On Wed, Mar 16, 2016 at 04:58:55PM -0400, Owen Taylor wrote:
So I think the messaging around Fedora 25 is that this is a new way
to install Fedora - that it is suitable in some cases, and not in other
cases. And we can start with an even narrower use case.
We basically need:
* A central registry for xdg-app repositories
* A way of deciding what goes into that registry; the
3rd-party-software proposal is meant to cover that, but basically
just defers the hard part to the working group.
* Perhaps a "bootstrap" repository where we make existing
Fedora applications available as xdg-apps
We have a big existing community which works to provide applications in
Fedora as RPMs. Rather than making this basically be "thanks for all
that but we don't need you anymore because it's all third-party with
xdg-app", I hope we can bring those contributors in and along.
And, I think we're very likely to need more than just bootstrap, for
First, the existing packaging model includes ongoing ownership and
responsibility, and for all the warts that has, I think it'd be a big
step backwards in Fedora quality and the trust users put in us to drop
And second, while I hope that we can achieve World Domination™ with
this, convincing all application developers and upstreams is not going
to be instant.
Maybe I'm just reading too much into the word "bootstrap". In any case,
I think this needs to be high priority and invested in for this to
succeed, rather than a "perhaps".
> For example, if we could produce xdg-apps through Fedora
> Engineering using the same build pipeline that's being worked on for
> Docker images, the work required would collapse to creating
> documentation for how to do that and any related policies, rather
> than a whole *new* releng change.
That sounds like a good idea to keep things sane on the Fedora side,
but I doubt many users understand the complexity of release engineering
and would pay attention to that.
The release engineering complexity is on the backend (and, yes, very
real). The target here is the Fedora contributor community, and the
goal is that applications could be packaged for xdg-app using dist-git
(ideally, future dist-git with pagure), built with fedpkg, and fed to
automated tests with taskotron, and released for testing feedback with
That's some complexity, but it's also functionality that we'd want to
duplicate somehow or somewhere _anyway_, and I think there's huge value
in using a familiar workflow.
Alignment that make reusing the name makes sense, to me, would be:
* Consistent talking points about what advantages "Atomic" brings -
stability, isolation, etc.
* Good support in the Workstation product for developing applications
to deploy on Atomic Host.
It certainly would be a problem if someone downloaded "Fedora Atomic
Workstation" then found they need to reinstall a non-atomic "Fedora
Workstation" to actually write apps for Atomic...
I definitely agree on these points. I'm not sure it's a complete list,
though. For example, I think "uses Nulecule" would be valuable, too.
(And if Nulecule needs to be extended to cover desktop needs, we should
get the people working on xdg-app involved in extending it.)
Fedora Project Leader