"kukeltje" wrote : Shouldn't this behaviour be the default btw? that is a tricky one. for jPDL 4 we could review this choice. but i'm leaning towards keeping as is. if you show a diagram with a fork and an end-state in one of the branches, I would guess that a majority assumes the rest of the process will not be cancelled. introducing this was related to auto-completion. i believe that was triggered by vanderaalst patterns. a process should finish when there is nothing left to do. there is some logic that each time a token ends, it checks if there is some parent token still active. if not, the process instance is ended. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4154930#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...