The number of iPads bought by those users might be just a drop in the bucket, but that is too shortsighted to look at those users. Those users are the ones who might push development of more advanced apps on the iPad, be it as developers themselves or because they would attract new app development to the iPad. Engineers and scientists might be a small market, but I think it was vital for OS X, that early on Apple laptops started to appear at conferences, slowly growing in share.
The point I objected to comes at the end of this quote:
> I've often wondered whether Apple could do well by doing something similar to Crostini on ChromeOS, to allow these things to actually work as intended, but without impacting security. But that also probably doesn't sell software as effectively, as free software doesn't pay Apple's cut.
The implication is that Apple wouldn’t develop such a feature because it might cut into their profits. I don’t buy that reasoning: Apple wouldn’t lose much if any money by offering that.
Update: and I also don’t believe that Apple would deprioritize it because would cost them some money. Far more likely they’d decide it isn’t sufficiently useful, or would cause other problems, be too confusing, etc.
If those users are so unimportant, why did Apple bother to introduce the Hypervisor framework on macOS? If there's a big, important market for virtualisation on macOS, surely it would be worthwhile to also address that market on iPadOS? Especially since Apple is trying hard to convince all kinds of professional users to adopt the iPad.
I think Apple has a different vision for needs of people who use Macs and iPads for work.
For example the new Sidecar feature (use iPad as a screen) supports the pencil but not touch. Apple believes that touch on macOS interface is not a good experience. I suppose that they feel that text entry and chaining CLI tools on iPad is not one either.
Apple has been trying to pull people away from writing kernel extensions for a while, so I don't see why the introduction of Hypervisor.framework (and with it, another user-space way to do something which previously required working in the kernel) is noteworthy as anything other than this.