On Sat, Oct 6, 2018 at 10:21 AM Ben Cotton <bcotton(a)redhat.com> wrote:
From a format perspective, please allow me to say "wow".
This is
great. My only suggestion would be to put a quick summary of the
desired end state at the beginning in case I'm too lazy to read it all
the way through.
Agreed -- I'll try to add that early in the discovery part so that we
know there's broad agreement on that state. I have a few things in
mind, like modernizing compose in a way that makes it more adaptable
to other users, but we also need to be careful about not being too
pie-in-the-sky.
As for the substance, what I'd like to see explicitly mentioned
is the
benefit to Fedora. You talk about not having to constrain ourselves to
the platform's release schedule, but I feel like there's one more step
there. What will the decoupling of application and platform truly mean
for someone who is using a Fedora deliverable?
Right now, contributors have to line up behind a single process, with
little latitude to own anything. We've made it easy to package lots of
things, but we clutch tightly the reins that allow the community to
own more of that process. It also means we have an extraordinarily
large release surface that's a struggle, even though, for example, our
QA and rel-eng teams are phenomenal. If we cut down what we need to
encompass in a platform release, it potentially yields less material
to track, compose, and test.
But we can't do just that; we also need to give working groups, SIGs,
and other contributors ways to release their stacks in a coordinated
way outside the platform. Modularity is one way to do that, although
there are likely others. From the user perspective, one benefit is
additional options for the platform itself without requiring longer
term maintenance of volunteer-provided packages. Others include less
lag time for a popular new stack, and being able to have the same
stack across multiple platforms (Fedora N-1 and N, for example).
I ask this in part because I would want to pre-address concerns from
the community, press, etc, that Fedora as an OS is dead. It's an
uncharitable interpretation of your proposal, but I can see it coming
up. My conversation with Adam Samalik last night clarified it some for
me, but I feel like there's some assumed knowledge in this that I
don't have.
None of this is to say I'm in any way opposed to the proposal, I just
don't quiiiiite get it, and I'd like to.
Dead? No way. But maybe "steady state" is a better way of saying it.
We haven't tried extraordinarily hard in the past to carve out the
platform like this. In one sense, that's probably because people are
worried about engendering a "Core vs. Extras" worry from the old days
of Fedora. That's not what this proposal is about. Even though we're
talking about a content split, it's not designed to wall off the OS
from community input or contribution. It's in service of letting us
diversify what, how, and when we make -- IOW, making our process serve
our goals, rather than letting the reverse be true.
--
Paul