How can we add a software infrastructure that does not result in a more rigid system than the physical/mechanical subsystem (relative to a system with humans in control)? Today, automation tends to freeze the RMS by making it hard and time-consuming to adapt (the software). How can we reduce lock-in caused by outdated technologies? Which novel techniques shall we adapt? In this respect, manufacturing - by itself - is lacking critical mass for it to support software technologies on its own. We need to adapt mainstream technologies and find a way to get our own requirements answered within the mainstream developments (not in our own separate developments). E.g. we need to team up and group users of real-time, of very reliable services, ... The user mass exists but users are disorganized and fail to offer an attractive non-fragmented market. Personally, I am looking at Erlang/OTP looking very promising for the soft real time, distributed and high availability aspects as well as hot code updating (www.erlang.org).
The reconfigurable machine tools themselves or at least machine tools that can be easily reconfigured (having standard interfaces) are very important on the top of the system level reconfigurability. The reconfigurable systems are designed for long term so the machine tools (or the elements of the system) should be reliable, easy to maintain and easy to upgrade.
basically it is to come up with cost effective technology, that is flexible enough to machine an ant size object and an elephant size one, with various combinations of operations that are needed to be performed on different objects with a push of a button.