build should allow local build when on the machine to be built; same for apply?

[Atemu]

When running colmena build --on A,B on machine A in order to sanity check the current config, Colmena fails to build because machine A is a desktop machine that I intend to only deploy locally.

What I would like to happen here is for Colmena to build config B on machine B remotely but build config A locally on machine A.

The same applies to apply actually; when I'm on the machine being applied itself, I'd expect Colmena to do apply-local on this particular machine and apply remotely on all the others I've specified.
I never understood why the distinction between apply and apply-local has to even exist though: apply-local can only ever be done when on the machine itself, so it never makes sense to use when hostname != target. Reversely, it never makes sense to use remote deployment when local to the machine you intend to deploy.
Am I missing something? What is the rationale behind this design decision?

If there is concern about edge-cases, this could perhaps be alleviated via a flag instead; apply --force-local to allow/force local deployment even if the hostname does not match and apply --force-remote to do the opposite.

A workaround for both of these issues would be to allow local "remote" builds via SSH and always use apply but I don't really want my local SSH key to be able to log into my machine to mitigate privilege escalations.

[fpringle]

Seconded! This restriction doesn't make sense to me either; if I want to build my entire fleet I have to do colmena apply switch --on '@not-local' && colmena apply-local switch, which isn't very ergonomic.