►
From YouTube: 20190129 sig cluster lifecycle
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
So
I
personally
didn't
put
something
on
here:
Paris
sent
out
an
email,
just
sort
of
reminding
everybody
on
behalf
of
say
contradicts
that
if
you
have
not
already,
please
join
the
communities.
Dev
mailing
list,
a
lot
of
important
announcements
get
sent
there
and,
if
you're
not
on
that
list,
you're
probably
missing
announcements.
It's
also
for
folks
that
are
creating
Google,
Docs
and
sharing
them.
B
A
lot
of
organizations
will
prevent
you
from
sharing
Docs
publicly
until
what
people
tend
to
do
to
work
around
that
at
the
share
Docs
with
the
community
is
they
will
share
with
the
sequester
lifecycle
mailing
list
or
with
the
Google
Victoria's
dead
minute,
like
Google
Groups
mailing
lists
until
a
lot
of
times,
if
you
can't
get
access
to
a
doc,
if
you
join
one
or
both
those
manual
Isis,
that
will
fix
the
problem
for
you.
So
there's
lots
of
great
reasons
to
join
the
mailing
list.
Please
do
so.
C
B
But
if
you're
not
friendly
reminder
to
do
that,
next
Justin
sent
out
a
pull
request.
I
think.
Yesterday
we
talked
last
time
about
sort
of
moving
forward
with
our
plan
for
add-ons
for
the
earth.
So
we
finally
actually
I
think
maybe
have
the
beginning
of
a
plan
has
a
sink
Justin.
You
want
to
talk
about
your
request.
I
know
that
some
of
there
have
been
some
people
that
have
looked
at
it,
but
it
just
went
out
yesterday,
so
some
people
haven't
also
yeah.
D
Thank
You
general:
the
link
is
in
the
agenda,
but
it
is
an
in-house
one.
Seven,
four
six.
It
basically
describes
a
broad
plan
to
investigate
managing
add-ons
using
operators,
though
we
have
code
that
Jeff
Johnson
might
have
worked
on
for
the
GK
on-prem
effort.
We
don't
imagine
that
we
can
just
like
code
dump
that
into
OSS,
and
so
we
are
proposing.
Basically,
you
know
examining
it
or
going
from
from
first
principles
like
thing.
D
This
is
a
capture
sort
of
act
as
the
umbrella
for
that
work
and-
and
hopefully,
yeah
discusses
sort
of
the
broad
ideas
that
we
have
and
hopefully,
if
anyone
has
any
feedback
that
we
very
welcome.
There
are
some
very
open
issues
about
how
we
structure
the
project
but
I
think
before
we
even
get
to
that
like
be
like.
What
are
we
building?
That
thing
is
sort
of
the
first
more
important
thing,
because
if
so,
even
kappa
yeah
so.
A
D
A
Great,
is
there
like
some
again
with
administrivia?
Did
you
want
to
have
like
the
typical
pattern
that
we
did
with?
The
other
kept
is
like
give
it
a
couple
weeks,
sort
of
time
out
for
feedback
and
then
one
more
time
during
the
cig
meeting
to
try
and
see
if
we
can
settle
in
any
ambiguities
or
questions
yeah.
D
D
Also
so
like
there
was
some
great
feedback
already
about
you
know
identifying
from
Mirabel
like
I
didn't
find.
The
the
use
of
patches
does
make
updating
the
underlying
manifests
harder,
which
is
I,
think
good,
a
good
feedback
to
a
piece
of
a
good
observation
and
something
we
should
call
attention
to
as
a
I
would
still
like
to
proceed
as
a
trade
off.
But
you
know
something
which
definite
call
attention
to,
and
so
that
sort
of
feedbacks
super
helpful.
E
D
It
just
punted
on
that
question:
I,
don't
think
one
of
the
things
we
were
very
does
call
out
explicitly.
Is
that
no
we're
not
gonna
require
tooling
to
adopt
this
right?
