Lennart Talks Up systemd's SD-Boot + Boot Loader Specification

[stan]

Lennart gets a lot of criticism, most of it warranted because he's a particularly prickly guy. But he seems to hate Grub with a passion, and that makes him ok in my book.
He addressed that in the video :
1) He has contributed to Grub, they didn't accept his patches.

Interesting, a quick check of the grub-devel mailinglist archives going back to 2004 don't show Lennart Poettering ever having sent patches to GRUB. Could you point me to where these patches were submitted and rejected?

As for GRUB being of bad quality, well the fact that it has been able to boot pretty much every computer for the past 24+ years shows that it does what it is supposed to do. If Poettering wants to improve the quality, he should submit patches that do so. It is less work (and less wasteful and disruptive) than starting his own boot loader.

[R41N3R]

As UEFI is almost always used these days, sd-boot works perfectly fine and I really like it's simplicity. Grub is just terrible to use and overcomplicated, I cannot understand why this is still the default in most distributions.

[dreich]

As UEFI is almost always used these days, sd-boot works perfectly fine and I really like it's simplicity.

I find it simpler to just create EFI entries using efibootmgr. Like you said, UEFI is almost always used these days.

Grub is just terrible to use and overcomplicated, I cannot understand why this is still the default in most distributions.

To be fair, systemd is not exactly epitomizing simplicity and reliability but then again it is the default in most distributions.

[Marc Driftmeyer]

Well, you see, I would but:



My system is MBR only

Unify that and get back to me

Buy a new computer and problem solved. I too have MBR not because it's the only option, but because at the time of the purchase Linux EFI was piss poor so I formatted with MBR and used the MBR option on the Motherboard. Moving foward, I'll use UEFI.

[Sniperfox47]

You mean GRUB functionality like native native ATA and USB support to have available as a backup when buggy EFI implementations can't read your hard drive or USB drive?

You mean like how you can download the latest up to date UEFI drivers for those and load them as EFI modules? And download support for additional filesystems as EFI modules? Rather than relying on the stuff baked into GRUB which may work and may not work but needs static configs generated for it even if it does work?

How about in a couple of years when EFI is replaced by the next great firmware, ModernFW. Will you want to throw away all the previous work and start a new bootloader that works for this and only this?

As mentioned already, ModernFW is a UEFI implementation, but even if it wasn't then *yes*. Make a new clean bootloader for the new firmware without all the crud of legacy support. If it's better than EFI enough to get industry support it's probably different enough to justify its own bootloader. rEFInd and other modern bootloader's designed around EFI have many features that GRUB simply does not support, like say dynamic drive detection.

With grub if you plug in a USB drive how do you boot it? Open root, search for the drive, tell GRUB to boot the partition? Do you know what you do with rEFInd? You just plug in the drive and it automatically shows up as a boot option if it has a bootable MBR or EFI boot file. The fact that it's 2019 and you still need to manually search for flash drives in GRUB is a joke.
GRUB is useful for one thing and one thing only, as a low level barebones bootloader for when you need something, anything, to boot your drive. But if your system has EFI then there are a ton of better options than GRUB in 2019.

Also, I thought RedHat, which has now been bought by IBM, is going to try to work better on PPC like RaptorCS motherboards. Will they make a bootloader just for PPC (after all, systemd-boot just works for EFI), so that they can have the pleasure of maintaining yet another architecture-specific bootloader? BTW, GRUB already works on RaptorCS boards.

And what's wrong with rolling another bootloader for those systems that takes advantage of the unique aspects of their firmware...?

[kaprikawn]

Interesting, a quick check of the grub-devel mailinglist archives going back to 2004 don't show Lennart Poettering ever having sent patches to GRUB. Could you point me to where these patches were submitted and rejected?

As for GRUB being of bad quality, well the fact that it has been able to boot pretty much every computer for the past 24+ years shows that it does what it is supposed to do. If Poettering wants to improve the quality, he should submit patches that do so. It is less work (and less wasteful and disruptive) than starting his own boot loader.

He said he sent patches in the video, I have no reason to think that's a lie. I've got better things to do than fact-check that.

Just because Grub can load systems doesn't make it quality. X can display windows, they're replacing it with Wayland because it's massively over-complicated for what it does. Same thing here. systemd-boot is a paradigm shift from Grub. You can't turn the mess that is Grub's config system into something as streamlined as systemd-boot with a few patches.

Yes, Grub does what it's supposed to do, systemd-boot does it much much better. Grub is a bloated mess, systemd-boot abides by KISS principles. Plus they're trying to be good FOSS citizens with sensible standards which Grub are ignoring.

[Slithery]

Why doesn’t Lennart contribute improvements to Grub?

He sent in a patch 4 years ago for GRUB to support the specification. They still haven't replied yet.

[skeevy420]

Buy a new computer and problem solved. I too have MBR not because it's the only option, but because at the time of the purchase Linux EFI was piss poor so I formatted with MBR and used the MBR option on the Motherboard. Moving foward, I'll use UEFI.

Yeah, I meant BIOS in that post. One of those lack of coffee posts and I didn't even catch it until you quoted me which is extra funny based on what I left in post 15

My system has no issues in regards to GPT or MBR disks. Doesn't even need an EFI with GPT partitions...why would it with GRUB? I don't have a UEFI so sd-boot w/o workarounds, am I'm not sure if those exist, is a no-go for me. Sucks because quite a bit of the rest of my system is managed by a something-or-other-ctl command.

I feel it to be a bit disingenuous to go back and fix it after this long and hope y'all get some laughs out of my mess-ups

[F.Ultra]

Why doesn’t Lennart contribute improvements to Grub? It is already so much more featureful and has been tested in the field with actual hardware for 24+ years? Making yet another boot loader is going to increase the maintenance burden and splits/dilutes the developer effort and bug tracking capacity of users in the future, similar to when Lilo was actively maintained.

This is not "making another boot loader", systemd-boot/sd-boot is an older project (gummiboot) that simply was renamed when it was included in the systemd project.

[Delgarde]

Make a new clean bootloader for the new firmware without all the crud of legacy support. If it's better than EFI enough to get industry support it's probably different enough to justify its own bootloader. rEFInd and other modern bootloader's designed around EFI have many features that GRUB simply does not support, like say dynamic drive detection.

Agreed — in theory, a bootloader is a pretty simple thing... GRUB is only complicated because it's trying to support every bit of hardware (BIOS, EFI, etc) from the last few decades, dealing with an accumulated history of edge cases. So yeah, it wouldn't necessarily be a bad thing to have one specialised EFI loader and one specialised next-gen loader, etc, if that makes them easier to maintain than a consolidated codebase that tries to support everything.