W3C home > Mailing lists > Public > public-ws-chor@w3.org > March 2005

Re: Issue 1008 - Fault Handling

From: Charlton Barreto <charlton_b@mac.com>
Date: Tue, 22 Mar 2005 12:09:16 -0800
Message-Id: <d7e1bc8abe61245e90b1de7885332961@mac.com>
To: Gary Brown <gary@enigmatec.net>, WS-Choreography Working Group <public-ws-chor@w3.org>

Hi Gary,

At the same time, BPEL 1.1 does something similar to your proposal in 
order to deal with the WSDL 1.1 scenario (as well as accommodate future 
revs of WSDL which have a better fault-naming model). As such, we 
already have a precedent and thus use cases for the scenario.

As such I am more comfortable with pursuing your proposal.



On 22/03/2005, at 10:50, Charlton Barreto wrote:

> Hi Gary,
> Although we should not let Java APIs fault handling semantics drive 
> any constraints on CDL, and we are primarily bound to WSDL, we do have 
> to understand that the bindings for such cases w.r.t. WSDL 1.1 do 
> prohibit this situation. While WSDL 2.0 is a non-issue in this 
> respect, it doesn't make much sense to me to allow CDL to be able to 
> define a choreography based on WSDL 1.1 which is inconsistent with 
> what the bindings permit, especially when we are otherwise 
> constraining CDL w.r.t. WSDL 1.1 in part based on these bindings.
> If we wish to add 'faultName' to the exchange to identify such faults 
> and support this for WSDL 1.1 exchanges, I feel we need to see a use 
> case for this.
> Cheers,
> -Charlton.
> On 15/03/2005, at 13:12, Gary Brown wrote:
>> Hi
>> My view is that we should not be constraining CDL based on the way 
>> Java APIs deal with faults. We are primarily bound to WSDL, and if 
>> WSDL allows a service to be declared with an operation having 
>> multiple named faults, all of the same type, then we need to be able 
>> to define a choreography that fully supports that operation.
>> I think we simply need to add a 'faultName' attribute to the exchange 
>> which can be used to identify the fault - possibly only relevant if 
>> the associated information type is an exception type? - or to be more 
>> generic we could just re-use the 'name' attribute.
>> Regards
>> Gary
>> ----- Original Message ----- From: "Charlton Barreto" 
>> <charlton_b@mac.com>
>> To: <public-ws-chor@w3.org>
>> Sent: Tuesday, March 15, 2005 7:49 PM
>> Subject: Re: Issue 1008 - Fault Handling
>>> In general, Java Web Services provide two approaches to handling 
>>> faults - deserializing them as Exceptions or processing the message 
>>> directly in a handler. If an operation throws multiple faults with 
>>> the same message type as in WSDL 1.1, where these faults are 
>>> differentiated solely by name, only the latter approach provides any 
>>> mechanism to distinguishing them. Once the fault is deserialized 
>>> into an Exception there is no way to differentiate between faults 
>>> having the same message type.
>>> Generally the lower level language bindings prohibit this situation, 
>>> recognizing it as a degenerate scenario. As such, unless someone can 
>>> come up with a good use case for this, I would propose that CDL 
>>> should strongly recommend against it, rather than provide facilities 
>>> to support it.
Received on Tuesday, 22 March 2005 20:09:33 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:01:07 UTC