►
From YouTube: Kubernetes Community Meeting 20160422
Description
We have PUBLIC and RECORDED weekly video meetings every Thursday at 10am US Pacific Time.
https://docs.google.com/document/d/1VQDIAB0OqiSjIHI8AWMvSdceWhnz56jNpZrLs6o7NJY
Demo Kargo by kubespray, Third Party Resources, Kubernetes Feature Workflow.
A
There
we
go
you're,
muted,
now,
hello,
hey
Cooper,
Nettie's
community
meeting
today
is
April
19th
20th
21st
holy
cow
today
is
April
21st
and
we
are
going
to
go
through
a
couple
of
things.
Today
we
will
have
a
demo
from
Maine
which
is
I,
we're
doing
the
audio
checks.
Just
a
moment
ago
about
cargo
x,
cube
spray,
a
cluster
deployment
management
tool.
Then
we're
going
to
hear
from
Brendan
Byrnes
talking
about
third-party
objects
and
a
coruña
DS
workflow
proposal.
This
is
sort
of
a
trimmed
down
version
of
the
broader
cabrinha
DS
efforts
proposal.
A
While
we
continue
to
talk
about
that
and
figure
out
what
the
solution
is
there
we
end
up,
maybe
moving
forward
with
this
feature
workload
proposal
and
then
there
and
then
Clayton,
came
in
and
said
he
wanted
to
talk
about
some
protobuf
work.
So
we've
got
a
pretty
big
agenda
today,
as
well
as
any
sick
updates.
If
we
have
time
and
notices.
So
let
me
put
my
notices
up
top
the
week
of
May
5th
or
our
weekly
meeting
the
first
week
of
May
will
be
at
the
asia
friendly
time
of
5
p.m.
pacific
time.
A
It's
unfriend
life,
unfriendly
for
a
lot
of
others,
but
our
community
members
from
asia
have
been
dialing
in
at
crazy
odd
hours.
So
we
want
to
say
thank
you
to
them
and
do
something:
that's
a
little
bit
gracious
to
them.
So
let's
get
started
and
I
will
find
fame
and
unmute
him,
and
then
we
debug,
ok.
B
So
so
I'm
smiley
member,
a
member
of
a
team
coops
play
and
the
studio
you
can
have
a
look
through
our
projects
on
github.
We
have
two
main
projects
we
have
JPL
which
allows
to
deploy
applications
on
top
of
communities.
Actually
Antoine
will
do
a
demo
next
week
and
we
have
cargo,
that's
what
I'm
going
to
show.
B
So
what
is
cargo
Chicago
is
an
open
source
project
who
started
about
one
year
ago
started
it
from
the
production
news
came.
We
had
the
video
on
demand,
the
application
on
the
hill,
and
we
needed
to
vibrate
this
application
into
our
own
data
centers.
So
it
can
be
used
to
deploy
your
criminal
discuss.
Iran
parameter
all
at
the
US
DC
and
recently
we
have
contributions
on
OpenStack.
B
So
it
comes
with
a
simple
command-line
tool
which
is
really
simple
to
you.
You
just
you
can
run
a
production-ready
cluster
with
it,
but
you
can
also
run
it
a
false
development,
but
once
again
la
la
so
as
I
said
it's
very
simple.
Yesterday
I
was
picking
up
to
the
headaches.
It's
the
most
important
development
conference
are
in
France
and
we
were
able
to
boost
our
to
run
60
clusters
of
three
notes
on
each
Japanese
have
been
able
to
to
run.
This
is
chris
erroneous,
laptop
good
god.
B
That's
important.
As
I
said
it's
simple,
yes,
but
so
the
default
options
are
really
fine.
You
can
do
you
different
options,
but
you
have
a
lot
of
options
to
fits
your
needs.
So,
for
instance,
you
can
shoot
the
network
again.
You
want
to
run
inside
your
cursor,
you
can
move
in
the
future.
You
will
be
able
to
choose
the
container
gene,
the
care
of
the
key.
B
If
you
will
be
able
to
choose
the
configuration
that
there's
a
TV
picture
so
and
we
did
a
special
focus
on
the
s,
because
everything
moves
fast,
cabana
this
world
and
the
way
to
test
each
change
and
we
don't
want.
We
didn't
want
to
break
anything.
So
we
want
to
be
to
date
for
each
component.
We
have
on
the
playbook,
but
we
still
want
to
do
tests
to
trigger
test
each
time
we
do
change.
So
I
can
show
you
we
have
a
lot
of
tests.
B
Each
combination
is
a
kiss
tested,
for
instance,
planet
with
DB
and
reset
us
or
calico,
with
a
way
that
with
the
coalesce,
so
we
each
combination
is
tested
and
it
wins
a
cluster
of
three
nodes.
We
tested
everything
works,
fine,
so
it's
render
cluster
and
I.
At
the
end,
it
checks
that
each
pod
Control
Commission
each
other,
so
the
best
is
pretty
complete.
So
I'm
going
to
start
the
demo.
