[protomyth]

> It is not a systemd bug to mount efivars read/write. The efitools - efibootmgr et al - require write access to that table. By the spec, this should not brick computers.

Why a filesystem and not a library? Write access doesn't require a filesystem mount.

[pilif]

Part of the UNIX philosophy is to expose as a file what can be exposed as a file. That way, open, read, write and close is all you need.

[retrogradeorbit]

Is it? I understand there are many things in UNIX that are exposed as file like objects, but is it part of the philosophy that this should be done with anything that it _can_ be done on? Can you cite me any sources?

[protomyth]

I've heard and read that for Plan 9, but I too would like the source for UNIX.

[em3rgent0rdr]

Everything a file, even the computer itself: https://media.giphy.com/media/P7PmvHY6kzAqY/giphy.gif

[protomyth]

That's not a real good reason given the damage and since we're talking about systemd.

[ustolemyname]

Given that much of the criticism of systemd is based on, "it's not like Unix," that certainly seems like enough of a reason to me. People would complain even more if systemd required a custom library to install your bootloader than they do when calling rm -rf / on a system without GNU extensions (--preserve-root by default) bricks a motherboard.

[protomyth]

FreeBSD uses a library and I don't think it suffers from criticism that its not UNIX. UNIX isn't Plan 9, and exposing something that can brick your system in the filesystem is just irresponsible. You can blame the vendor, but it should have never gotten as far as the vendor's bad decisions.