I wish they would fix the UI and make it idiot-proof.

I'm dealing with servers every day and Grub still gives me headache every time I have to deal with it.

I know it's a low-level tool and tackling a hairy problem. But the overall handling is just terrible (chrooting, insufficient error/debug output, insufficient transparency about what the current state is and what grub is doing/going to do, etc.).

When shuffling raid-sets around often enough I'm not even sure from which drive it booted (or didn't boot) in a particular instance, there's way too much trial & error involved for my taste.

After so many years. That's great news.

I remember when I started to use GNU/Linux and they were talking how great will be grub 2.0. It took maybe too much time, but it's a great tool.

A cautionary tale in how not to implement configuration files. Don't make them a Turing-complete programming language.

simple solution: don't even try think of them as configuration files. vim and emacs have the same issue and you don't hear many complaints.

Well that's not so convenient in virt-v2v, where we have to parse grub2 configuration files to work out what they mean, what kernel is actually going to boot by default, what options it'll be booted with etc.

In fact the current approach we take is to ignore grub2 configuration. Very tellingly, all the important distributions ignore the files too. They generate the files from ordinary static configuration files stored elsewhere (in different places for each distro, naturally). So we parse those files instead.

But programming in your programming is so awesome (Yo dawg). Just look at how amazing Sendmail is.

Would be cool to have this signed by a trusted-boot key that ends up in all consumer x86 BIOSes, since it will probably be used by a lot of distributions.