Even with Irridium, you’d still have a cost / infrastructure issue: First, I doubt the Irridium structure can handle the necessary through-put (we’re talking about loads aof data coming from loads of aircraft), but even if it could, it still has to go somewhere, and that somewhere has to be built. The quantities of data would be staggering, and the storage and processing power to recieve, process, store, retrieve, and validate it would be stupendous. Not that it’s necessarily a bad idea, just that it’s not economically feasable at this time.
As for the last minute data-dump, in many cases, it would the disaster happens before a useful dump can be made. Adding built-in dump capability to the “Black Box” would only make them more complex, which isn’t a good idea when you’re building for maximum durability.
A combination of ideas might be doable, however:
Include the ability to manually or automatically start transmitting data in case of an inflight emergency, such as sudden depressurization, setting the transponder to “inflight emergency” or “hijack”, or any of a number of pre-programmed scenarios. It wouldn’t capture all the data, necessarily, nor in all cases, but it would give some clues, such as perhaps what happened to flights that go down at sea without warning, but don’t actually break-up until they hit, and the data recorders aren’t readily recoverable.
The through-put would be far less, and the necessary storage and processing infrastructure could be greatly reduced. Further, this could capture data in cases of near-disasters that don’t result in crash, but take longer than the recording time of the data recorders to reach the ground. When a dangerous event takes place in mid-air, if it takes longer than the recording span of the FDR/CVR to reach the ground, the data is over-written, partially, or completely. A manual in-flight dump could store that data while allowing the recorders to remain fairly simple and inexpensive. The transmitters for this could be phased into service at a fairly leisurly pace, reducing cost burden, and limiting the resultant rise in ticket prices.