>
> From: metamodel-owner@isounifi.com 
> [mailto:metamodel-owner@isounifi.com] On Behalf Of 
> matthew.d.rawlings@jpmchase.com
> Sent: Wednesday, Jul 12, 2006 1:54 AM
> To: metamodel@isounifi.com
> Subject: Emailing: CSF for a Process Modelling.doc
>
> <<CSF for a Process Modelling.doc>>
> Please see the updated Critical Success Factors for selecting a 
> process modelling tool.
>
> Matthew Rawlings
> +44 791 539 7824
Received on Tuesday, 18 July 2006 17:48:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 18 December 2010 01:01:45 GMT