Updated Fedora Workstation PRD draft

[Adam Williamson]

On Thu, 2013-11-28 at 09:01 +0000, Richard Hughes wrote:
> On 27 November 2013 21:28, Adam Williamson <awilliam at redhat.com> wrote:
> > to a rough approximation, no-one likes gnome-packagekit
> 
> What a way to be awesome.

Okay, a bit harsh, but c'mon...it's kinda true. This isn't really your
fault, because you're spread so thin you don't have time to make it a
really great app (as you note below). I just see it as a consequence of
Fedora's dev cycle and resource allocations: we just don't consider it
that important to have a great graphical package manager.

> >GNOME Software is a classic Fedora feature: in F20, it's there and it just about manages the most basic functions.
> 
> You know what, I don't know why I bother. Do you know what the
> alternative is? A developer goes into a cave, and comes out two years
> later with a finished design and implementation that cannot be
> changed. Real (not theoretical) users try to use the finished code,
> and find it unsuitable for X, Y, Z reasons. I thought Fedora was *all
> about* release early, release often?
> 
> If I had 5 people working on software and application management full
> time we could have done something better than gnome-packagekit years
> before, and we certainly could have prototyped, designed, implemented
> and tested a working offline update in less than 6 months. But the
> reality was that until very recently we had a toxic environment for
> the package stack with an unstable API that nobody was allowed to
> alter (remember the debacle with Zif?) and a single person (me)
> working about one day a week on the whole middleware and UI stack.
> 
> Anyway, I best get back to writing confusing code.

Sorry, Richard, I was hoping the mail wouldn't read that way to someone
involved in one of the things I highlighted as 'prototype' examples,
because that wasn't my intent. I debated including a sideline on this,
but decided against it because it would make an already-long mail even
longer. Still, obviously I should have done, so here it is. I saw this
as kind of a subtext to my mail, but obviously it wasn't clear enough.

The things I identified as examples of Fedora's prototype-y approach
were _just that_: examples of the prototype-y approach. I didn't intend
to suggest that they were Objectively Bad Things, or that the people
involved in doing those things made the wrong decisions or were
incompetent. I actually agree precisely with what you wrote above,
Richard - "I thought Fedora was *all about* release early, release
often?" - and the point I was trying to convey with my mail is that I
think Fedora is actually _intentionally_ this kind of 'prototype
project', and I don't think that's inherently a bad thing...but it's a
major reason why we don't get more 'mass market' / 'regular' users, and
we have to be aware of that. The goals of 'have an aggressive release
cycle, develop big new features and release them as early as possible,
generally push the envelope' and 'develop a mass user base' are at least
to an extent mutually exclusive.

Again, though, I'm not trying to suggest that I think we don't know what
we're doing, or that the prototype approach is the wrong approach. My
_personal_ opinion is actually that the prototype approach is a good,
useful and interesting thing for Fedora to be doing, and we should think
very hard before moving away from it.

So again, sorry to give the wrong impression, and I wasn't actually
trying to suggest that I think we should have landed GNOME Software
later or worked harder on GPK in the last few years or whatever. That
wasn't the intent of bringing up that example. I was just trying to
highlight that we need to make sure all our thinking is joined-up: it
doesn't necessarily make sense for this group to be having this PRD
discussion which (at least by my reading) seems informed by a desire to
broaden the Fedora user base without considering this question of
Fedora's 'prototype' approach. And if we all agree that the 'prototype'
nature of Fedora is a key part of Fedora's identity that we want to
retain, we need to recognize that that places significant limitations on
our ability to go in some of the directions this whole
three-product-proposal appears to be heading.

As I said, I actually quite enjoy the prototype approach and think it is
a good thing, long term, for the whole ecosystem. If there wasn't a
project like Fedora which pushes the envelope, pushes out new
technologies quickly and provides a nearly-stable environment to whack
them all against each other and see where the sparks fly, I think that
would be bad for the ecosystem: it would slow down development and cause
more friction between developers trying to push new stuff and a
relatively conservative set of distros trying to provide a stable
user-friendly experience. I think by being the envelope-pushing distro,
Fedora provides a valuable service to _everyone_. But being that distro
comes with consequences that we have to keep in mind. 

Releasing Fedora 20 with GNOME Software in its current state is exactly
the right call for Fedora-as-I-see-it, Fedora the 'prototype project'.
It's the classic example of what Fedora does; we put it out there as
early as possible, and that helps us work out the kinks. It would be the
wrong call for Fedora-as-an-attempt-to-build-a-credible-end-user-OS, and
that's why I brought it up in a specific context in my original email,
because I was kinda invoking hypotheticals. But I don't want to suggest
it's a bad project or F20 shouldn't include it or anyone's making bad
calls, because I don't think that.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net