Max Spevack wrote:
(6) We figure out a plan for maintenance of the spin, and for hosting
the ISOs. But until there is something actually working, this is not
much of a worry, because it's simply a bunch of text in a .ks file, and
hosting that is trivial.
Basically there *cannot* be a problem in getting a live spin in any
language pass Q&A if the (default) language is all that has changed, or
if OOo(-langpack) is all that has been added. Part of how Release
Engineering and Q&A wanted to compose the live spins had to do
reproducibility, so if the same spin off the same tree is being
reproduced several times with only a different language each, giving a
thumbs up to each of those spins should be a breeze. Quality cannot be a
problem considering the way the compose tools have been set up, and if
quantity is a problem, really, you're back to a quality problem in the
compose tools and reproducibility becomes utopia.
Hosting the ISOs (considering the footprint) is the tricky part I'd like
to get involved with. Or actually, it's the amount of downloading from
the user perspective that I'm concerned with, but it's the same difference.
Presumably, ~90% of the localized Live Spins (or any Live Spin?) is the
same, give or take a dozen. I'm seeking a way to not have to
host/download that ~90% twice or even more often. It's a squashfs in
ISO9660, I'm sure there is a way of finding out what blocks match
between the spins (squashfs) and get something jigdo-like to reduce the
spin's footprints to a bare minimum.
This could potentially include modifications to livecd-tools as to make
the common bits end up on ext3 in a way that makes them go in the same
place on the squashfs image, and finally on the ISO9660 image, but that
is beyond the scope of what I would want to do for a showcase.
Jeroen van Meeuwen