If
you
don't
want
to
adopt
these
operators,
you
don't
have
to
I.
Think
the
operator
operator
will
be
something
that
is
even
more
dependent
like
I.
Do
hope
that
everyone
adopt
operators
for
like
the
core
components.
D
I
think
the
operator
operator
is
one
that,
if
we
build,
would
be
very
much
up
very
like
very
optional
and
like
some
people
will
use
it,
some
people
won't.
There
will
be
different
operators
for
different
operator
operators
for
different
scenarios
like
if
you're
doing,
cluster
API.
It
might
be
that
the
thing
which
is
building
your
cluster
is
a
natural
operator
operator
right,
so
that
one
is
there
will
have
to
be
something
that
applies
the
bottom
turtle,
but
it's
it's
not
clear
that
that
is
necessary.
An
operator
in
the
same
way.
E
D
Kept
as
outlined
describes
the
idea
of
that
each
atom
would
have
its
own
kind.
The
two
reasons
for
that
as
I
see
them
are.
We
want
operators
to
be
built
by
the
component
teams
so
like
the
coup,
DNS
team
should
build
their
own
operator
or
the
Cortinas
team
should
build
their
an
operator,
ideally
in
our
end
state,
because
they
best
know
how
to
how
to
manage
it,
and
so,
from
that
point
of
view
it
it
was
a
weird.
No
that's
it
seems
to
fit
that
better,
the
other
one
of
which
is
I.
D
My
feeling
with
cluster
API
is
if
the
extension
block
is
sort
of
awkward
I,
don't
know
how
other
people
feel
about
that
and
that
it
might
be
easier
to
not
to
use
duck
typing.
Sorry
might
be
easier
use
duck
typing
for
common
fields
rather
than
a
common
fields
with
a
wrapper,
but
I
think
this
is
something
we
can
investigate
in
our
prototypes.
We
did
different
kinds,
it
does
mean
you
end
up
with
a
plethora
of
kinds:
it's
yeah
if
you'll
have
views
and
if
your
view
so
I'm
happy
to
hear
that.
Well,.
A
B
Can
we
don't
have
a
lot
on
the
agenda
today?
I'd
be
curious
for
people
that
have
sort
of
built
their
own
add-on
manager
systems?
If
they're
here
to
maybe
give
like
a
brief
overview
of
what
the
requirements
might
be,
and
then
we
can
also
point
them
if
they
kept
to
review
those,
but
not
another
time
about
how
you
solve
your
problems
but
like
what
the
problems
were,
that
you
found
right
like
I,
think
a
mini
cube
may
have
different
set
of
requirements
than
gke,
then
cops,
then
something
else
right
and.
D
I
can
certainly
speak
to
like
there
are
a
lot
of
problems
in
the
add-on,
the
bash
add-on
manager,
which
I
think
everyone
wants
to
avoid
the
cops
this
one
tries
to
avoid
them,
but
make
some
other
trade-offs.
Most
notably,
is
the
one
that
it's
very
hard
to
change,
one
of
the
built-in
add-ons
in
any
form,
and
that
when
you
do
change
it,
you
get
you
get
a.
It
gets
referred
like
a
minute
later
or
five
minutes
later,
which
is
sort
of
a
pretty
obnoxious,
behavior
I.
Think
of
the.
D
D
A
B
So,
at
least
for
mini
cube
I'm,
not
well
versed
in
the
add-on
system,
yeah,
but
it
seems
pretty
basic.
It's
just
a
couple
of
EML
files
that
get
reconciled
and
bundled
into
the
binary.
Unfortunately,
that's
kind
of
the
painful
part
that
we
have
is
we
basically
support
like
an
open
marketplace
where
anyone
can
add
their,
you
know
submit
their
add-ons
and
we
definitely.
We
have
some
some
growing
pains
in
that
area
like
how
do
you?
How
do
you
vet
it
if
an
add-on
is
actually
properly
working?
