On 12/23/05, David Geary <[EMAIL PROTECTED]> wrote: > > At JavaPolis, I showed a couple of Ajax demos using Shale remote method > calls. After the demos, I got to thinking... > > The back-end stuff for remoting--AbstractListCompletions and company--is > great. But I don't understand the requirement that remote calls must be > made > to Commons Chain commands. That requirement seems artificial to me, and > induces another layer that developers must understand and configure just > to > make a remote method call. > > As it stands, to invoke a remote method call, you must: > > 1. Implement a Jakarta Commons Chain command. > 2. Add an entry to the remote catalog for your command in chain-config.xml > . > 3. Invoke the appropriate URL using JavaScript. > > It seems to me that steps 1&2 could be eliminated if we got rid of the > Chain > command requirement and instead invoked methods on managed beans using > reflection. Sort of a remote value binding.
After implementing the Dialog feature (which essentially includes calls to action states with a method binding, I'm inclined to agree ... not only would configuration be simpler, but the request can becoma a "real" JSF request. This gives you access to the FacesContext, lets you execute value binding and method binding expressions, and even use the standard JSF ResponseWriter to create the response if you want. All we probably need is a mechanism to map a request URL pattern to a particular method on a particular bean ... and this might be a place where all of the mechanisms (conventional configuration files, annotated service classes, and/or convention over configuration techniques) might have a role. david > > Craig
