Fedora update server connections (daily by
IP)(jwf,
12:56:05)
Many caveats, but good for comparing release to
release even if not completely representative(jwf,
12:56:19)
Steady growth since F20, with F25 very high on
charts, and as of this last week, F26 just crossed over the F25
update server connections(jwf,
12:56:50)
40 days for this to happen, record for Fedora
upgrades(jwf,
12:57:01)
Disclaimer, again: We don't do invasive
tracking so these metrics are very approximate and tough to draw
decisive conclusions from this data(jwf,
12:57:37)
Why are we doing crazy things this time around?(jwf, 12:59:26)
Longer release cycles not to make longer
release times only, lot of new type of work going into
releases(jwf,
12:59:45)
Fedora.next: Helped make some of the things
that made Fedora as awesome as it is today; numbers indicate growth
starting from that time or so(jwf,
13:00:58)
32bit systems: Around F20/F21, numbers dipped
and instead of upgrading, just went away; only have observational
data(jwf,
13:01:41)
Last year: Neither growth or decline, just
flat… concerning?(jwf,
13:02:03)
Different type of metric: hotspot check-ins;
every five minutes, make check to make sure they're not behind a
captive portal where you need to log in with a public hotspot; hits
specific URL on Fedora servers; we count the number of times people
hit that(jwf,
13:03:37)
Not form of tracking; we just see people
hitting that URL(jwf,
13:03:47)
Given that laptops might be on everyday;
servers don't do this; not every spin will or people might have
changed it / turned it off; so there are caveats again(jwf,
13:04:56)
StackOverflow survey: ~20% developers using
Linux as primary OS for development(jwf,
13:05:14)
We reached a plateau; we should be growing to
have the impact we want to have; maybe fires necessary to continue
growth instead of flatline we're at now, even if it's a current
record? Want to continue trends, not just hold line(jwf,
13:06:09)
Innovators => Early Adopters < > Early
Majority => Late Majority => Laggards / Long Tail(jwf,
13:07:03)
Fedora lives in the Innovators / Early Adopters
space(jwf,
13:07:22)
Fedora isn't just the bleeding edge (let's
cross that off the list of ways we describe Fedora)(jwf,
13:07:38)
We don't want to be too far ahead because we
want people to actually use Fedora; we want people to use it in
their life, we want to be leading, but not totally bleeding(jwf,
13:07:56)
Can't say: "We invented Fedora" and 14 years
later, doing the same thing(jwf,
13:08:29)
See: Raspbian, CoreOS Container Linux, Solus,
ArchLinux (really bleeding edge with awesome docs), lots and lots of
change going on in the Linux desktop right now(jwf,
13:09:02)
Pushing the innovation curve for things we need
to keep up with(jwf,
13:09:09)
Lot of people use Fedora because our
downstreams too slow: trade-offs about RHEL / CentOS; need newer
versions to do things(jwf,
13:09:38)
Red Hat tries to solve this problem with
software collections with newer things; so, tl;dr: we can't say
we're special just because we package things faster(jwf,
13:10:06)
Using infrastructure that was meant for lower
thousands of packages but we have 17,000+ packages(jwf,
13:11:31)
Our "Fedora machine" isn't always keeping up
with some of the change that we're growing with(jwf,
13:12:01)
Need to find ways to automate what we're doing
with Fedora(jwf,
13:12:08)
Big, awesome changes that will get us to where
we need to go to keep growing Fedora(jwf,
13:12:19)
"Why do we keep blowing things up when we have
infra that does what we need for Rust?" Example of this is that it's
our baseline, Rust upstream not super interested in packages, other
things we need to do to keep up(jwf,
13:12:57)
What's happening at Flock on these things this year?(jwf, 13:13:06)
Every change is tested as a light QA
environment(jwf,
13:13:23)
Generally testing will happen with basic
pass/fail(jwf,
13:13:40)
Continuous integration: Every change is run
through robotic, automatic testing and if it breaks something,
you're told about it right away, and bad changes are rejected(jwf,
13:14:08)
"Sorry, pls fix before pushing to Fedora
thx"(jwf,
13:14:16)
What we have now with Atomic: (1) Build; (2)
Test; (3) Release; (4) Present(jwf,
13:14:49)
WhatWeHaveNow.jpg (see slides for pretty
graphs)(jwf,
13:15:04)
Atomic_Host_CI_CD_diagram.png (see slides for
pretty graphs)(jwf,
13:15:27)
tl;dr: More automation and help us move things
faster and scale up because we're working together with the robots
instead of fighting them(jwf,
13:15:53)
Taking comps groups and making them super-comps
groups(jwf,
13:16:13)
example: Not just install a web server, but
install a web server environment; could install in different
environments and builds in different platforms(jwf,
