The community would have what done in no time? Accelerated indirect support which may or may not require a massive overhaul to the firmware depending on how 2D and overlays are implemented on ATI cards?
What you're ignoring is the fact that FGLRX is based on, in large part, the Windows driver. In order to opensource that driver they would have to weed through a huge ton of code looking for patented materials and trade secrets. If it's not worth AMD's time to fix their driver then it's not worth AMD's time to look through their code. So opensourcing the firmware is out of the question. It'd never make it into the kernel anyway.
What AMD is more likely to do is release specifications for their cards. Given how complicated the cards are, an external interface specification would be functionally useless. It takes a team of hundreds of professional driver programmers over a year to write a driver for a new graphics chip and they have access to both the internal design specifications as well as the hardware engineers. Even a relatively simple GPU, the R200, has had open specifications for years and the opensource r200 driver isn't feature-complete and runs about half as fast as FGLRX did (when it still had r200 support). There are perhaps a handful of people in the world who are both willing to do this work for free and have the capability. I doubt there would be a lot of people willing to do serious work on the firmware if it were opensourced anyway.
No, this situation is very much in the hands of AMD. But that's okay, because we're going to get a driver that's faster and more feature-complete. Linux is also a growing market while Windows is more or less stagnant, so making their Linux offering competitive is going to be a pretty huge draw to OEMs (assuming Dell's experiment is a success).