B
B
B
So
a
profit,
so
it's
going
to
create
the
end
and
you
can
do
the
same
come
online
and
simple
nineties.
A
look
is
very
well
for
anybody
yet,
and
you
have
some
options
in
added
you
to
have
a
look
to
the
options
and
for
for
the
permit
themselves.
If
you
have
already
have
your
servers,
you
can
use
the
comment
preparer,
which
should
only
generate
the
inventory.
You
will
use
it
for
the
deployment
and
you
can
add
another
to
a
running
cluster
like
that.
B
So
with
the
option
add
you
can
let
your
current
inventory
and
then
you
will
have
to
deploy
again
with
the
same
options
of
the
previous,
sir,
then
so
all
become
online.
You
can
find
on
the
other
direction.
Sorry
you
can
find
in
the
config
file.
You
can
override
it
with
the
command
line.
It's
a
really
useful
when
you
just
have
to
change
the
type
of
the
server
or
the
image
you
want
to
use,
so
it
will
run
to
create
tpms.
I
will
show
you.
The
revisions
are
creative,
okay,.
B
B
B
So
I've
chosen
a
topic
in
calico
and
independent
is
running,
so
it
downloads
the
winery's
configure
the
attd
cursor.
So
you
should
have
already
you
should
have
pre-printed.
At
least
it
creates
the
certificate
and
synchronize
the
certificates
between
nodes
masala
all
the
packages,
for
instance,
dachau
of
them
see
Patek
achieve
the
required
packages
and
it
in
to
configure
the
Masters,
so
it
installs
the
API
servers
on
two
nods,
as
well
as
the
dollar
and
schedule
and
with
the
master
election
process,
stopped
everything
and,
at
the
end
of
these
demo,
I
check
that
everything
is
working.
B
B
So
let
the
city
the
projects
we
are
working
closely
with
the
network
ground
out
from
the
beginning,
we
started
to
talk
with
a
calico
with,
and
we
soon
you
able
to
integrate
to
an
attitude
and
the
reopen
dish.
We
talk
often
with
them,
and
we
have.
We
had
some
much
similar
configuration
from
other
companies
like
racking
total
net
for
OpenStack,
and
we
need
something
to
the
contribution.
So
we
need
you,
we
need
some
tests.
We
need
you
to
test
you
to
connect
to
open
issues
and
eventually
to
add
the
confirmation.
B
B
So
what
to
add
nodes
is
already
possible
to
remove
em,
to
break
the
humanities,
to
a
specific
reaction
you
are
working
on.
Integrating
there
are
some
new
network.
Plugins
like
women
are
right
and
we
want
to
give
the
option
for
users
to
choose
if
they
want
to
run
all
the
components
inside
parts,
because
we
did
the
choice
for
persian
reasons.
B
This
tries
to
run
some
components
inside
the
nuts
as
a
service
of
the
system
cells,
and
we
are
looking
for
rather
looking
to
at
the
end
of
the
year,
I
think
so
you
can
find
the
link
you
can
check
out
our
repository
that
steep.
As
I
said,
you
don't
hesitate
to
ping
me
on
slack
and
I
was
like
and
I'm
participate
to
the
secret
ops
2,
so
you
can
join
it.
We
are
working
on
this
business
kind
of
the
issues
so
uneasy
day.
Thank
you
all.
B
A
B
C
It
will
be
added
to
the
crystal
yes
certificates
kind
of
quite
playful
to
link
them.
So
this
is
major
parts.
Every
single
we
communicate
with
the
gamemaster
through
gns
https
on
me,
so
we
synchronize
each
better
right.
It's
not
too.
We
add
each
not
in
the
certificate
to
be
able
to
contact
the
abuser.
A
C
Won't
you
so
this
is
part
of
the
life
cycle.
Engine
run
map,
so
so
now
we
don't
have
the
command,
but
it's
something
we
want
definitely
want
to
add
to
make
people
just
say:
okay,
I
want
to
read
from
1
30
to
2,
1,
cod3
and
the
script
we
just
run
in
the
best
practice
to
get
just
a
rating
update,
/
kristina,
actually.
B
A
Soon,
okay,
thank
you
both
for
the
demo.
If
there
are
any
other
questions,
you
had
some
example
or
you
had
some
contact
information
on
there.
Can
you
add
that
to
the
document
where
we
put
you
in
the
agenda
so
that
you
can
so
people
can
reach
out
to
you
and
then
maybe
a
couple
of
notes
of
key
points,
because
we
didn't
have
a
note-taker
during
your
demo?
Sadly,
so
so,
if
you've
got
key
points
you
want
to
put
in
there,
that
would
be
great
okay,.
A
Thank
you
so
now
I!
So
now
I
am
going
to
go
ahead
and
ask
if
we
can
get
a
note
taker
for
the
rest
of
this
before
I
kick
off.
The
third
party
objects
update
from
Brendan
Byrnes.
So
Brendan
are
you
about
come
here
awesome
and
semana.