13:16:43)
Modularity_pipeline.png (see slides for pretty
graphs)(jwf,
13:16:57)
Working on infrastructure to automate creation
and testing of modules(jwf,
13:17:11)
EPEL / Fedors OS update server connections(jwf, 13:17:23)
Earlier, said half a million people use Fedora,
but not limit of our impact(jwf,
13:17:33)
EPEL: Extra Packages for Enterprise Linux;
things we take and package for RHEL / CentOS and make available for
those(jwf,
13:17:49)
Started exceeding popularity of Fedora
connections and is off the charts, literally; somewhere around 1250k
connections (daily????)(jwf,
13:18:16)
10 year commitments for packaging (oof);
modularity helps make packages available for audience and people who
need them without making ten year commitments to packaging and with
Fedora maintenance(jwf,
13:19:02)
Have thing you want moving fast at the speed
you need it(jwf,
13:19:16)
Remember CentOS and RHEL are downstream, which
is also including part of our impact; but we still want to grow
Fedora(jwf,
13:20:01)
Proposal: Ambassadors and Fedora strategy(jwf, 13:20:10)
Proposal for how we spend Fedora money and
efforts in where we show Fedora to the world(jwf,
13:20:40)
Traditionally, go to Linux confs / shows, talk
with people, which is good… but impact is not gigantic(jwf,
13:20:57)
At Linux confs, people know what Fedora is, and
people usually pretty set in their ways(jwf,
13:21:23)
Would like people to focus on new things this
year and help make a difference on the future of Fedora(jwf,
13:21:34)
Council asked this before, but people asked
what we should promote and do; thus, proposal(jwf,
13:21:53)
When we want to spend money on Fedora events,
we want to spend money directly related to Fedora objectives and
mission statement(jwf,
13:22:26)
Objective leads: Identified area that we think
is important for Fedora, lead helps drives forward efforts on the
initiative(jwf,
13:22:52)
Want to empower people working on the
initiatives with the support they need(jwf,
13:23:03)
We should do things that directly support these
objectives(jwf,
13:23:17)
Go to automation confs! Go to where people
using slow-moving distros are with modularity! Go to container /
container orchestration confs! Go to developer conferences and tell
people why they should use Fedora instead of Linux
conferences.(jwf,
13:24:01)
We've been growing for a while, but we're at a
plateau; let's set (good) fires to keep growing and continue the
past trends of the last few releases and break the plateau(jwf,
13:25:35)
Things counted: Editing wiki pages, getting
badges, pushing packages, etc.(jwf,
13:26:22)
Not counted: Going to events, organizing Flock,
writing for Fedora Magazine, etc.(jwf,
13:26:39)
Counted as active: Quarter of the weeks in the
year(jwf,
13:27:06)
Divided into old school / intermediate / new
contributors / all contributors including less active(jwf,
13:27:25)
250 people every week who are showing up
working on Fedora; lots of people!(jwf,
13:27:41)
Intermediate / new users: First active just
this year but at least 13 weeks this year(jwf,
13:27:53)
Good influx of new contributors and flow into
intermediate contributors(jwf,
13:28:03)
We'd like to see this graph going up; probably
can't handle **exponential growth**, but an upward growth would be
nice to see and would be maintainable(jwf,
13:29:15)
Yeah, but that's how it should be. We just want
the fires to be in a good way(jwf,
13:29:47)
This Flock is focused on doing things; let's
find things this Flock that are on fire in a bad way and work on
changing the direction there(jwf,
13:30:07)
Figure out where problems are, get them
solved(jwf,
13:30:13)
Explore automation to make next generation of
Fedora be awesome(jwf,
13:30:22)
Q: "With metadata for updates, if you use
containers, what is counted?"(jwf,
13:31:10)
A: If you use DNF to upgrade system, there's
metadata that describes updates available; gigantic blob of 100s of
MBs of data; modularity has its own data; if have small container
like a web server, your metadata is way bigger than container
itself, which is crazy; but I don't know what we're doing with that,
DNF team might understand(jwf,
13:32:04)
Q, clarification: Hitting mirrors, are
containers hitting mirrors more often and influencing stats?(jwf,
13:32:24)
A: Since coming from single IP address, only
counted once, just like if you `sudo dnf upgrade`, only counted
once; but maybe we should?!?(jwf,
13:33:09)
Q: Do we have a world map of users from IP
data?(jwf,
13:36:10)
A: No, and also a dinosaur (myth) of this data;
with low-cost broadband connections are underrepresented in our
data, so might be lots of Fedora in some regions of world, but a
world map would not represent this well(jwf,
13:36:56)