B
And
how
do
you
actually
like
get
the
tickets
to
the
person
who's?
Maintaining
it?
We
would
really
love
to
move
to
a
model
where
every
add-on
ends
up
with
its
own,
like
git
repository
and
people
open
bugs
on
the
specific
get
repository
and
it's
updated
separately
from
the
mini
cube
releases.
So
that's
decoupling.
That
would
be
really
nice
for
us.
I.
Think.
A
Also,
decoupling
for
the
air-gap
scenario,
too,
would
be
nice,
especially
as
the
add-ons
grow,
because
that
one
of
the
common
problems
and
the
things
that
we
saw
in
Cuba
TM
was
that
to
get
a
user
story
that
wasn't
a
terrible
terrible
for
how
to
manage
their
container
life
cycle
for
the
air-gapped
environment.
And
we
do
the
exact
same
thing.
We
just
apply
this
but
lava
VM,
oh,
but
that's
not
tenable
across
all
these
there's
writers.
F
Yeah
we're
we're
in
pretty
much
the
same
boat
and
air
gapping
is
important
to
kind
because
anywhere
we're
using
the
Internet
it's
somewhere.
We
can
flake
and
CI
and
users
have
just
recently
started
requesting
add-ons
and
I
think
we're
trending
towards
the
same
direction
of
we
don't
really
want
to
be
shipping
them
ourselves,
but
having
a
good
way
to
manage
that,
that
is
something
that
can
be
added
on
itself
would
be
really
useful.
D
The
there
is
a
potential
solution,
I
wanted
like
fixate
on
the
solution,
but
the
Josh
hoax
cluster
bundle,
work
which
is
open
sourced,
potentially
provides
a
solution
there
and
that
it
is
a
specification
that
can
live
like
on
an
HTTP
endpoint
somewhere.
So
you
can
pull
it
or
because
those
objects
are
also
well
API
to
have
CR
DS.
They
could
be
put
into
the
cluster
by
bottom
turtle
and
then
the
the
then
your
your
then
a
or
gaps.
At
that
point.
A
What
becomes
a
little
bit
thorny
to
cuz
we're
kind
of
treading
into
Pandora's
box,
but
I
guess
we
got
time,
might
as
well
dip
the
toe
in
to
the
tornado
and
see
what
happens.
The
you
were
entering
add-ons
when
doing
with
with
some
of
this
stuff,
because
it
could
mean
iam,
explicitly
punted
this
stuff
to
say,
like
here's,
the
images
in
the
amyl
and
you
can
go
apply
it.
But
whenever
we
get
to
the
application
space,
then
we're
starting
to
get
tread
into
that
sort
of
meta,
templating
problem
that
exists
in
helm.
D
Very
much
want
to
close
that
Pandora's
Box.
The
anything
to
do
with
application
management
is
very
much
at
a
scope
of
this
work.
Even
if
users
choose
to
use
it
for
it,
we
are
not
going
to
target
that
with
I.
Do
think
that
some
of
the
templating
issues
do
arise
and,
for
example,
users
often
want
to
change
the
resources
allocated
to
add-on
right.
We
might
have
other
solutions
for
that,
like
autoscaler
type
things,
but
we
don't
want
to
get
into
the
world
where
we
have
to
map
every
single
field.
D
In
my
opinion,
and
that's
where
that's
where
template
that's,
where
patches
like
to
cuddle,
apply
or
like
customize,
so
I
could
cuddle
patch,
where
that
customize
could
come
in.
That's
where
Lumiere
pointed
out
correctly
that
you
know
there
is
a
if
the
underlying
structure
has
changed
significantly
your
passion
or
longer
apply,
and
that's
I
think
the
trade-off
you
if
you're
not
going
to
map
it.
You
can't
yeah,
you
risk
breaking
the
underlying
structure
and
then
breaking
patches.