If
its
main,
if
you
could
stop
sharing
your
your
other
screen,
that
would
also
be
helpful
and
then
I
think
Brendan
can
share.
If
he's
going
to
or
arrow.
A
A
D
So
you
know
your
kernel
that
ships
with
the
operating
system
is
monolithic
and
it's
hard
to
add
things
and
remove
things,
but
kernel
modules
enable
you
to
dynamically
add
support
for
various
features
and
I
think
that
going
forward
I
think
this
is
how
we
should
think
about
extending
the
Cooper
Nettie's
API
rather
than
adding.
You
know,
lots
and
lots
of
API
objects
to
the
mainline
thinking
about
having
modules
that
different
users
can
and
even
take
from
third
parties
that
different
users
can
insert
into
their
cluster
for
a
time
being.
D
There's
nothing
really
running
there
right
now,
and
so
what
I'm
going
to
do
is
I'm
going
to
do.
Q
control
run
this
demo
module
all
right,
I
guess
before
that,
I
should
do
that,
like
there's
nothing
up
my
sleeves
part,
which
is
also
a
cube,
control,
get
demos
so
I'm
going
to
create
this
third-party
object
called
a
demo
and
it's
going
to
say:
oh,
it
doesn't
have
the
resource
type
demos,
I,
don't
know
about
that
resource
type.
So
now
I'm
going
to
run
and
insert
this
module.
D
D
So
one
of
the
things
that's
important
here
is
that
all
these
objects
are
prefixed
by
a
domain
name,
so
that
we
can
have
different
namespaces
for
objects
so
that
if
I
choose
the
name
demo
and
somebody
else
chooses
the
name
demo
they're
not
going
to
conflict
with
each
other,
it
says
watching
for
new
objects
and
then
briefly,
you'll
see
I
should
have
said
dash
F,
so
I
follow
it
briefly.
You'll
see
here
it
says
the
server
doesn't
have
the
resource
type
demos
doesn't
have
the
resource
site.
D
Demos
doesn't
so
that
error
message
is
coming
back
because
the
you
create
a
third-party
object,
but
it's
a
little
bit
asynchronous
before
the
actual
API
paths
are
our
PAB
din.
So
for
a
brief
moment,
it's
not
there.
But
now,
if
we
go
back
over
here
and
we
say,
cube
control
get
demos,
we
get
an
empty
list.
So
it
didn't
give
me
that
error.
There's
no
demos
created
so
just
give
me
the
empty
list
and
then
do
device.
80,
JSON
you'll
see
that
we
get
back
a
list,
and
this
is
actually
a
you
know.
D
Cooper
Nettie's
style
object.
It
has
the
API
version.
It
has
a
metadata
that
you
would
expect
okay.
So
now,
let's
take
a
look
at
one
of
my
demo
objects
actually
before
I.
Do
that,
let
me
just
create
it
just
really
exciting
to
say:
cube,
control,
create
f,
demo
JSON,
and
now
this
object
is
created
and
if
you're
looking
over
here,
my
that
module
that
I
registered
now
said
object
added
right.
D
So
at
this
point,
if
it
were
a
more
complete
module
and
not
just
a
demo,
it
now
knows
about
this
object
and
it
can
start
managing
it
right.
So
you
can
imagine
that
the
replication
controller
could
be
implemented
by
one
of
these
modules.
We're
lacking
some
features
and
I'll
talk
about
that
at
the
very
end.
D
To
make
us
actually
want
to
do
that
right
now,
but
you
can
easily
imagine
creating
new
objects
and
then
managing
those
objects
and
by
managing
them
you
know
creating
other
objects
so
creating
something
that
then
creates
pods
or
creating
something
that
then
creates
services,
that's
sort
of
them.
So
let's
take
a
look
at
that
demo,
JSON
so
again,
you'll
see
here.
This
is
the
standard.
A
kubernetes
object
metadata,
so
we
have
kind.
D
D
Benetti's
API
itself
uses,
so
users
get
to
do
those
themselves
and
I
can
actually
then
so,
and
then
we
can
say
if
I
try
in
so
let's
say,
cube
control,
get
demos,
my
object,
20
JSON,
we
get
back
the
JSON,
and
what
you
can
see
here
is
that
there's
this
resource
version
2002,
and
so
now,
if
I
go
in
and
vo
demo,
JSON
and
I
actually
add
the
resource
version
field.
Here.
D
D
D
So
now
sorry
that
now
you'll
see
this
error
and
it
says
error
from
server
error
when
replacing
demo
JSON
the
server
reported
a
conflict
so
you're.
You
know
these
objects,
despite
being
defined
by
end
users,
have
all
of
the
semantics
of
a
standard
crew.
Benetti's
object,
you
can
do
compare
and
swap
you
can
do
watch
you
can
do
all
of
the
stuff
that
you
would
expect
to
be
able
to
do
and
they
integrate.
Naturally,
as
you
can
see,
with
cube
control
commands,
and
so
that's
about
it.
D
I
don't
really
have
anything
else
to
talk
about
there,
but
I
think
as
people
are
thinking
about
things
that
they
may
want
to
add
to
their
cluster
I
think.
