Re: GRUB 15_ostree

[Chris Murphy <lists colorremedies com>]

On Tue, Jul 21, 2015 at 12:28 PM, Fabian Deutsch <fdeutsch redhat com> wrote:
> On Tue, Jul 21, 2015 at 8:01 PM, Colin Walters <walters verbum org> wrote:
> > On Tue, Jul 21, 2015, at 01:05 PM, Chris Murphy wrote:
> >
> > > Next, after 'atomic host upgrade' there are now two conf files in
> > > /boot/loader/entries, which are sucked into a new grub.cfg. However
> > > the boot order is wrong. The old tree and kernel are the default boot
> > > options. Is this intended or a bug?
> >
> > This is https://bugzilla.redhat.com/show_bug.cgi?id=1226520
> >
> > > If that hypothetic pie in the sky were to materialize, would ostree consider dropping 15_ostree in favor 
> > > of in-GRUB BLS support?
> >
> > Definitely yes.
>
> Yes, that would be great.
>
> Is there an RFE open against grub? We should somehow have one to be
> able to push this forward.

There are two specs proposed [1], and no consensus yet which tells me
there won't be consensus. So I'm not even sure what the RFE should
propose.

Upstream GRUB aren't fans of either BLS version, yet they also won't
say what changes are needed to get them to accept it. And they have
their own idea of drop-in scripts instead of rewriting grub.cfg each
time [2] which of course aren't compatible with extlinux or gummiboot.

Consensus is needed because the spec implies work for bootloaders,
installers, ostree/rpm-ostree, and on RH/Fedora grubby since it
supports a Fedora variant of BootLoaderSpec in order to support
gummiboot. It's a lot of cooperation and work needed to make any
version of BLS work.

I'm trying to convince the (open)SUSE folks that their snapper+Btrfs
design is flawed and to look at OSTree at least for conceptual ideas
for a way forward. If not OSTree, at the least it demonstrates BLS
could be used as glue to support what are ultimately two really
different ways to boot and assemble a system and implement snapshots
and rollbacks.[3] The divergence in boot and assemble, snapshot and
rollback functionality between the two paradigms is very significant.
It's BSD vs Fedora, or Windows vs OS X level of difference.

So I don't know at what point a line is drawn in the sand and you just
go with some variant spec based on whoever is willing to do the work.
I think that's always deceptively easy, but often fails to gather
enough critical mass to make sure it gets carried forward. See
existing blscfg.mod work...


[1]
http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec/
http://www.freedesktop.org/wiki/MatthewGarrett/BootLoaderSpec/

[2]
http://lists.gnu.org/archive/html/grub-devel/2013-06/msg00023.html

[3] There are more than one but this is one of them...
http://lists.opensuse.org/opensuse-factory/2015-07/msg00629.html


-- 
Chris Murphy