A
G
B
Big
takeaway
for
us
I
know
it's
mentioned
in
Justin's
kept,
but
I
think
it's
something
that
maybe
we
should
highlight
a
little
bit
more
and
make
sure
that
we
sort
of
really
are
addressing
that
as
a
core
use.
Kiss
that's
kind
of
what
I
was
going
for
is
like.
Are
there
important
things
that
we
really
need
to
talk
about
in
the
cap
to
make
sure
we're
actually
going
to
get
the
broad
adoption
that
Justin's
targeting
like?
B
E
But
what
is
the
other
difference?
Is
there?
Is
there
any
other
difference
from
like
say
a
thing
that
install
is
installed
a
CSI
thing
to
my
cluster,
to
make
all
my
net
storage
drivers
start
devices
work
or
a
CD
operator
which
just
is
used
back
for
my
web
service,
whatever
form
applications?
What
is
they
both
use?
Operators,
they
both
use
the
same
series.
They
have
different
API
groups,
but
one
is
required
for
the
cluster
to
function
of
the
mine
is
not,
but
they
still
have
today
no
visible
difference
right.
So
so
what
is?
E
What
is
a?
How
can
I
know
that
this
is
actually
and
I?
Don't
that
is
managed
as
part
of
the
cluster
is.
That
is
one
of
the
key
things
we
want
to
scope
it
as
like.
This
is
why
why
is
this
an
atom?
Well,
it's
an
another
big
course:
it's
if,
if
I
don't
upgrade
it,
when
I
have
trained
my
cluster
I'm
gonna,
be
in
pro
book
yeah.
B
Last
part
is
the
most
important
bit
and
Justin
touched
on
it
and
it
kept,
which
is
we
want
to
limit
the
use
of
this
for
Adam,
specifically
and
add
on
certify
and
as
things
that
are
tied
to
the
lifecycle.
The
cluster
itself,
whereas
generally
applications
are
their
life
cycle,
is
independent
of
the
life
cycle.
The
cluster,
where,
as
add-ons
like
you
said,
need
to
be
upgraded
sort
of
with
the
cluster
right
at
the
same
time
and
I.
B
Think,
given
that
definition,
we
can
vary
drastically
limit
the
scope
of
things
that
we
target
for
add-ons
where
something
like
Helmer
application
manager
has,
you
know,
very
broad
applicability.
Even
if
we
use
some
of
the
same
techniques,
we
can
say
we
don't
have
to
support
a
whole
bunch
of
these
use
cases,
because
we
have
a
very
targeted
set
of
things.
B
Can
tell
programmatically
what
is
being
sort
of
considered
a
cluster
add-on
I
think
that
will
help
a
lot
to
historically
Adams
have
also
run
it
run
a
cube
system
namespace,
which
is
another
sort
of
hint
that
it's
an
add-on
as
opposed
to
an
application.
So
I
think
there
are
different
different
things
we
can
do
in
terms
of
where
we
place
them
and
how
we
apply
metadata
and
consistently.
So
it's
easy
for
operators
to
understand.
What's:
what's
an
add-on,
it
was
just
an
application.
Yeah.
A
A
These
other
application,
things
which
are
kind
of
like
you
know
how
we're
gonna
do
HPA's
and
BPA's
what
what
am
I
going
to
use
heap
store,
whatever
technology,
the
matrix
server,
whatever
there's
a
there's
a
demarcation
line,
and
some
people
want
to
opt
into
those
things,
and
some
people
will
want
to
like
not
opt
into
them
because
it
was
considered
them
application.
Space,
yeah.
B
I
think
that's
important
to
consider
is
they're
like
you're,
you
and
him,
or
you
and
Lucas
we're
saying
there
are
things
that
are
required
for
the
cluster
to
actually
function,
and
there
are
other
things
that
people
might
look
like
on
gke,
you
can
say,
give
me
ingress
right
or
give
me
logging
and
monitoring
right
and
those
things
we
tie
their
lifecycle
to
the
life
cycle
of
the
clusters.