Hopefully
this
is
going
to
mean
that
you
can
effectively
create
things
for
a
cluster
prototype
things
for
cluster
without
ever
landing
any
code
in
the
main
crew,
benetti's
repo
and
potentially
maintain
it
forever
that
way
right
and
so
that
I
think
this
is
going
to
provide
a
great
deal
more
flexibility.
D
You
can
implement
your
API
objects
in
Java
or
Python
or
bash,
or
whatever
you
want
to
do
right
and
I.
Think
that's
going
to
add
a
lot
of
flexibility
for
users
as
we
go
forward,
so
I
highly
encourage
people
to
think
about
how
they
might
structure
things
like
that
allegations.
That's
it.
Oh
I
should
also
say
that,
as
I
did
this,
I
found
some
bugs.
So
I'm
going
to
be
continuing
to
sort
of
push
out
and
fix
the
bugs
that
that
exist
still
cube
control
edit,
doesn't
work
with
these
objects.
For
example,.
E
D
It's
a
little
bit
different
right,
I
mean
there's,
there's
the
there's
the
this
is
really
about
introducing
new,
whereas
an
experimental
scheduler
is
about
different
binary,
manipulating
existing
objects,
all
right
and
so
I
think
we
have
a
different
mechanism
for
experimental,
schedulers
and
I.
Think
it's
necessary
because
we
want
to
change.
D
E
D
I
talked
about
this
in
different
contexts
and
I.
Think
so
one
option
would
be,
of
course,
that
you
could
actually
I
mean
you
do
an
experimental
scheduler.
This
way
you
could
create
sort
of
a
shadow
pod
object
and
then
have
the
experimental
scheduler
read
those
shadow
pods
and
then
schedule
the
pods
directly
by
setting
host
I
mean.
D
Sense,
that's
exactly
what
Damon
said
does
right,
but
at
the
same
time
we've
had
a
discussion
about
how
you
have
opaque
params
like
I.
Think
what
you're
talking
about
is
more
like
opaque
params,
which
is
I
kind
of
want
to
shove.
Some
data
inside
a
pod
object
and
have
the
scheduler
know
about
it
without
having
the
the
underlying
system
know
about
it
and
right
now
that
would
have
to
be
done
via
annotations
and
it's
kind
of
cruft
II
we've.
F
And
we
have,
we
are
putting
JSON
and
side
of
annotations
I'm.
This
is
exactly
how
we're
adding
new
experimental
scheduling
features,
we're
adding
them
as
annotations
to
exit
pods,
so
pod
affinity
and
anti
affinity
notified
Indian
anti
affinity
paints.
All
these
things
are
being
added
via
annotations
today
right.
So
that's
the
right
to
do
that
kind
of
thing.
A
D
D
The
other
question
was
the
actual,
a
third-party
object
definition.
So
what
what
the
very
first
thing
the
script
does
is.
It
goes
and
it
creates
an
extension
object
using
this
object
description
and
then
it
goes
to
the
api
path
for
those
newly
created
object,
resources
and
it
just
watches
the
objects
in
those
resources,
and
it
looks
for
changes,
and
you
know
more
thorough
example.
Would
then
given
the
changes
would
actually
take
some
effect
on
the
on
the
world?
D
D
That's
the
big
weakness
right
now,
which
is
that
that
all
of
the
mainline
furinkan
structure
that
we've
built
for
object,
conversion
and
object,
validation
and
admissions
control
currently
does
not
apply
to
this
stuff
and
there's
been
a
great
deal
of
discussion
about
how
we
might
go
forward
with
that
and
I.
It's
clearly
something
that's
necessary
in
the
limit,
but
I,
don't
think
there's
a
complete
consensus.
Yet
Brian
might
correct
me
if
I'm
wrong,
but
I,
don't
think,
there's
a
complete
consensus
about
how
to
do
it.
We
have
some
ideas.
D
You
know
it
would
involve
hooking
out,
so
this
module
container
would
implement
a
web
service
in
addition
to
monitoring
the
api's
and
so
validation
and
object,
translation
would
happen
by
having
the
API
server
call
out
to
this
module
to
say:
hey
you
know,
validate
this
or
you
know
default.
You
add
defaults
to
this
or
whatever
else,
yeah.
F
And
we
need
something
similar
to
out
of
process
admission
control
yeah,
so
probably
I
would
try
to
find
a
way.
You
said
same
mechanism,
whatever
we
build.
One
thing
we're
discussing
is
called
the
initializers
proposal,
where
we
would
postpone
any
action
in
the
API
server
on
an
object
until
all
initialization,
including
validation
and
whatnot,
were
finished.
That
changes
a
client
experience
pretty
dramatically
so
I'll
weirded
still
discussing
other
alternatives
is.
D
B
G
F
D
Also
part
of
it
right
is
that
there's
also
been
a
discussion
of
actually
separating
spec
and
status
in
a
couple
of
resources.
Cuz.
