[
https://issues.apache.org/jira/browse/TUSCANY-3788?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12932950#action_12932950
]
Simon Laws commented on TUSCANY-3788:
-------------------------------------
Looking at the reference implementation (AsyncJDKInvocationHandler) I note that
it creates the response endpoint on the fly when the forward message is first
sent. This is problematic if the composite/component making the call is
stopped/started while an async response is outstanding as, when the component
comes back up again, the response won't have any where to go to. I.e. the
response service is not created until a forward call is made.
There are complications here though as by rights we don't know how the response
service should be configured (is it async? what binding does it have?) until
the reference is resolved which also may only happen when the first message is
sent.
> Consider what reliable async would look like
> --------------------------------------------
>
> Key: TUSCANY-3788
> URL: https://issues.apache.org/jira/browse/TUSCANY-3788
> Project: Tuscany
> Issue Type: Sub-task
> Affects Versions: Java-SCA-2.0-Beta1
> Environment: All
> Reporter: Simon Laws
> Fix For: Java-SCA-2.0
>
>
> At the moment out Java implementation is no reliable. I..e the async
> correlation information is not persisted between calls. What would happen if
> an implementation type wanted to use a reliable async protocol. I think the
> we would need to give implementation providers access to any correlation
> information that's generated.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