They're
considered
add-ons,
but
they
are
not
strictly
required
for
the
cluster
to
function.
Right
and
so.
B
I
think
that
the
definition
of
add-ons
is
things
that
you
tie
to
the
life
cycle.
The
cluster
and
therefore
different,
stallers
or
different
distributions
will
have
a
different
set
of
add-ons,
and
the
hope
is
that
the
things
that
are
common
across
those
distributions,
we
use
the
same
CR
DS
and
the
same
operators,
so
that,
if
you're
using
cube,
DNS
or
Cordy
and
s
that
everybody
uses
the
same
coordinates
operator.
B
E
So
getting
this
new
flow
into
the
DNS
whatever
and
then
adopting
that
across
all
solutions,
which
means
that
that
is,
that
is
a
proof
of
concept
or
just
proof
that
that
this
works
and
it
it
doesn't
harm
anything
because
it's
like
whatever
it's
just
a
same
coach
thing
but
but
still
it
would
be.
The
proof
that
here
is
the
the
new
I.
Don't
think
now
go
ahead
and
adopt
it
for
yourself.
That
would
be
and
then
offer
beta
in
GF,
of
course,
but
I
think.
D
That's
a
good
call,
I
think
I
did
put
in
some
graduation
criteria
that
we're
basically
more
along
those
lines,
but
fuzzier
like
saying
that
you
know
we
have
operators
for
most
for
most
things
and
that
most
tooling
adopts
it.
But
I
like
the
idea
of
having
like
concrete
graduation
criteria
of
that
nature.
I
do
worry
that,
like
cops
holds
out,
doesn't
adopt
the
add
ons
that
we
see
we
never
graduate
it
right,
but.
E
The
other
thing
is
like
the
proxy:
that's
the
proxy
need
an
operator
or
doesn't.
Are
we
sure
that
the
cube
controller
manager
a
find
in
a
deep
demon,
said
it's
cube
controller
manager
good
enough
to
address
all
the
business
legitimate
proxy?
This
thing
will
pounded
on
in
Cuba,
diem
and
said,
let's
hope
so
but
like
I,
don't
know
if
it
works.
In
fact,
yeah.
D
Actually
so
the
proxy
specifically
they
one
of
the
things
I
thought
might
happen
is
coughs,
does
not
use
a
diamond
seppuku
proxy.
Yet
and
in
theory
we
should
run
different.
Daemon
sets
based
on
the
node
version,
which
probably
means
multiple
daemon
set
C.
So
maybe
maybe
we
really
do
need
an
operator
there
to
like
manage
those
multiple
daemon
sets.
They
don't
become
like
you,
don't
have
200
daemon
sets
in
there
by
the
time
room
like
terminates
120,
right,
I,.
A
H
H
D
B
E
J
B
Thing
I
was
gonna,
have
Lumiere
suggested
on
the
PR
that
we
have.
You
know
separate
sub
projects,
slash
working
group
with
a
separate
meeting
for
this.
It's
like
the
other
thing
for
people
to
take
away
is
if
this
is
a
space
that
you
are
particularly
interested
in,
there
will
probably
be
sort
of
a
separate
place
to
store
dive
deep
and
discuss
this
in
depth
outside
of
this
meeting.
I
think
this.
D
I
will
but
I
I
welcome
all
help
with
logistics
at
all
times,
I'm
terrible
at
gistic,
but
I
suspect
I
will
create
a
beating,
I
presume.
We
should
do
then
I
don't
know.
What's
the
what's.
The
next
step
is
to
do
that.
What
we'll
meet
we
can
be
next
in
two
weeks
time
and
then
schedule
a
meeting
or
do
you
want
in
between
so
we
we.
E
Have
problem
with
the
with
the
current
zoom
setup?
Yes,
I
see
you
maybe
so
we
kind
of
we
ended
7
or
from
my
time,
7
p.m.
sharp
with
a
Ouija
component
standard
meeting.
