> "that variable is passed to that process only not to the current shell" Exactly, so my question was - why can't you do that to pass a variable into a subshell, which is a separate process? I think @chepner was right - it's a syntactic limitation because bash's compound statement need to start with keyword.

@JaredSmith Oh that's a good link, thanks. But yeah, you're right. I'll settle for a good old as :)

@JaredSmith I'm certainly not looking to fudge safety - I was hoping TS would be able to express a safe way of doing this. So, the way I've ended up thinking about this is that if you take kelly's example, to make it generic, you would need the function type to be a type parameter, but it requires type parameters itself. This is what other languages call a high-kinded type, and TS does not havs those.

@kellys That's the one! Thank you. Edit - right, it's dependent on the actual operation. Well, that's good progress.

@JaredSmith Yeah, I just keep thinking if the correctly-shaped type of the input is known there must be a way to get the corresponding output.

This is absolutely filthy. I love it. Thank you.

For anyone who needs it, the command to remove that taint is kubectl taint nodes --all node-role.kubernetes.io/master-