On Wed, 2013-11-27 at 19:39 +0000, Matthew Garrett wrote:
On Wed, Nov 27, 2013 at 02:34:16PM -0500, Máirín Duffy wrote:
> See my other post to you. Didn't you experience how, even though there
> weren't nearly as many software titles available for Mac and how much
> more popular PCs were, the Mac people were rabidly fanatical about Macs?
> I mean, it was hard to find a lukewarm / apathetic Mac user. The people
> who used it loved it. They broadened from that base.
To do that we need an audience who's able to evangelise to a wider base.
Are developers going to do that? And if they are, how do we attract them
in the first place?
So I've been reading this thread, and I don't have all the answers, but
one observation to throw in:
I'm not sure there's that much of a difference between targeting
developers and targeting 'casual' users.
In my experience, what a 'developer' wants and what a 'casual user'
wants are often quite similar. Often, they both seem principally to want
a system that's functional, reliable and stable (in both senses of that
overloaded term). I'd say 'developers' and 'casual users' have more
common with each other than either has in common with the 'power user',
who wants to install three OSes with five desktops each onto a complex
partition layout, be able to pick any of those at the drop of a hat, and
change all their configuration settings every Wednesday.
If the elephant in this room is the 'why don't more people use Fedora?'
debate, then I think some of the major reasons for that aren't really
things we're answering in this discussion at all. My impression is not
based on rigorous scientific data, it's based on observation of
list/forum/comment thread/irc/etc etc discussions. But if I can be
allowed to be a bit immodest I'd say I've done quite a _lot_ of that
observation, possibly more than most. I'd summarize the Hive Mind's
Opinion Of Fedora as this:
"Fedora? Hey, I like Fedora. They're good guys. We like Red Hat because
they're the Good F/OSS Company and Fedora is basically like a beta for
Red Hat, right? I wouldn't run it, though. It changes too quickly and
breaks things too often and it's kind of a pain to install proprietary
stuff on, so why wouldn't I just use Mint or Ubuntu?"
People generally don't have a negative impression of Fedora. They think
we're good folks doing good work. But they often don't run Fedora, and
the reasons why really do always seem to boil down to the above: too
unstable - both in terms of changing things fast and without great
documentation, and in terms of our quality bar - and our F/OSS
principles are a barrier for pragmatists.
I don't think we could do a lot about the second point; I'm not in
favour of compromising our principles, I think there has to be a major
distro which doesn't compromise but pushes for proper solutions and I
think it's Fedora's natural role to be that distro. But I think Fedora
could potentially do more about the first point, and I'm not sure the
three product proposal and the discussion this WG is having at the
moment really touches on it.
To point out some practical examples of what I'm talking about:
* We migrated to PulseAudio and systemd very early and without anything
much in terms of hand-holding for users. We didn't publish a systemd
Survival Guide or anything, we just threw it at users and let them
figure it out.
* We decided to use GPT disklabels for BIOS system installs for a whole
release cycle, pushed the change out despite knowing it caused quite a
lot of problems, and then eventually backed it back out again with the
* We replaced the method that's been used for doing Fedora upgrades
since Fedora *first existed* with a completely new and incomplete system
which was completed sometime after the last possible minute (fedup),
with minimal notice to users.
* We've had one or more major change to how we configure how you input
characters into the operating system in _every single release_ from
Fedora 18 through Fedora 20. This mail would be way too long if I went
into the details, but suffice it to say, if you're a Russian or Japanese
Fedora user, you probably had a heck of a rollercoaster ride trying to
type for the last year and a half.
* We never really make a concerted effort to define baseline
functionalities of our OS and consider how they're changing from release
to release. This is something a mature, grown-up, 'proper' OS would do.
We wouldn't ship two releases in a row with system-config-keyboard not
actually working at all, for instance. We would be checking that our OS
actually still conforms to our documentation on how to deploy it and how
to use it, at each release. There are individual superstars doing their
best to keep up with the firehose of changes in these areas, but is it
an organized effort that the distro buys into? Does the entity called
'Fedora' consider it important to make sure that, if you download Fedora
XX and read the Fedora XX manual about how to do things in Fedora XX,
it's actually correct, and we haven't lost or massively changed whole
areas of functionality without fixing the documentation and making sure
we're not dropping important capabilities? I'm not sure we do.
* Our quality bar is pretty damn low for a 'real' operating system. This
is something I think I have a decent feel for as I'm heavily involved in
the release validation process. As a QA guy I try to push for the
quality bar to be as high as possible, but you get a feel for what
'Fedora' as a whole has as its expectations and you can't really push
much higher than that, and what we have is pretty damn low. The
installer in Fedora 18 is not something that a project with high
standards of quality would ever have released.
. It took two releases for us to have a consistent story about how updates are supposed to
happen, after the partial introduction of offline updates in Fedora 18, F18 and F19 were
confusing messes in this area. Our graphical package manager was an acknowledged weak area
of the distribution for 10+ releases: to a rough approximation, no-one likes
gnome-packagekit, but we're only doing something about it in Fedora 20 (and GNOME
Software is a classic Fedora feature: in F20, it's there and it just about manages the
most basic functions. But there are all sorts of features it's missing compared to
what even gnome-packagekit had. You can't tell how big a package is. You can't
configure repositories. There is no longer any graphical configuration of settings like
'should updates be downloaded in the background or not?' This stuff is coming
back...in Fedora 21 or 22. Probably.) And so on, and so on. I can pull out as many
examples as you like.
When people ask me to describe Fedora's niche, I tend to say that we
make a prototype of something that could be a really great operating
system a year later. But we never stop and turn it into a really great
operating system: instead we introduce another dozen shiny things that
aren't quite finished yet and turn out another prototype. We never build
a Toyota Corolla, we're perpetually building motor show prototypes -
something with all sorts of shiny amazing features that isn't really
intended to work satisfactorily in the real world. We're not interested
in doing the last 20% of boring work to turn our super-exciting
prototype into something Joe Normal will drive to work every day: we
just want to keep building more super-exciting prototypes.
This kind of stuff is the reason more people don't use Fedora. If we
slowed down our pace of development and improved our documentation and
our quality standards, we would likely build something that more people
wanted to use...and we wouldn't necessarily need the three-product
proposal or the WGs to achieve that. It's something that we could
theoretically do under that new model, _or_ under our old model. It's
not really a part of the current proposals.
*but*, I'm not saying that's actually what we should do. I quite like
building exciting prototypes. Building Corollas probably ain't as much
fun. Still, there is an obvious corollary; I think it's vitally
important that in any debate which touches on this question, we bear the
above in mind. No matter how we re-arrange our deliverables or talk
about 'target audiences' and the like, as long as we maintain our
current focus on building lots of shiny new things and landing them as
soon as we possibly can and releasing often and not sweating the small
stuff, we are building prototypes, and we're not going to get a mass
user base. So I think it would be a mistake to make decisions as a part
of this process based on the idea that we're trying to make Fedora a
credible operating system for 'regular folks' *or* for 'developers' who
want a stable, reliable operating system more than they want the latest
shiny version of absolutely everything, *without* addressing the more
fundamental stuff I'm talking about above.
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net