[Fedora-livecd-list] Kadischi: roadmap from Board meeting ?

[Jeremy Katz]

On Sun, 2006-04-16 at 19:05 -0700, Jane Dogalt wrote:
> --- Chitlesh GOORAH <chitlesh at fedoraproject.org> wrote:
> > The board meeting http://fedoraproject.org/wiki/Board/Meetings/2006-04-07:
> > 
> > Live CD (Kadischi) -- Lots of people like the idea of this project,
> > but there are still some problems to solve. We should empower people
> ...
> > It is now the time to discuss and share thoughts about how shall we
> > boost Kadischi :)
> 
> 1) get a name that draws interest and is intuitive.  Unless of course the
> motivation is to keep users away until the project is sufficiently mature and
> useful.
> 
> I vote for system-config-liveiso, or livecdgen, or anything whatsoever would be
> better than the cutesy 'kadischi'.  It sounds like something that real men
> don't eat.

Naming is, to be perfectly honest, the least of my concerns.

> 2) unionfs via initramfs (forget read-only root)

Since unionfs isn't in the Fedora kernels, it's not really something we
can count on right now.  And there are significant concerns about it
from some of our kernel developers that are likely to keep it from being
a good option.

> 3) x autoconfig

As was kind of alluded to last week, we actually can be pretty close to
this now.  Chris was working on the rhpxl changes so that the X configs
written out were more generic.  I halfway wonder if the way to go about
this while we wait on X to get the final pieces that would let us kill
xorg.conf is to write basically a wrapper script that gets invoked
instead of /usr/bin/Xorg.  This would probe for the video card and
create an xorg.conf and have a simple/sane config file for overriding
some of the defaults.

> 4) run as user, not root (thus build system cannot be corrupted by liveiso
> generator bugs)

This is going to be harder as you generally can't do everything needed
for package installation as a user.  I'd love to be proven wrong, but
don't have high hopes :(

> 7) selinux, via ext2 image as an option

Trying to support multiple filesystems really is going to be a losing
battle.  Adding xattr support to squashfs shouldn't be that difficult --
look at the patches which added xattr support for tmpfs for the xattr
piece and then just handling putting the metadata appropriately on the
filesystem.

Jeremy