Well, from the perspective of just what the JSF spec says, the correct thing is happening - a redirect sets up a new request, and everything you list (except Locale actually, which should be preserved) gets blown away.
But from the perspective of making a users life non-painful, I find the fact that we discard messages to be extremely frustrating - but that we discard request beans to be unavoidable, at least as a default. Communication between pages should not happen via the request scope, which doesn't really work very well. For example, say I use the request scope to pass a parameter from page A to page B. OK, great, page B gets set up fine. Then I click a button which generates a validation error, so page B gets displayed. That's clearly a new request - so where's that parameter that B needed? One thing about navigational redirects has seemed a clear mistake to me - we should not have actually executed the redirect during Invoke Application right in the navigation handler, but should have simply marked the request as such and postponed actual execution to the Render Response phase as a responsibility of the ViewHandler. The current approach makes it very, very painful to specialize the behavior of a redirect. -- Adam On 9/2/05, Martin Marinschek <[EMAIL PROTECTED]> wrote: > and not only the beans, but also the locale, messages and everything > else that is relevant to JSF should be saved over a redirect... > > we should really find someone to suggest this to the expert group. > > well, why don't we do it ourselves? Let's ask Craig, Adam and Manfred > what they think of this - I am including their addresses here... > > regards, > > Martin > > On 9/2/05, Mario Ivankovits <[EMAIL PROTECTED]> wrote: > > CONNER, BRENDAN (SBCSI) wrote: > > >That's a good point about redirect; we don't do that. ;-) > > > > > >Regarding the "browser back" button, I believe that that works OK if you > > >set "client-side" state saving, rather than "server-side." Have you > > >tried that? We have to do "client-side" state saving anyway, because > > >otherwise JSF gets confused when we return from our pop-up windows. > > > > > No, this wont work. > > > > If it comes to the point where a redirect will be issued due to the > > navigation rule there is no chance for JSF to send data to the client. > > The state on the client-side is rendered into the html-page. > > > > Normally the flow is (very, very simplified): > > > > 1) Request > > 2) JSF > > 3) Your action code > > 4) you modify your request beans > > 5) the navigation rule will be applied > > 6) the new page will be rendered > > > > In case of redirect it is: > > > > 1) Request > > 2) JSF > > 3) Your action code > > 4) you modify your request beans > > 5) the navigation rule will be applied > > 6) REDIRECT !!!! -> > > 7) NEW REQUEST (your "request beans" are lost now as we have a new request) > > 8) the new page will be rendered > > > > Now the challenge it to save the beans (maybe marked with saveState) > > during the transition from 6 to 7 > > > > --- > > Mario > > > > > > > -- > > http://www.irian.at > Your JSF powerhouse - > JSF Trainings in English and German >
