> If you really feel like publishing the event, how bad would it be to create a one shot FacesContext just for delivering the event and then release it immediately after?

Summary: Strictly speaking this is a Mojarra issue, not a spec issue. There are spec implications, but these were addressed during 2.0 Rev a. Details: >>>>> On Wed, 20 Oct 2010 21:10:17 +0200, Martin Marinschek >>>>> <mmarinschek(a)apache.org&gt; said: MM> you never responded to this, but I see on the patch for 1512 that you MM> implemented Blake's suggestion. Is this right? Here is what we did to address this. The spec portion of this issue is very simple. In 2.0 Rev a we modified the last sentence in FacesServlet.service to say: The implementation must make it so FacesContext.release() is called within a finally block as late as possible in the processing for the JSF related portion of this request. The implementation portion is more complicated. As a convenience to app developers, Ryan made an implementation decision in Mojarra 2.0 to agressively dirty session scoped managed beans by providing an HttpServletRequestListener.requestDestroyed() impl that did the dirtying. Blake pointed out that this was in poor taste and also bad for performance. In Mojarra 2.1 I elected to make this behavior only happen when you specify the impl specific context param com.sun.faces.enableAgressiveSessionDirtying. Ed -- | edward.burns(a)oracle.com | office: +1 407 458 0017 | homepage: | http://ridingthecrest.com/ | 14 work days until German Oracle User's Group Conference