Please look closely. All the commits there are tree-wide cleanups, or automated patches, documentation fixes, warning fixes, and mostly adaptations to internal kernel API changes,...
This is normal "noise" for all code inside the kernel. That's the reason why having the driver mainline is good. But try to find any other activity since the start of git history in 2005.
It doesn't show the driver is in use in the wild. The hardware might as well be extinct.
Maybe I'm being swayed by my ideological views on licensing, and I understand that pragmatism is important, but if it's not in use in the wild then is that really a good argument for keeping? The status quo is a sensible default, but if the issue at hand is magic bits that are not even licensed under MIT, unlike most of the others that are sprinkled throughout the kernel source, is the cost really that high?
Then again, I fully understand that kernel hackers have better things to do than listen to my ideological quibbles and start removing kernel code that might be important to someone running a hospital with Xeroxes that only work with AppleTalk, but by that point what is the chance that the same hospital would at the same time be going around and updating all the devices to 5.10?
