Linux multi-boot release criterion discussion, redux

[Peter Jones]

On Fri, Nov 07, 2014 at 11:34:16AM -0800, Adam Williamson wrote:
> More criteria time, folks!
> 
> So we managed to get the Windows multi-boot criterion revised and an OS
> X multi-boot criterion added, but we did not yet manage to come to a
> consensus on exactly what should be required for Linux multi-boot.
> 
> I think Chris will re-propose his last idea and we can discuss it a bit
> more, but I'd like to suggest a more restricted criterion we can
> hopefully agree on immediately for the short term:
> 
> ===
> 
> When installing to a system containing an existing installation of
> either the same Fedora release or either of the two previous releases,
> the installer must configure the new installation's bootloader such that
> it can successfully boot the existing installation.
> 
> [Footnote] Typical configurations only: This criterion applies only to
> installations (both existing and new) using default or very common
> storage and bootloader configurations.
> 
> [Footnote] Platforms: This criterion applies to all supported
> configurations described in
> [[Fedora_21_Alpha_Release_Criteria#Release-blocking_images_must_boot|the
> Alpha criteria]], but does not apply to mixed configurations, e.g. it
> does not require that a UEFI native installation of one Fedora release
> be able to configure its bootloader to boot a BIOS native installation
> of another Fedora release.
> 
> ===
> 
> How does that look? I think we had at least a consensus that this much
> was reasonable, and we have two bugs currently that would likely violate
> the proposed criterion:
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=825236
> https://bugzilla.redhat.com/show_bug.cgi?id=964828
> 
> I think it's reasonable to consider these blockers for F21, but we
> should justify it ASAP to give the devs sufficient time to fix them.

I'm all for getting those bugs fixed, but making it an official
criterion is basically adding an anaconda requirement and a new feature
to the distro.  Doing that at the end of a cycle isn't really okay, and
doing it without any mention on anaconda-devel-list or fedora-devel-list
for discussion isn't really that great, either.

Also this criterion as written is going to bring in more than just those
two bugs - for example, right now we don't /really/ support two UEFI
Fedora installations on a single disk without the user making some
fairly strange choices, and when you do that, some things such as
fallback (which admittedly aren't in the criteria AFAIK) don't work.

Fixing that would be a fairly substantial RFE, so either we need to have
language in any potential new criterion to accept the current conditions
(e.g. by requiring two separate disks to put an ESP on each), or we need
to plan on adding support for that before we apply any new criteria.

-- 
        Peter