Nenz schrieb: > Looks great! Just what I need. > > Finally something to help us abstract the joys of "process scope" in a web > app. I truely believe we have taken the (stateless) protocol itself well > beyond what it was originally intended to do, thus all the hacks we are > having to do to get around a fundamental "flaw" as opposed to perhaps > developing a new protocol itself (I know that's a much bigger conversation) > > Not knowing the inner workings of this other then it uses a filter, are > there any differences between using server state saving / client state > saving... the Sun RI vs MyFaces Impl? > > I guess the true question is ... on a high level.. how is this maintaining > state... I'm assuming it's either swapping session or serializing objects in > hidden form fields and using url parms to further maintain this info? > > Thanks! > > Mario may correct me if I am wrong since I know the mechanisms only from long Discussions I had with him, he basically was starting the thing when I gave him the idea on a tag I once did which was a savestate replacment which used the session directly instead of going the savestate route. And what he did was some genious kind of work indeed.
What it does as I assume is saving the affected objects into the session and using a servlet filter for the garbage collecting the objects. Hence pure client side savestating still will push those objects into the session, as I assume, the rest of the functionality is built around this idea. I dont know how far mario has driven the idea, from an earlier discussion we had on this one, it was clear some generic orm sesssion handler interface for orm mappers also must be integrated to ease the burdens on the orm using people. I am not sure how far mario has driven this by now. The whole thing was thought originally to replace the basic scope mechanisms of seam which are excellent ideawise, but too strongly bound to ejb3, with a tag based approach, which we both liked from savestate, so that you neither have to deal with EJB3 on a server level nor with xml configuration graves like other scope approaches follow.
