Some of the feedback we got on the EF design blog about our early N-Tier plans, highlighted that lots of people just want the whole "tier thing" to just work out of the box.
Fortunately now that we have adopted T4 to generate our entities we have lots of options, in fact in the latest design blog post Jeff Derstadt a colleague of mine talks about our plans around something called Self-Tracking Entities.
Self-Tracking entities essentially solve the "what is the intent" problem I discussed a while back, by remembering what has been done to them.
This is what makes them "self-tracking", and allows us to remove the ambiguity about what the final entity graph state actually means.
It is this tracking that makes it possible to distinguish between a collection that is empty simply because its contents were never known and a collection that is empty because it has been explicitly emptied, which are of course very different things.
These are the sort of things you must know in order to work out what changes another tier really intended.
Anyway read this post for more information.