But
that's
probably
I,
don't
know
if
there's
some
like
in
two
minute
lag
or
one
minute
lag
or
whatever
that
people
can't
join
this
we're
on
the
same
account
of
this
meeting
so.
A
E
A
Is
the
horse
sufficiently
bitten
to
death
here
in
this
one
all
right?
The
next
topic
I
want
to
discuss
with
the
broader
sig
is
Kuk
on
EU
in
developer
sessions.
Robbie
and
I
chatted
about
it
informally,
but
I
wanted
to
open
up
this
thing
to
see
what
sessions
folks
thought
were
appropriate.
We
have
a
lot
of
sub
projects
now,
and
the
question
is
like
how
many
sessions
should
we
start
to
apply
for
and
what
has
the
most?
What
has
broader
interests.
A
Does
I
don't
think
there
is
a
bound?
So
that's
the
interesting
thing,
but
I
do
think
we
should
do
some
some
level
of
logistical
grouping
right
having
a
per
provider.
Implementation
of
cluster
API
probably
be
too
much,
potentially
lumping
things
like
mini,
cube
and
kind
together.
If
folks
wanted
to
to
sort
of
have
this
sort
of
developer,
centric
view
of
the
world
is
another
possibility,
if
that
folks
can
start
to
talk
about.
I.
Definitely
think
that
cube
ATM
and
cluster
API
is
the
two
central
pieces
or
pillars
should
still
have
sessions.
B
Also
to
answer
that
question:
technically
they
say:
do
you
want
to
a
deep
dive
for
your
sig
last
time
we
asked
for
two
and
they
gave
us
two
so
I
think
this
time
we
can
try
to
ask
for
more
and
see
what
they
say.
There's
also
an
effort
to
combine
all
of
the
different
cloud
provider
SIG's
into
a
single
cloud
provider
sig
and
one
of
the
things
they're
deciding
as
they
do.
That
is,
if
we
take
you
know
these
five
different
SIG's
entry
into
one.
Does
that
mean
we
only
get
one
session
a
cubic?
B
Maybe
we
shouldn't
combine
SIG's
I,
think
that's
the
wrong
incentive.
I
think
for
people
I
think
we
should
group
people
are
working
together,
regardless
of
like
how
many
free
talks
you
get
Q,
Khan,
so
I
think
as
a
large
thing,
with
lots
and
lots
of
sub
truck
objects.
We
should
ask
for
more
and
as
long
as
we
can
justify
those
I
think
that's
reasonable
I.
D
E
B
It's
still
35
minutes,
it's
just
okay.
The
audience
is
supposed
to
be
different
right,
so
I
think
what
we
should
do.
They
should
do
one
intro
for
the
sake
and
the
intro
can
sort
of
do
a
sort
of
spectrum
of
like
here
that
the
high
level
topics
is
sick
is
interested
in
tell
them
to
put
that
for
a
little
in
the
week
and
then
have
that
have
four
pointers
to
the
various
deep
dives
and
say
you
know
we
have
a
really
big
sig
with
a
broad
set
of
interests.
B
We
have
a
number
of
deep
dives
at
our
stake,
is
sponsoring
here,
the
different
ones
and
and
here's
which
ones
you
should
go
to
based
on
what
you're
interested
in,
and
we
should
ask
for
sort
of,
as
many
as
we
think
is
reasonable
right.
So
I
think
that
you
know
Joe
and
I
were
sort
of
chatting
and
I
came
up
with
like
about
four
ish,
four
or
five,
and
then
we
can
see
what
they
say.
I.
E
Yeah
I
think
that
just
to
have
something
I
think
that's
the
same.
You
you
talked
about
well,
one
for
Cuba
name,
one
for
API,
then
I
like
the
Tim,
said
about
like
mini
Cuban
kind
being
together
and
then
maybe
we'll
just
make
cops
and
cubes
right
together
as
well.