We
don't
want
to
have
to
do
because
we,
there
are
objects
out
there
that
have
lots
of
status
updates,
but
very
little
speck
changes.
Pods
are
an
example
of
this,
and
you
don't
actually
want
to
have
to
like,
read
and
write
that
full
object
all
the
time
and
have
conflicts.
The.
H
H
D
I
mean,
I
hope
the
stuff
just
is
a
is
a
question
of
making
sure
that
when
we
write
a
lot
of
this
stuff,
it
is
dynamic
in
terms
of
discovering
objects
via
the
API
endpoints
right.
So
a
lot
of
the
work
that
I
did
to
support.
This
was
basically
going
through
cube,
control
and
a
bunch
of
other
places
and
in
places
where
things
were
hard
coded
making
it
talk
to
the
server
and
download
all
the
API
objects
front.
D
All
right
and
so
I
would
hope
that
if
we
do
that
in
all
the
right
places,
it
should
just
work
because
these
things
should
be
imperceptible,
like
they
shouldn't
be
perceptual
differences
between
them,
but
obviously
it's
hard.
If
you
want
to
pre
assert
like
if
you
want
to
pre
assert
that
relationship
before
the
object
exists.
Just
in
case
it
may
be
exists
that
becomes
difficult.
I.
H
Guess
what
I
was
looking
for
is
like,
whatever
we
do
to
allow
you
to
describe
a
kind
in
the
API
server
machinery.
We
need
to
make
sure
it
stays
in
sync
with
how
I
describe
my
third
party
resource
so
like.
If
we
do
allow
concepts
will
say
you
know,
this
is
not
something
that
the
supports
dependencies.
You
know
I
want
to
make
sure
I
have
that
knob
on
third-party
resource
and.
H
D
I
mean
I
think
that
if
we
do
that,
if
you
do
all
that
stuff
and
you
make
sure
that
it's
all
keyed
by
API
group
and
version,
it
should
just
work,
I
would
think
right
like
if
you
make
sure
that
when
you
declare
those
things
you
declare
them
with
a
full
API
group
version,
etc.
Api
group
version
kind,
then
it
should
I,
would
imagine
it
should
work
right.
A
Awesome-
and
there
was
one
more
question
which
may
potentially
be
Quentin
just
trolling
us
have
you
given
any
thought
to
portability
between
clusters?
Ie.
My
application
relies
on
external
resources,
XY
and
Z,
which
happened
to
exist
in
my
test
cluster,
but
nobody
installed
them
in
my
production,
cluster.
D
D
We
may
have
some
sort
like
I,
think
higher
level
tools
start
looking
at
these
kind
of
dependency
management,
where
it's
like
make
sure
that
Cassandra
is
deployed,
I,
guess,
I,
don't
see
it
any
different
than
any
other
kind
of
dependency
and
I
would
hope
that
any
dependency
management
tools
that
are
built
on
top
of
Cooper
Nettie's
would
understand
those
dependencies
as
well
and
understand
kiss
I.
Don't
think
that
I
want
to
do
any
kind
of
deep
dependency
stuff
here
any
more
than
like.
D
If
you
haven't
inserted
the
module
that
handles
your
USB,
you
know
wireless
driver
and
you
insert
that
USB
stick
I
mean
I.
Guess
at
this
point
Linux
actually
will
they
do
actually
have
a
dependency
manager,
but
it's
not
the
colonel
I.
Don't
think
the
colonel!
You
know
the
kernel
itself
doesn't
go.
Hey
I
should
insert
this
module,
it's
some
other
third-party
Damon
that
looks
in
scans
USB
and
sees
a
version
code
and
then
does
the
in
smoke.
For
you.
It's.
F
B
H
D
A
D
D
So
I
would
say
that
the
stuff
that's
in
there
is
just
it
was
not
quite
done,
and
so
it's
you
can
turn
it
on.
You
can
kind
of
play
around
with
it,
but
it
doesn't
have
cute
control,
so
probably
isn't
a
very
good
control
support
and-
and
it's
not
forwards,
compatible
with
the
stuff.
That's
it
head.
Then.
A
D
There's
this
stuff
all
becomes
a
very
meta
because
there's
the
like
the
definition
of
the
there's,
the
definition
of
the
object,
which
is
like
the
class
definition
and
then
there's
the
instance
of
the
object
itself
and
they're,
both
third
party
and
so
I
sort
of
use
third-party
resource
to
define
to
describe
the
definitions
of
the
objects
and
third-party
object
to
describe
the
in
of
that
resource.
Although
actually
in
the
code,
it's
called
third-party
resource
data
anyway,
it's
the
tailor,
use
we're
kind
of
the
code.
Reviews
were
kind
of
fun.
A
G
G
The
problem
for
people
who
are
not
following
closely
on
that
proposal
is
that
a
couple
things
contributors
don't
know
what
the
expectations
are
for
a
PR.
