What is it? What is the specific fragment of algorithmic information that only exists in "live" things and never in "dead" things (or vice versa)? If the term has meaning, such a thing is a describable and we will have found our universal metric.
Also, "successful communication" is a subjective rather than objective measure. Every successful communication can be viewed as an unsuccessful communication in a myriad of other contexts -- this has been brought up before. A "dead" thing very successfully communicates in many other contexts as it decays. What constitutes true decay? DNA is a very stable material, and we can drop ancient DNA into modern cells and get the whole system to start running. If the DNA is running in a molecular machine normally associated with organisms, is it "alive" again? If so, that would seem like successful communication even within your chosen context even though the organism was "dead" for tens of thousands of years. Not that this is any different than single-celled crystalline spores that can automagically turn into living things after millions of years, just a more complex behavior of the system that is our universe.
The problem with the notion of "successful communication" is that it can mean whatever anyone wants it to mean -- there is no objective metric. The system churns with boring regularity and will produce the results regardless, letting other parts of the system (us) argue about "alive" or "dead". At the bottom of it all, there is nothing more but the state space of the system interacting with itself to produce a different conformation. This is what is actually happening. That we've dressed it up with all sorts of neato terms like "alive" or "dead" is not a reflection on the basic character of the system -- the system is what it is and nothing more.
My point for the day: The map is NOT the territory. Many people here have been treating the map as interchangeable with the territory (a very, very common reasoning flaw -- no great sin there). I'm trying to get people to stop arguing over various interpretations of the map and to start walking the territory, but it seems that there is a lot of reluctance to do so for fear of what might be out there in the great wild yonder.
It is not however algorithmic information theory - it's plain old, vanilla, information theory but the mathematics give a measurement which is useful to the purpose at hand.
One of these days perhaps the likes of Adami will come up with something better in algorithmic information theory. But he's not there yet.
I've gotta go now but will try to come back and address some of your other points later this evening.