[jira] Commented: (CAMEL-2249) Wrong handling of useMessageIDAsCorrelationID

["Marat Bedretdinov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/activemq/browse/CAMEL-2249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=56232#action_56232 ] 

Marat Bedretdinov commented on CAMEL-2249:
------------------------------------------

Hi Christian,

I generally agree that JMS Message correlationID value should supersede the messageID value. Realize though that both configuration assertions and the statement I just made in the previous sentence are all out of band agreements. So one can argue either way.

However, let me give you an example when this will break down. Certain brokers, namely IBM MQ will modify your correlationID value using their proprietary encoding. So what you're going to get back in the reply message will be very different from what you set on the request message on the producer side.

So here the scenario: 

produce  ->       1     -> consume  - > produce ->      2       -> consume
                     ActiveMQ                                               IBM MQ
consume <-      4    <-  produce    <- consume <-      3     <- produce

1. correlationID = C,  messageID = M
2. correlationID = C!,   messageID = M1
3. correlationID = M1,   messageID = M2
4. correlationID = C

C! - is IBM mangled C value

I'd check with Suemas and ask him to review your changes and run against the tests suites as I suspect if your changes were to be applied certain tests that depend on Camel will fail.

Cheers,
Marat

> Wrong handling of useMessageIDAsCorrelationID
> ---------------------------------------------
>
>                 Key: CAMEL-2249
>                 URL: https://issues.apache.org/activemq/browse/CAMEL-2249
>             Project: Apache Camel
>          Issue Type: Bug
>    Affects Versions: 2.0.0
>            Reporter: Christian Schneider
>         Attachments: camel-jms-correlation.patch
>
>
> Camel jms seems to contain two bug in correlation id handling.
> The first shows when you have a sender that has UseMessageIDAsCorrelationID="false" and a server that has UseMessageIDAsCorrelationID="true". If you send a message with correlationId="a" then the response message will contain correlationId="<request message id>". Even if this could be a valid behaviour as you wanted UseMessageIDAsCorrelationID="true" I don´t think it makes sense as the sender will not be able to correlate the message. So for this case I propose to only set the correlation id to the request message id on the server if the correlation id of the request was not set.
> The second bug seems to hide the first bug. Perhaps someone found a quick (and wrong way to make the tests work). It shows when you set UseMessageIDAsCorrelationID="true" on both client and server. If you send a message with correlation id = "a" the client sends it out with this correlation id. The server then sets the correlation id to the request message id (first bug). Then on the client the reply is received. After that the correlation id is set back to "a" on the client. So the tests think all is well. This part of the code should be removed.
> I have marked both problems in the code with FIXME markers in my patch. I can also provide a patch with the solution but first I wanted to only show the problem and provide a failing test. 
> Hope my explanations were not to confused ;-)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.