They
don't
know
they
should
code
first
or
if
they
should
design
first
or
how
many
tests
they
should
write,
how
good
it
should
be
when
they
first
contributed
versus
later
on
and
who
they
need
to
go
to
you
to
get
approval
for
specific
parts
of
code
for
some
or
specific
parts
of
the
feature
addition
process
for
code.
G
We
have
a
proposal
to
add
owners,
files
and
I'm
assuming
that
that
gets
implemented.
Another
problem
is
there's
like
multiple
approvals
needed,
but
there's
no
one
who's
single
job,
so
multiple
approvals,
including
it
violates
first,
stick
a
new
API
reviews,
inquest
legs
course
that
can
decide
if
you're
using
the
right
test.
Primitives,
we
have
documentation
experts,
there's
no
single
expert,
you
can
sort
of
say.
G
Is
this
feature
done
any
of
there
were
single
person
like
we
don't
have
sort
of
the
staffing
to
have
people
that
just
Shepherd
issues
all
the
time
from
google
or
from
other
people
who
work
full-time.
So
we
would
like
to
expand
to
the
community
so
that
project
management
can
be
more
distributed
process.
If
you
want
to
drive
something,
will
tell
you
what
you
need
to
do
to
drive
it,
but
you,
as
that
feature
owner,
are
the
driver
for
it.
G
So
it's
about
not
just
giving
you
responsibility
to
decide
what
happens
with
your
feature,
humming
some
power,
but
also
responsibility
to
make
sure
those
through
all
those
steps
and
the
way
we
make
sure
that
you
sort
of
follow
up
follow
through
on
that
responsibility
is
you're
going
to
generate
a
summary
for
us
that
someone
else
can
look
at
quickly
and
say:
yeah.
They
did
do
all
the
steps,
but
the
sort
of
busy
work
of
going
through
all
the
conversations
of
getting
an
approval.
G
That's
on
someone,
who's
driving
a
feature
and
one
a
kind
of
related
issue
which
is
not
totally
addressing
but
got
me
thinking,
was
that
we
had
the
problem.
The
community
cannot
label
issues
or
PRS.
This
is
only
somewhat
solving
that
problem,
but
if
people
familiar
that
problem-
and
there
has
actually
been
to
work
around,
this
you've
been
discovered
recently.
But
generally
we
don't
want
everyone
to
be
a
committer,
but
only
committers
can
put
labels
on
things
at
the
get
up.
G
So
the
solution
that
I'm
proposing
involves
a
new
thing,
which
I
call
a
feature
issue
and
a
feature
issue-
is
a
structure,
type
of
issues
and
umbrella.
There's
going
to
be
many
other
might
be
sub
issues
under
there,
but
for
each
major
feature
that
we're
going
to
propose
to
add
we're
going
to
have
a
feature
issue
that
tracks
that
feature-
and
you
might
say
why
not
also
efforts
like
making
testing
less
flaky
or
making
documentation
better
well
a
feature
I
can
figure
out,
I
figured
out
what
are
defined
lifetime
instead
of
steps
are
for
that.
G
I
have
not
been
able
to
generalize
that
two
more
open-ended
efforts
like
have
no
test
flakes
or
have
awesome
docks
where,
as
a
feature,
it
goes
through
a
series
of
releases,
and
you
can
kind
of
you
know
what
you
have
to
do
and
you
can
know
when
it's
done.
So
that's
why
I'm
focusing
on
features
and
so
with
a
feature.
There
is
this
notion
of
a
checklist
of
approvals
that
you
need
this
checklist?
Will
grow?
Hopefully
it
won't
be
too
heavy
weight,
but
we
also
want
to
maintain
quality.
G
G
So
this
is
maybe
the
most
controversial
part
of
the
proposal,
which
is
that
I
want
to
put
these
feature
issues
in
a
new
repository
called
Corona
d,
/
features
and
the
thought
there
I'll
explain
why
another
repository
in
a
second
and
the
fourth
aspect
to
this
proposal
is
that
we
need
some
kind
of
policy
for
which
changes
require
feature
issues.
We
definitely
don't
want.
Every
time
you
set
a
PR,
you
have
to
follow
this
heavyweight
process.
We
only
want
this
heavy
a
process
for
big
changes
to
the
system
and
I.
G
Don't
have
an
exact
talking
for
that.
But
if
people
have
read
of
this,
we
would
define
a
document.
That
said,
like
this
type
of
thing
requires
a
feature
issue:
can
this
type
of
thing
doesn't
so
I
would
imagine
stuff?
That's
not
visible
to
users
like
fixing
a
test.
Refactoring
code,
you
know
adding
a
rate
limit
to
a
server
that
would
not
require
a
feature
issue,
something
that
has
an
API
change
or
a
new
type
of
object.
That
would
definitely
be
a
feature
issue
as
a
middle
ground.
G
G
Doc
update
so,
if
you
add
a
feature,
then
part
of
the
checklist
for
that
feature
should
be
to
add
documentation
for
and
a
documentation
owner
should
like.
