New comment by heliocat on void-packages repository

https://github.com/void-linux/void-packages/pull/29115#issuecomment-787584309

Comment:
> > First, someone (it might have even been you) had mentioned in IRC that xbps-alternatives was buggy and so I didn't want to put something as critical as startup and shutdown on that.
> 
> Yes it was me. But virtual packages have similar issues too, and are generally avoided when possible. So I understand that committers would not want to take responsibilities for this.
> 
> > The second is that I don't know at what point alternatives become available in the life of a new system
> 
> The symlink should be there as soon as it is installed, so I don't see the issue here... but maybe I'm missing something.
> 
> As a start it would be good to do what ahesford said and move away from `runit-void` things that should be elsewhere. The `ignorepkg` approach is better than nothing...

I've got a half written reply to aheford vis-a-vis moving away from runit-void but the gist of is that I agree but didn't want to break that work off as part of this, at least not as part of the initial PR. As for the symlink presence, the concern is really about the guarantees around when alternatives symlinks are present. The dependency graph guarantees that packages, and by extension binaries, will be present but I don't know if alternatives provide those same guarantees in all circumstances.