dracut http://git.kernel.org/cgit/boot/dracut/dracut.git/tree/README
or stick with initramfs-tools

some problems with people using dracut.
some are using it in Debian, with software raid
one user has issues with btrfs and dracut

options to have a generic or a rescue initrd - no triggers for populating
grub config but update-grub appears to work with multiple initrds

Needs more testing, esp. with LVM or crypto.

https://tracker.debian.org/pkg/dracut

In theory: crypto should work.
make initramfs-tools & dracut co-installable?
	alternative grub configs.
bwh: what we can do is allow each to generate versioned initrd.gz files and
hardlinks or symlinks to get the bootloader to find each one.

Missing support in dracut? Needs testing and bug filing.

Alternative workflow to backup existing initramfs and alternative config.

difference between dracut and initramfs-tools:
	plugins and modules could be harder to work with in dracut?
	dracut currently checks only /usr/lib/ for modules, not in /etc/
		needs a change

initramfs-tools has option to backup but dracut simply overrides.
instead make a backup when the switch happens.

make two packages out of dracut? dracut and dracut-debian.
allow dracut to be co-installable but dracut-debian not co-installable?

A lot of investment in hooks in initramfs-tools, end users especially.
if we want to keep that, what is required to keep initramfs-tools with udev.
Also debian derivatives.

Any historical initramfs tools? nobody in favour.

No initramfs at all? Only for simple cases, not for general purpose devices.

Carrying on as we are with udev: systemd input ...
	once kdbus exists, drop udev's current IPC
	udevadm settle often added to avoid race issues
		- may be the wrong fix but it is a common fix.

dracut does not use busybox, uses libc with shell scripts and some C
programs. Errors fail to a shell prompt. Extension mechanisms allow
for modules which are shell scripts. Include additional binaries - quite
similar to current.

initramfs-tools activity: Debian and Ubuntu are each upstreams.
codesearch for hooks.

2 years ago, dracut maintainer found:
37 packages that ship hooks, 5 config-hooks.d, 28 shipping scripts.

Those maintainers need help to port those scripts, without porting
legacy.

Which needs more work - keeping initramfs-tools or switching to dracut?

dracut has useful steps when umounting the rootfs instead of just
remounting as read-only. The LVM or crypto remains active. dracut can re-exec
into a ram filesystem independent of the original rootfs.

dracut can use systemd for booting without initramfs to allow starting
some storage daemons.

dracut starts journald in the initramfs as well.

some offers of helping with initramfs-tools

only one maintainer for dracut in Debian.
	dracut sprint.

dracut NFS support for NFS4 - fails with klibc.

could drop klibc support from initramfs-tools

dracut has callbacks into udev rather than simple timeouts as in
initramfs-tools.

In error cases where something fails, a timeout is the final step, falling
back to a shell.

bwh does not want to be the main maintainer of initramfs-tools
	no preference for keeping it.
	questionable benefit of the work involved in porting initramfs-tools
	to systemd and kdbus.

dracut and glibc would make an initrd for certain armel and other small
systems.

For people doing TFTP, 32Mb limit which may preclude glibc use (such as
dracut).

Next steps
==========

Need people testing dracut in their use cases
Post gobby notes to -devel.
Sprint?

Needs help with triggers and dracut
Need to port the hooks to dracut or dismiss as irrelevant/historical.

Maintaining initramfs-tools:
 3 significant differences:
 	smaller images than with dracut
 		many can have their modules optimised
 	size needs evaluation
 	
 	stack teardown post shutdown
 	
 	systemd/journald integration.
 		sysvinit users
 		udev without kdbus

stack teardown is fixable in initramfs-tools.

Proceed on assumption that we are not to tie people into systemd in Stretch
Some non-systemd config needs to be retained, maintained and supported.

dracut itself is not ready for full switch yet.
	problems with LVM or software raid command line options can require
	manual changes.

Needs more people to test it, including VM tests.

Split both packages to allow co-installability without overwriting initrd.gz

No Debian Installer support for dracut.