Look
at
your
feature,
description
and
say:
okay,
he's
adding
a
food
kind
and
I,
don't
see
a
page
for
kind,
foo,
you're
not
done
but
yeah,
but
thought
should
be
easier
to
understand.
That's
not
a
well
defined
process.
No.
E
G
Think
it
can,
you
said,
I,
don't
think
it's
the
same
process
if
someone
else
wants
to
come,
posle
convincing
me
what
the
checklist
is
for,
like
a
docs
change,
I'd
love
to
hear
that
this
isn't
a
process
to
generate
free
labor.
It's
a
process
to
like
enable
someone
to
follow
me.
Our
standards,
I
guess.
I
Another
way
of
thinking
about
that
is
is
whether
we
want
to
like
create
some
retroactive
feature
issues.
If
we
know
that
there's
a
feature
that
exists,
but
it
doesn't
add
bucks
and
it
doesn't
have
tests
and
doesn't
have
whatever,
rather
than
creating
feature
issues
to
fix
the
docs
or
to
fix
the
test.
You
know
go
back
and
create
a
feature
issue
for
the
whole
feature,
and
then
you
know
tick
off
the
stuff.
That's
there
and
the
stuff
that
isn't
there,
you
know,
is
now
reflected
in
each.
G
G
Have
these
like
requirements
for
graduation
to
alpha
and
beta
I
think
all
those
that
would
be
kind
of
a
model
for
we're
talking
about,
and
we
would
probably
open
issues
for
things
that
are
still
in
alpha
or
beta
and
track
them
under
this
new
system
and
I
guess
other
stuff
is
already
stable,
but
we
want
something
on.
Maybe
we
don't
get
issues
for
those
go
ahead
line.
F
F
G
I
welcome
to
be
a
federal
system
where,
like,
if
you're,
just
changing,
docks
that
there's
no
top-level
process
talk
with
the
docks
community.
If
you're,
just
changing
code
then
like
in
couplet
to
like
fix
a
bug,
then
just
talk
to
them
only
if
it
cuts
across
the
whole
user
experience.
Do
you
need
to
track
it
at
this
eye
level?
That's
my
thinking.
So.
A
Let's
double
back
to
Bob's
can
contribute,
though,
because
I
think
as
much
as
the
flip
answer
of
contributes
to
go
away
because
it's
confused
it
is,
it
is
definitely
an
answer.
I
would
just
be
slightly
broader
in
this
with
which
is
Bob,
that's
out
of
scope
for
this
discussion,
simply
because
we
need
to
figure
out
what
to
do
with
Q
trip.
It's
got
a
mishmash
of
stuff
in
it
right
now,
and
we
can.
We
can
come
up
with
a
broader
plan
around
that,
but
I
would
say
that
this
is
not
specific.
E
A
That
was
not.
That
was
not
my
intent
either
it's
more
I.
We
should
address
contribs
with
intent
and
vision
and
figure
that
out
because
contribs
a
very
useful
space,
but
it
may
not
be
as
a
repo
under
Cooper
Nettie's
org,
because
that
limits
who
has
rights
on
it
so
we've
talked
about
a
broader
organization
potentially
or
it.
F
A
G
Pitcher
issues
is
that
you
don't
open
one
until
you're
like
it's,
not
for
the
initial
discussion
of
what
we
should
do.
We
can
still
keep
using.
We
can
use
cigs
or
the
you
know:
Cooper
Nettie's,
/
communities,
issues
to
talk
about
that
not
proposing
to
move
all
our
existing
issues
like
it's
something
you
open
when
you're
really
ready
to
drive
something
and
the
person
who
opens
it
should
be
the
person
when
they
open
their
saying
I
want
to
be
the
driver
for
this.
G
So
it's
a
way
to
say
I'm,
taking
ownership
of
this
issue
and
you
open
it
when
you're
ready
to
start
work
and
it's
for
things
that
have
user
visible
behavior,
not
for
internal
refactoring
and
I
expect.
These
issues
are
going
to
live
a
long
time
because
they'll
track
the
alpha
beta
and
stable
stages
of
a
features
lifetime.
So
that
probably
means
they
live
for
at
least
three
releases
in
most
cases.
G
So
the
reason
I
said
we
should
put
the
exam
repository
was
one
we
can
have
different
ACLs.
Another
one
is
that
issue
templates
seem
like
a
really
good
way
to
present
the
workflow
to
users,
because
it
puts
this
thing
in
a
checklist
at
the
top
of
the
issue,
and
you
can
only
have
one
issue
template
for
repository.
So
that's
kind
of
thing
that
was
really
driving
me
to
a
new
repository.
The
other
thought
was
that
big
features
are
going
to
span
multiple
repositories
already.
G
You
need
dachshund
docks,
koduku
benetti's
we've
talked
about
splitting
pernetti's,
like
maybe
Kublai,
goes
to
its
own
repository
a
year
from
now
so
you're
going
to
be
spanning
repositories
with
large
work
anyhow.
