@Suiseiseki
> As a result, Theo is quite happy if the manufacturer slaps most of the driver into some loadable firmware that runs on the hardware itself

That is quite false.
As seen in https://web.archive.org/web/20130424125958/http://kerneltrap.org/node/6550 one of the things is wanting redistributable documentation (while Linux signs NDAs and might even just accept whatever passes code review from vendors themselves) and being aware that proprietary device-software is a security issue.
There is a difference between accepting free-dist device-software and not wanting to go against it, of course FSF/GNU doesn't makes that difference, together with forgetting that devices can be freed.

> usually only when the hardware doesn't use digital handcuffs to prevent replacement of the software

Which is rather rare as digitally signing a software is very expensive, specially for peripherals.

Speaking of which, I *hate* SecureBoot, I don't even get why GNU/FSF even accepts that garbage.
And of course some cryptonerds being all rejoicing at an horribly flawed implementation that doesn't protects them anyway given that you can cheat your way out via grub-shim and run malware.
I'm not even sure SecureBoot can properly let the OS check that the boot-chain wasn't tampered with.
(Meaning CoreBoot + tamper-evident tape/paint is the actual thing they want)

>It's impossible for a copyleft license to be compatible with a different copyleft license (aside from an upgrade mechanism) - otherwise the copyleft would have a hole big enough to drive a truck through.

Ever looked at the Mozilla Public License or the LGPL? Those are copyleft, sadly they allow proprietary software to use the code instead of requiring free software.

> If you are still worried that later versions will be different to freedom somehow there's always

The part I'm worried is contributing to software and using it to the point of heavily depending on it, and then being effectively forbidden to use it because of a relicensing I implicitly accepted.
In a way, OpenSolaris death gave me two licensing pains that likely will never forget:
- Copyright assignment meant Oracle just threw the license out, that's the biggest pain of all
- Linux distros had either no support of garbage support for ZFS, the most resilient filesystem out there among other qualities, mostly because of the CDDL being libre software but GPL incompatible

And sure, relicensing a massive piece of software with asking consent of everyone takes a long time, but I'd rather have that than taking the risk of having a piece of software so complex everyone depends on it, suddenly being unusable because of the decision of one organisation that's borderline a fanclub of RMS (FSF) and one person (relicensing).