WS-Addressing : dealing with ReplyTo address set to none

[Sergey Beryozkin <sb...@gmail.com>]

Hi

I've attached a patch to

https://issues.apache.org/jira/browse/CXF-2972

to do with returning a fault in case of ReplyTo address set to none.
It appears 202 status is expected but a related JAXWS TCK test actually
expects a fault.

CXF treats a none address as a generic (nearly as an anonymous) address
which is not a problem
in cases when a restriction is applied that only non-anonymous addresses are
expected.

But if an anonymous address is expected then treating none as if was an
anonymous address is problematic.

However, it does seem to me now that indeed returning 202 seems equally
problematic. Why would anyone ever event want to emulate a oneway request by
setting a reply to address to none and then executing a two way request ?
Probably JAXWS TCK is correct here after all on expecting a fault instead.

Let me know if you have any concerns re the proposed fix, I reckon it's ok
to go

cheers, Sergey

Re: WS-Addressing : dealing with ReplyTo address set to none

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

On Friday 03 September 2010 12:13:48 pm Sergey Beryozkin wrote:
> Hi
> 
> I've attached a patch to
> 
> https://issues.apache.org/jira/browse/CXF-2972
> 
> to do with returning a fault in case of ReplyTo address set to none.
> It appears 202 status is expected but a related JAXWS TCK test actually
> expects a fault.
> 
> CXF treats a none address as a generic (nearly as an anonymous) address
> which is not a problem
> in cases when a restriction is applied that only non-anonymous addresses
> are expected.
> 
> But if an anonymous address is expected then treating none as if was an
> anonymous address is problematic.
> 
> However, it does seem to me now that indeed returning 202 seems equally
> problematic. Why would anyone ever event want to emulate a oneway request
> by setting a reply to address to none and then executing a two way request
> ? Probably JAXWS TCK is correct here after all on expecting a fault
> instead.
> 
> Let me know if you have any concerns re the proposed fix, I reckon it's ok
> to go
 
The ONLY use case I can think of is if the client really doesn't care about 
the response at all, fault or not.   But I really think it's a relatively 
bogus use case so your fix is fine by me.


-- 
Daniel Kulp
dkulp@apache.org
http://dankulp.com/blog