Should Linux Have a Binary Kernel Driver Layer?

For those of you that follow LKML at all, you know this is a hotly debated topic. Greg K-H recently posted aboutan OSDL meeting on this topic, along with a follow up. This is a really interesting topic to me. Now, we all know that binary drivers are bad. Here's a post from a Microsoft employee with some details. Like most stories though, there are multiple sides with multiple views here. There are certainly some cases where a company releasing a 100% GPL driver is impossible, due to licensing, 3rd party agreements, whatever. I've even heard (warning, I've not taken the time to verify this) that in some cases it would be illegal to even release proper documentation on API's and interfaces, which would allow kernel maintainers to write a driver. In some other cases, a company may simply decide they don't want to release a Linux driver (be it company culture, company policy, FUD, to prevent tampering, a PHB's whim – whatever). The reality though is that the less driver support there is for Linux, the less mass adoption we'll get. I don't think that is arguable. BUT, the reason I like Linux and one of the reasons it has gotten as far as it has is because it values technical correctness, stability and performance over “adoption”. The question is, and I posted something else about this recently, what price are we willing to pay for mass adoption. It's a tough question and the answer certainly depends on what side of the coin you're on. I for one hope we continue to value technical correctness, speed and stability. For now, I think we will. As for the long run, well – I'm not sure.
On an unrelated note, I find Greg's comments on OSDL “technical management” a bit troubling. OSDL has become quite important to Linux and I'd hate to see them drop the ball. Let's hope this was just an anomaly or a genuine one time miscommunication (which certainly happen).
Linux, Open Source, kernel OSDL
–jeremy