With acknowledgement for this idea to Owen Stephens' recent Tweet. My interpretation of this idea is that 'repositories' are best viewed as a 'type' of data store supporting a variety of services, embedded in various workflows. This fits nicely with Paul Walk's concept of a 'source repository' (see http://tiny.cc/FIHwc) being a simple system with complexity moved to specialised services. I suppose this approach isn't that far removed from the original OAI concepts of data provider and service provider, though the focus there was on access whereas now we are considering a wider context for repositories..
Idea No. 13