So
it's
kind
of
not
a
big
deal
to
have
a
separate
repository
and
the
fact
that
it's
single
purpose
to
me
when
you
go
there,
it's
more
focused
what
you're
thinking
about
actually
found
the
fact
that
Doc's
removed
to
a
new
repository
I
was
really
skeptical
down
at
the
beginning.
G
I
found
it's
been
fantastic,
that
repo
has
its
own
workflow
and
it's
it
works.
Well,
so
I
think
by
its
own
repo
he'll
have
its
own
workflow.
Its
own
automation
it'll
be
clear
when
you
go
there
what's
going
to
happen,
so
this
is
like
a
sketch
of
a
feature
checklist
that
you
might
have
it
has
alpha
beta
and
stable
different
stages
have
different
things
that
you
need
to
do
like
get
an
API
design
review.
G
Make
sure
that
you
have
some
testing
and
it
has
a
list
of
teams
that
you
go
to
for
testing
so
rather
than
like
spending.
A
lot
of
time
on
this
slide.
I
actually
want
to
go
to
a
demo
which
I've,
created
and
Sierra's
going
to
help
me
with
a
demo,
so
we're
both
going
to
go
to
github.com
/
eric
tune.
/
features,
hi
I'm,
someone
who
wants
to
drive
a
new
feature
so
I'm
going
to
create
a
new
issue.
If
you
guys
all
see
github
so
I
make
a
new
issue.
G
I
never
done
this
before,
but
I
really
want
resources
and
cooper
Neddy's
to
have
a
color.
So
I'm
going
to
title
this
colored,
we
sort
colors
for
resources,
I'm
a
semantic
issue.
That
issue
is
going
to
make
the
template
be
submitting
it's
going
to
make
the
template
be
readable.
Now
I'm
going
to
read
the
template
and
see
what
I
need
to
do:
okay,
there's
advice
on
what
alpha
and
beta
is
table
mean
before
its
alpha.
I
know
I
need
to
propose
my
feature.
G
G
Okay,
so
here's
my
design
proposal,
all
objects
forget
how
to
color.
I
presided
a
standard
for
how
those
colors
should
be
specified.
I
think
my
design
proposals
is
thorough
enough
that
someone
can
review
it
now
then
ask
for
a
first
step
is
a
new
feature:
reviews
911
set
of
this
team,
so
I'm
going
to
pretend
the
team
is
somebody
else.
I'm
gonna
say.
G
And
then
I'm
in
a
way
Sarah's
been
at
mentioned.
This
would
actually
be
a
team
to
be
multiple
people.
There
be
some
protocol
for
them
to
decide
who
picks
it
up
that
beeper
team,
she's
gonna,
say
it
looks
good
to
me.
I'm
gonna,
head
at
this
okay
I
proposed
it
and
I
got
an
LG
team
in
the
design
review.
Now
anyone
who's,
not
Sarah,
not
me,
but
like
maybe
a
release
manager
can
go
through
and
say:
okay,
I,
look
at
as
we're
coming
up
to
the
release.
Here.
G
H
E
H
H
G
H
G
J
J
Think
just
adding
a
step
to
this
checklist
that
says
just
write
up
how
the
feature
was
actually
done
and
maybe
add
enough
to
the
original
design,
doc
or
shuffle
things
around
whatever
I
think
that
captures
that
discussion
in
source
control
and
it's
I
have
no
preference
about
whether
it's
this
repo
or
the
Cooper
daddy's
repo,
but
wherever
the
docs
for
this
sort
of
stuff
ultimately
live.
So
one
of
the
Dutchman's
checklist
should.
G
K
H
See
me
go
to
the
documentation
for
that
right,
but
I
want
to
make
sure,
but
I
don't
I
don't
think
the
summary
right
up
at
the
end
is
enough
right.
I
I
I
think
if,
if
sarah
was
to
comment
on
your
initial
things,
I
don't
like
your
color
choice
like
I
want
to
avoid
having
those
long
discussions
in
this
feature
issue.
So
I
still
think
we
need
to
have
a
doc.
That's
the
in-progress
iteration
on
the
one-page
summary
of
your
future
I.
Think.
H
J
A
F
G
F
That
sounds
good.
One
quick
comment,
I
have
is
that
generally
alpha
beta
is
stable,
are
going
to
be
in
different
releases,
so
be
useful
to
have
them
as
separate
tracking
issues,
so
you
can
assign
milestones
to
the
individual
like
before
alpha.
You
know,
alpha
presumably
will
be
targeted
in
particular
release.
Milestone
made
it
with
me
a
different
release,
milestone.
F
G
G
F
A
It's
great
that
we
can
go
forward
and
iterate
I'm
fine
with
that
I
haven't
heard
anything
that
slows
us
down
in
that
way
other
than
trying
to
make
it
better
and
it
is
11
0
1,
so
people
are
going
to
start
disappearing.
So
thank
you
all
and
if
please
raise
your
hand
if
there
was
any
big
concerns
about
the
future
stuff,
but
it
sounds
like
a
great
spot
to
start
and
iterate
from
here
see
you
all
next
week,
great
work
Eric
see
comments.
Yay.