JAX-WS spec interpretation issue....

[Daniel Kulp <dk...@apache.org>]

Could a couple of you check the JAX-WS spec and let me know what your 
interpretation of what is supposed to happen for a specific scenario?   

Scenario:
I'm calling a one way method on my client proxy
I have a Handler registered
That handler throws an exception, specifically a SOAPFaultException.  
(obviously on the outbound handleMessage call as there won't be an 
inbound)  Obviously, nothing makes it to the server.

What happens?   Does the exception make it back to the user or not?


In section 9.3.2.1 of the jaxws spec, it says for an exception during 
processing in a handler and no response is expected (one way):

Normal message processing stops, close is called on each previously 
invoked handler in the chain, the exception is dispatched (see section 
9.1.2.3).


To me, that means the exception is re-thrown back to the client to get 
caught.  That's what CXF is currently doing.  

The RI, however, swollows the exception.   It's gone.  Disappears into 
the ether someplace without any log or anything.   User never sees it 
and thus has no indication that their message never made it to the 
server.

Anyway, this is causing an issue with the 2.1 TCK and I'm trying to 
debate whether it's better to make CXF behave like the RI or start the 
whole process of challenging some tests.     Making it work like the RI 
is easy which would allow those tests to pass, but I'm pretty 
uncomfortable with just swollowing the exceptions.   I'm kind of leaning 
toward challenging the tests, but that's just not a fun process.   :-(   
That said, I already have to challenge a couple others, so what's a few 
more?   :-(


What are peoples thoughts?

-- 
J. Daniel Kulp
Principal Engineer, IONA
dkulp@apache.org
http://www.dankulp.com/blog

RE: JAX-WS spec interpretation issue....

["Paibir, Ajay" <aj...@iona.com>]

Hi Dan,

Based on 
Section 6.4.2 One-way Operations explicitly says a one-way operation
*MUST* throw a WebServiceException if error is detected during message
send. 
And Section 4.3.2 Conformance for Failed Dispatch.invokeOneway which
mentions a WebserviceException to be thrown in case of any protocol
error detected.

To me it appears CXF is handling it correctly.
Also 9.3.2.1 says if a ProtocolException or its subclass is thrown
(SOAPFaultException in the below case) then further handling depends on
the MEP in use. 
Now since the MEP is a oneway then section 6.4.2 becomes applicable.

Regards
Ajay
> -----Original Message-----
> From: Daniel Kulp [mailto:dkulp@apache.org] 
> Sent: 30 January 2008 09:42
> To: cxf-dev@incubator.apache.org
> Subject: JAX-WS spec interpretation issue....
> 
> 
> Could a couple of you check the JAX-WS spec and let me know what your 
> interpretation of what is supposed to happen for a specific 
> scenario?   
> 
> Scenario:
> I'm calling a one way method on my client proxy I have a 
> Handler registered That handler throws an exception, 
> specifically a SOAPFaultException.  
> (obviously on the outbound handleMessage call as there won't be an
> inbound)  Obviously, nothing makes it to the server.
> 
> What happens?   Does the exception make it back to the user or not?
> 
> 
> In section 9.3.2.1 of the jaxws spec, it says for an 
> exception during processing in a handler and no response is 
> expected (one way):
> 
> Normal message processing stops, close is called on each 
> previously invoked handler in the chain, the exception is 
> dispatched (see section 9.1.2.3).
> 
> 
> To me, that means the exception is re-thrown back to the 
> client to get caught.  That's what CXF is currently doing.  
> 
> The RI, however, swollows the exception.   It's gone.  
> Disappears into 
> the ether someplace without any log or anything.   User never sees it 
> and thus has no indication that their message never made it 
> to the server.
> 
> Anyway, this is causing an issue with the 2.1 TCK and I'm 
> trying to debate whether it's better to make CXF behave like 
> the RI or start the 
> whole process of challenging some tests.     Making it work 
> like the RI 
> is easy which would allow those tests to pass, but I'm pretty 
> uncomfortable with just swollowing the exceptions.   I'm kind 
> of leaning 
> toward challenging the tests, but that's just not a fun 
> process.   :-(   
> That said, I already have to challenge a couple others, so 
> what's a few 
> more?   :-(
> 
> 
> What are peoples thoughts?
> 
> -- 
> J. Daniel Kulp
> Principal Engineer, IONA
> dkulp@apache.org
> http://www.dankulp.com/blog
> 

----------------------------
IONA Technologies PLC (registered in Ireland)
Registered Number: 171387
Registered Address: The IONA Building, Shelbourne Road, Dublin 4, Ireland