On Thu, Feb 21, 2008 at 9:47 PM, ant elder <[EMAIL PROTECTED]> wrote:
> > > On Thu, Feb 21, 2008 at 9:16 PM, Simon Nash <[EMAIL PROTECTED]> wrote: > > > ant elder (JIRA) wrote: > > > [ > > https://issues.apache.org/jira/browse/TUSCANY-2051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12570894#action_12570894] > > > > > > ant elder commented on TUSCANY-2051: > > > ------------------------------------ > > > > > > The suggested fix has already been implemented in the trunk code, but > > as pointed out, its not an optimal solution. > > > > > > The JMS binding spec makes no mention of how to deal with faults or > > exceptions. For RPC style messaging we need a way to indicate that the > > response is either an application or system exception while invoking the > > service. Some messaging systems use a faultTo header on the request message > > similar to the existing replyTo header, but to add that would require a > > Tuscany specific change to the <binding.jms> scdl which doesn't seem > > ideal. > > > > > > How about Tuscany sets a header on the request message to indicate > > faults should be returned, then if thats set and an exception occurs the > > exception gets returned in a JMS Object message along with a header > > indicating if its an application or system exception, then Tuscany can spot > > the exception when receiving the response and throw it to the client. If the > > fault header is not set on the request then the non-optimal fix of just > > returning the exception string is used. > > > > > > Does that sound ok? Or any alternative suggestions? > > > > > Can you raise this as an OASIS issue? Sounds like we need some > > clarification/guidance from the spec gurus here. > > > > Yes ok that sounds like a good idea. > > ...ant > > Response from the SCA Bindings TC chair is that as there's no standard definition of fault messages in JMS so it would be application-specific, the JMS binding spec doesn't define this and the how to handle it is SCA runtime specific. I've added the minimum discussed above so exceptions get sent back in a JMS object message with a property set to indicate its a fault response. Open to any alternative proposals. Would be good to do something better on the differentiation of system fault or user fault as mentioned above. ...ant
