[
https://issues.apache.org/jira/browse/ORCHESTRA-29?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12627479#action_12627479
]
Jim commented on ORCHESTRA-29:
------------------------------
Petr is correct, NetAdvantage uses PhaseListener.beforePhase (for Render
Response phase) to handle AJAX requests. The NetAdvantage team regards this as
the safest - and fastest - way to AJAX.
Going forward, we are watching the JSF 2 spec to see how AJAX will be
standardized.
> Make access-scope work with ajax frameworks that "render" during beforePhase
> of a PhaseListener
> -----------------------------------------------------------------------------------------------
>
> Key: ORCHESTRA-29
> URL: https://issues.apache.org/jira/browse/ORCHESTRA-29
> Project: MyFaces Orchestra
> Issue Type: Improvement
> Components: Conversation
> Reporter: Simon Kitching
>
> As reported by Petr Juza, there is a problem with the NetAdvantage ajax
> library and orchestra access-scoped conversations.
> Orchestra's AccessScopePhaseListener runs for the RENDER_RESPONSE phase. In
> beforePhase it clears the "accessed" flag of all conversations, and then in
> afterPhase it discards conversations that are not accessed. In other words,
> it expects rendering to occur between its beforePhase and afterPhase are run.
> However in the NetAdvantage ajax framework (and possibly others) it appears
> that the ajax equivalent of "rendering" is done by a PhaseListener's
> beforePhase code, and that the PhaseListener then calls
> facesContext.responseComplete to block normal rendering. In other words,
> these ajax frameworks effectively move "rendering" (including access to all
> the beans) earlier in time than normal JSF processing.
> In this situation, *if* the Orchestra PhaseListener is registered first, then
> all works fine. But if it gets registered after the ajax phase listener, then
> the accesses to conversation-scoped beans all occur before its beforePhase
> (where the access flags are cleared). So at afterPhase no conversations
> appear to be accessed.
> I don't think this qualifies as an Orchestra *bug*; moving rendering out of
> its normal time is the problem.
> However the obvious solution is for Orchestra to somehow clear access flags
> at the very start of the render phase, before any PhaseListener objects are
> run.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
