W3C home > Mailing lists > Public > public-ws-chor@w3.org > April 2003

Re: Abstract Bindable Choreography

From: Assaf Arkin <arkin@intalio.com>
Date: Mon, 07 Apr 2003 12:53:07 -0700
Message-ID: <3E91D723.9020403@intalio.com>
To: Jean-Jacques Dubray <jjd@eigner.com>
CC: "'Burdett, David'" <david.burdett@commerceone.com>, "'Ricky Ho '" <riho@cisco.com>, "'WS Choreography (E-mail) '" <public-ws-chor@w3.org>, "'Martin Chapman'" <martin.chapman@oracle.com>

There are three issues here.

1. Does the fact that we use one notation prevents you from using a 
different notation? Using the notation in a non-normative fashion 
ensures that if you prefer to use a different notation you can. We don't 
marry the language to the notation, only use the notation to clarify the 
examples. So the notation should be considered on that merit.

2. Is the notation a good one to select for what you want to achieve? I 
argue that BPMN would be easier to use and read than, for exampe, UML. 
While it looks similar enough to UML, it has more explicit distinction 
between event types (messages, time-outs, faults) and better handling of 
roles, message direction, etc. So it leverages existing experience, but 
provides a notation that is easier to grasp at first glance.

3. Is the model the right one to use?

There is one way to model choreographies which both BPMN and UML take. 
See for example RN PIPs. And I have seen it used before with EDI and 
other communication specifications (e.g. Internet protocols, distributed 
algorithms). This approach is quite common in the B2B world, it depicts 
the separate responsibilities of each partner in the choreography. It 
also maps well to the underlying mobile process model, which can depict 
complex choreographies, allow for better composition and reuse, etc.

I know you disagree with this assesment and prefer a different model 
which shows one activity shared by both partners. Understandable. But if 
we decide that the mobile process model is the better one to use, then 
BPMN seems like the best notation we can utilize. Since we've so far 
discussed using BPMN to model choreographies that fit into this 
category, let's keep the discussion along these lines and not turn this 
into a permathread about which model to use.

arkin


Jean-Jacques Dubray wrote:

>  
>
>>>In fact, I think the last point is very important. The notation is
>>>      
>>>
>self
>  
>
>>>explanatory so it's easy to understand definitions - including those
>>>depicting multi-party choreographies - by just looking at the diagram.
>>>Just look at the example that Stephen gave and see if you can easily
>>>understand what's going on without having to read the BPMN spec.
>>>
>>>      
>>>
>
>Actually, I would content that this is not as true as one would believe,
>yes you can clearly see the message exchange in an otherwise forest of
>internal activities in each respective party.
>
>My main point of contention with WSCI remains and this is why I don't
>like BPMN when applied to choreography as well: the choreography itself
>is not represented, it is merely the association of "n" choreography
>specifications (one for each partner) that one has to use to understand
>(an navigate) the overall behavior of the choreography (this is well
>shown by Steven's examples). This creates an unnecessary level of
>complexity that would hinder the understanding of a large choreography
>(look at what is currently needed to support a few messages between two
>parties). Other notations exist which show one single "neutral"
>choreography without the use of "internal" activity and business logic.
>
>
>JJ-
>
>  
>
Received on Monday, 7 April 2003 15:55:03 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:00:58 UTC