Last week I attended the Flock 2019 conference in Budapest, like many Fedora community members. There was a good mix of paid and volunteer community members at the event. That was nice to see, because I often worry about the overall aging of the community.
Many people I know in Fedora have been with the project a long time. Over time, people’s lives change. Their jobs, family, or other circumstances move them in different directions. Sometimes this means they have less time for volunteer work, and they might not be active in a community like Fedora. So being able to refresh my view of who’s around and interested in an event like Flock was helpful.
Also, at last year’s Flock in Dresden, after the first night of the conference, something I ate got the better of me — or I might have picked up a norovirus. I was out of commission for most of the remaining time, confined to my room to ride out whatever was ailing my gut. (It wasn’t pretty.) So I was glad this year also to be perfectly well, and able to attend the whole event. That was despite trying this terrible, terrible libation called ArchieMite, provided by my buddy Dennis Gilmore:
<figure class="wp-block-image"><figcaption>One of the things in this picture is mostly harmless, in the words of Douglas Adams. The other is ArchieMite.</figcaption></figure>
I’m not going to exhaustively list all the sessions I attended, although I made it to virtually every hour of the event. Here are some of the highlights.
Plenary sessions at Flock
It’s always good to hear our project leader Matthew Miller talk about Fedora. I was happy to see the steadily rising popularity of Fedora especially as Linux distros overall seem to be flattening. A lot has been going well with both quality and reliability of the operating system, and it’s reflected in the graphs Matthew showed.
Speaking as a (very!) part time contributor, I also would have liked to hear a rallying cry or two, something aspirational and specific. But I also understand that Fedora’s resources out of Red Hat haven’t increased a lot over the last couple of years. And it seems like the evolution of tech that Fedora is incubating is doing well overall. So even without a “moonshot” type goal I’m still satisfied that the project is on the right track for the future.
The talk by Cate Houston from Automattic — makers of WordPress, which powers fine blogs like the one you’re reading, as well as Fedora Magazine — on building stronger teams was good as well. I did feel it lacked a bit on tactical approaches for community projects like Fedora. But there were plenty of lessons to chew on. One of my favorites I RT’d here.
I also thought about her talk in the context of Fedora’s institutional knowledge. We don’t always do a good job of capturing lessons learned. Sometimes we have retrospectives about things we do. As long as we pull out some improvement from each of these, and put it into play, in the long run we should end up with better processes and practices. It’s up to each of us to drive that treadmill of constant improvement.
My Flock appearance: RHEL 8 panel
I spoke on a panel with Brendan Conoboy (development coordinator for RHEL 8), Denise Dumas (Red Hat’s VP of Platform Engineering), and Aleksandra Fedorova (team lead for continuous integration efforts in Fedora and RHEL). We talked about the road to RHEL 8, and what we learned about differences in Fedora and RHEL.
One major point I made was about the huge weight a “secret schedule” puts on the Fedora community and project as a whole. I experienced this first hand as FPL, and I know Matthew Miller has too. It’s not always clear why certain features are important when they land in Fedora. Now that Red Hat has publicly committed to a new major release of RHEL every three years, we know that RHEL 9 will hit the streets in spring or summer 2022. That, plus knowing that Fedora 27/28 were vital to RHEL 8, gives us an idea about critical upcoming Fedora releases for RHEL 9. Think three years ahead, and it’s clear to me we’ll see efforts on technical change accelerating up through Fedora 33/34 (my estimate, not gospel).
I also attended several sessions on Modularity. One of them was Merlin Mathesius’ presentation on tools for building modules. Merlin is on my team at Red Hat and I happened to know he hadn’t done a lot of public speaking. But you wouldn’t have guessed from his talk! It was well organized and logically presented. He gave a nice overview of how maintainers can use the available tools to build modules for community use.
The Modularity group also held a discussion to hear about friction points with modularity. Much of the feedback lined up well with other inputs the group has received. We could solve some with better documentation and awareness. In some cases the tools could benefit from ease of use enhancements. In others, people were unaware of the difficult design decisions or choices that had to be made to produce a workable system. Fortunately there are some fixes on the way for tooling like the replacement for the so-called “Ursa Major” in Fedora. It allows normal packages to build against capabilities provided by modules.
Zuul + composes
There was an excellent presentation on Zuul, the Python-based CI system used by OpenStack. It seems quite mature at this point, incredibly flexible, and possibly of interest to Fedora. However, there are a number of CI systems in use. I’m not sure how many CI system maintainers were around for this talk. But I hope they will take note of the video when it’s issued and see if it’s worth investigating.
Lubomír Sedlá? presented on how to make faster composes. In large part, we stand to gain the most by not producing as much as we have been. One does wonder why we make so many separate expressions of the OS intended for servers, desktop computers, or other footprints.
Lubomír and others are looking at ways that we could make some composing work available to the community. Currently you can only compose with a tight connection to a very large (and understandably sensitive) NFS storage area where we keep package builds. Being able to make that more available to the community would be helpful. I’ve been wondering for a while why we can’t publish our builds somewhere besides a storage location that others can’t access in a read-only mode as freely (like an Amazon EBS or S3 store, or something like that not at Amazon). But Lubomír also talked about work in progress on the composer code, too, and it’s good to see that no one considers the compose “good enough, let’s not try to improve it.”
Another highlight was Dominik Perpeet’s talk on the CI Objective in Fedora. There are a number of people working on bringing the benefits of a CI/CD approach to Fedora. Dominik’s group plays a central part in that. We want to have two orders of magnitude more people run Rawhide daily than right now. A good CI process, along with properly iterating on that process, gets us closer to that goal. Education, awareness, and a shared sense of responsibility across package maintainers will, too! We don’t have to give up daily usability to be able to land new things in Rawhide. But we will need to work better together as a project. I think we’re on the right track but more is yet to come.
I attended a lot more talks, but I’m running out of time to get this blog out and feel like I added to the knowledge pool about Flock proceedings. Many, many other sessions happened at Flock — too many for any one blog to cover! I hope to see the videos soon on the Fedora YouTube channel, and then you too can see some of the fantastic work being done by contributors around the project.
In closing, I want to sincerely and deeply thank Jen Madriaga, Veronica Cooley, Matthew Miller, Brian Exelbierd, Ben Cotton, and all the folks who helped organize and run such a great conference. This was one of the best Flock events in recent memory and it happens through the hard work of organizers. Thanks to all of you, and hope to see you next year at Flock on this side of the pond!