WiFi is literally the only blob Linux will have to deal with as far as we can tell, and I'll make sure our installer pulls it from the system firmware partition (which will always exist on a functional Mac) so nobody will have to worry about it. All the other blobs are loaded by system firmware before Linux even gets to run, so they might as well be invisible to us (UEFI on PCs does much the same thing).

No DKMS, ever. All our drivers are headed to mainline. I'm not even going to bother with DKMS most likely; people who want to use bleeding edge drivers before they're merged can use our kernel trees (and we'll have builds available for those folks).

I wouldn't call it optimism per se, not in absolute terms at least. That will certainly continue to be a problem. But Apple messing up their own software either by accident or on purpose is a nice boon. And a single piece of consumer hardware being so good that that their a patreon for it is also interesting and good.

For the reasons you say, I personally would rather buy e.g. some pine64 thing just might benefit from this porting work. (Probably more in the userland than kernel itself, tbh.)

The one good path I see for hardware is that in the system on chip era, open designs can gain a foothold by saving the companies money, just as no one could compete with free linux for server software 2 decades prior. There was no hope getting a foothold with 1 design = 1 chip, but a world where people can mooch some free IP for their dohicky might do the trick.

Progress on the Apple GPU is already at the "can draw cubes" stage:

https://rosenzweig.io/blog/asahi-gpu-part-3.html

All the Asahi Linux support for Apple M1 is going into Linux mainline, so the dkms modules will not be necessary for the parts they work on.