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> 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