The 2.6.28-libre kernel is out

Yeah, right. This theory might have held some water back when it was first brought up, when firmwares were actual independent works merely packaged together. These days, a number of firmwares are actually developed alongside the corresponding drivers, so updates in one require updates in the other. It's very hard to show that the evolution is not maintained with clean-room techniques, but good luck defending 'mere aggregation' for those.

And then, this doesn't even cover the most common cases. Quite often people will take say firmwares in binary-only-disguised-as-source form under a permissive GPL-compatible license and assume that it is lawful to redistribute that under the GPL.

At other times, people will take bits extracted from non-Free firmwares, or from I/O dumps taken while running them, and assume it's lawful to distribute those dumps under the GPL without getting permission from the copyright holders.

> If you have evidence of distribution without the rightholders consent, as you imply further down, I suggest you take that up with the kernel maintainers, or possibly the owners.

I wouldn't betray the people who confided that to me, or who stopped responding once it became clear to them that a screwup was underway. Freedom has nothing to gain by the exposure of these errors, quite the opposite. And then, given upstream resistence and heat towards faulty generalizations towards this topic, why bother?

That the exclusion power of copyright cannot be used to stop certain uses doesn't mean the code doesn't qualify for copyright. I very much doubt you'd be able to get the same kind of exception the courts allowed, for Lexmark clone cartridge manufacturers to duplicate creative works added for the specific purpose of using copyright to prevent interoperability, to the non-Free programs that run on peripherals' CPUs, and that could be programmed differently and would function just fine, if only we knew how to make such arbitrary programs for them. But hey, IANAL, give it a try and maybe we'll all get lucky.

Not that getting permission to distribute those bits (which we already have, at least for some of them) would do much towards enabling them to be lawfully distributed as part of a larger GPL work. The interoperability claims would hardly get you permission to modify the work as extensively as required by the GPL. And, if you distribute it nevertheless, *any* of the copyright holders might notify you that your license is terminated, and get an injunction to stop you from modifying, distributing and, in some countries, even running the program, regardless of who induced you to the error. Then what? Why would you accept that kind of risk, when there are much safer alternatives?

I can see that people going from A to B might prefer to go through C rather than straight to B.

But that's not a reason for people who are at D, closer to B, to be dragged back to C, just because others haven't even reached C yet. There are other ways to draw the path that doesn't force this detour.

Furthermore, I dispute that it may be necessary to go through this particular C. If some people aren't ready to reach B, it might be better for them to wait a bit, or try to find another path to go through to reach B, than to risk falling in the trap set up in C and getting stuck there.