So, currently we don't have an available server-to-server API to do what you need, but we are working on it.
The workflow you describe is what the FilePublisher does (currently it also upload a file to our servers, but the core model simply passes URLs around), and when we deliver the server-to-server feature you will be able to do this via a REST interface.
Currently in Flex you can create an application that only has a limited subset of the available pods (i.e. only the chat).
That application will only subscribe to the chat collection (plus a couple of management collections) and will only receive (and send) chat messages.
If you were trying to limit bandwidth usage even more you could setup a collection modeled after the chat collection but with permissions set in such a way that the mobile clients can only publish items but not receive them.
I have had a think... about the AFCS concept and how to use it strengths.
The biggest value to me is that at its core the service can manage and 'push' data to clients in real-time - which is something I currently would find impossible to do. In the initial stages of an application design I thus need to differentiate between what data needs to be 'pushed' in real-time and what data can be 'pulled' when required.
In the above question I fell into the trap of believing I had to push image data around network, where as I don't I would simply need to send a 'notification' that some image, somewhere else on the web has uploaded/updated. The image could then be 'pulled' into the application upon being notified - I would then use the notification model for all non essential real-time data, whether it be a another webservice or an email from a mobile phone, or a flash lite application.
Is it possible to send messages from my own server directly to AFCS and tell it to push out a custom notification/message to all the connected clients?
To clarify here is an hypothetical situation:
User in the 'field' sends an email, or uploads an image to an address on my server, the processing scripts (on my server) in turn then tell AFCS some event has happened, which then pushes out the message to the collaborators clients - which then either goes and gets the new data (on my server) or just notifies them to do something.
If I need to send notifications the other way I can do that direct from one of the user clients to my webserver.
I know I keep mentioning mobile phones, but it could be anything.... that triggers the event - from twitter to a gps tracker.
I am approaching the use of AFCS in the right way?
Non flex would be a small flash application for example running on a mobile phone.
It does have to be chat... it could be a person submitting a photograph to the room for collaborators to use.
I am not overly technical in these matters but my perception of what the AFCS is might be wrong... I see your big server a place where resources can be placed for people to jointly interact with. How those resources, whether they be images or text get into that room does not nesscary have to be via a flex based application - is that right?
Actually thinking about it, nothing has to be stored on the Adobe server, it could just be collected from somwhere else. Just like a you tube video is collected from you tube. The Adobe server could be seen more as a messaging relay hub if you decided not to store resources on it. The developer can use the AFCS either way.
I need to have think about this - I am lacking the technical knowledge to really understand how the AFCS works and what it is.....