On Fri, 31 Jan 2014 14:37:02 -0800
Adam Williamson <awilliam(a)redhat.com> wrote:
So in my new constructive spirit ;) let me take a crack at some
answers to this:
I think the Spins process as it currently exists has a lot of
problems. We've been saying this for years, long before we even
thought about Fedora.next. You identify some of them above, and there
are others - we've never had coherent messaging about the spins, for
instance. This is especially silly with the desktop spins, where
there are all kinds of mixed messages.
Yeah. I think things got somewhat better in f20 at least due to the
requirement that someone/anyone actually booted the thing and it
worked.
...snip...
The desktop spins are the ones that seem most important to keep. I
think there's a reasonable argument for dropping most or all of the
non-desktop spins, because they're essentially just vehicles for
delivering package groups, when you look at them. Games provides a
bunch of games. Electronics Lab provides a bunch of electronics
tools. There's nothing particularly compelling about shipping these
particular bundles of packages as live images, or as images at all;
we can come up with any number of other mechanisms for letting people
get at them, very trivially. Hell, it's not particularly difficult to
do it right now.
I went down this same path a few years ago, but there are actually use
cases for the non desktop spins that aren't served by just installing
and then installing the packages. For example:
* Using the security spin booted live to examine a compromised install.
You don't want to attach it to a real install thats r/w. Booting off
a read only media means if something messes it up, you can just
reboot.
* You have 30 machines in a lab you can use for your electronics lab or
design class or gamer gathering. You're allowed to reboot them, but
not install anything on them (they have windows on them or something).
You can just walk around before class and boot them all up on live
dvd/cd's. If someone messes up their setup in the class, they just
reboot and get back to the desktop.
Now perhaps these are cases where we just say: hey, make your own for
this, but they are valid use cases not easily handled by dropping those
spins.
The desktop spins, though, do have a reasonable amount of value to
users of those desktops. People do use live media *just as live
media*, and we know there are Fedora users who want to use desktops
other than our default desktop, and Fedora contributors willing to do
the work of maintaining and testing live image deliverables for those
desktops. The desktop spins we have have mostly managed to meet
reasonable quality expectations in recent releases without imposing a
burden on the QA team. I just don't see any major problems to solve
in the area of the existing desktop spins *as small-p products that
are a part of the Fedora project*, though I certainly respect the
releng team's statement that their work scales more or less linearly
with the number of deliverables we decide to make a part of the
Fedora space.
I'm not going to speak for releng, but IMHO... the items that are
somewhat a burden still with spins are:
* Making sure someone tests and signs off at milestones.
(Perhaps this could be somehow automated?)
* The volume of things makes composes take longer.
(perhaps we could stop doing them as part of tc/rc composes, and just
do them after each of those so they don't gate those?)
* websites folks have to look at what was signed off and adjust the
websites for them.
(Perhaps we could make some kind of more self service site for spins?)
Even if we want to keep the alternative desktop live images as a
part
of the Fedora space, though, that affords us quite a bit of
flexibility to change other things about this process.
Agreed.
kevin