If
Cuba
is
interested,
I,
don't
know
if
there's
someone
here
from
that,
and
then
we
have
like
the
cluster
API,
which
is
like
well
respecting
talking
about
that.
B
E
B
F
A
L
D
A
C
C
So
mostly
agreed
when
I
think
of
a
user
focused
deep
dive,
I
wouldn't
just
look
at
cluster
API
I
would
consider
they
mean
a
presentation
on
cluster
API
Kubb
ADM
kind
I
mean
like
all
of
the
different
pillars
that
we
have
and
show
how
you
can
manage
an
entire
lifecycle
cluster
using
all
of
the
tools
together
rather
than
focusing
on
one,
then
visually
you're
right.
You
would
have
to
pick
a
provider,
but
that
would
just
be
a
small
detail
in
the
entire
architecture.
I
think.
A
A
M
B
Think,
realistically,
if
we're
gonna
sort
of
do
demos
for
like
a
cluster
API,
it's
gonna
serve
recur
over
multiple
coupons
and
were
talk
about
this
tomorrow.
But
I'd
like
to
sort
of
rotate
sort
of
true,
is
giving
those
deep,
dives
and
I'm,
hoping
that
as
we
do,
that
we
also
give
different
demos
right.
We're
not
always
demoing
on
AWS
are
always
doing
on.
Google
are
always
in
the
middle,
but
like
we
rotate
through,
because
you
can
go
back
and
watch
the
talks
from
old
coupons.
B
So
right
you
can
go
back
and
watch
a
demo
they're
working
on
Google
and
maybe
watch
I
can't
watch
the
demo
to
work
on
abs
and
it's
you
know
six
months
old
now,
but
that's
okay,
like
you,
still
get
the
idea
that
it
works
in
multiple
places,
right,
which
I
think
is
a
good
message
to
send
over
time,
but
I
love
to
see
that
sort
of
rotate
and
not
say.
Well,
you
know
we're
gonna
fight
politically,
it's
just
whoever's
giving
the
talk
whatever
they
feel
most
comfortable
with
us.
They
should
use
that
yeah.
B
G
Is
it
over
sir
point
that
with
least
mention
that
having
a
kind
provider
that
should
be
arrested
because
they're
like
a
dog
foodie,
we
creating
a
provider
for
all
something
of
this
also
in
the
same
ecosystems
and
actually
are
quite
interesting
on
that
somebody
else
is
interested
because
I
think
is
for
developer
is
the
best
option
at
all.
Well,
why.
A
Wouldn't
you
put
that?
Why
would
you
put
that
directly
in
the
core
as
a
potential
option
of
Kloster
API
itself
proper,
alright,
like
because,
right
now,
cluster
API
itself
is
more
like
a
framework
and
if
it
was
like
the
default
developer,
use
case
for
the
core
seems
much
more
applicable
than
trying
to
create
yet
another
provider
which
doesn't
have
its
own
life
cycle.
F
A
F
A
A
E
Right
but
I
like
we
could
think
about
the
idea
of
having
say
like
a
lightning
talk
inside.
We
have
35
minutes
in
the
intro
and
I'm
I'd
be
happy
to
like
have
some
kind
of,
because
we
talk
about
the
overall
architecture.
There
we
talked
about
how
the
components
or
the
in
different
sub
project
we
have
fits
together.
So
why
would
we
why
couldn't
we
put
in
a
5
minutes
or
10
minute
demo
of
something
something
that
uses
them
all
together?
You
know
that
expected
way.
E
B
E
E
P
P
I
do
believe
that
it's,
it
would
be
super
nice
to
present
kind
and
mini
cube
together,
because
it
can.
It
can
be
an
opportunity
to
express,
like
from
a
sick
perspective
that
this
this
is
these
two
products
are
together,
they
have
their
own
use
cases,
and-
and
this
is
where
like
how
you
can
use
them
so
I-
think
it's
it's
a
it's
good
for
the
users
to
clarify
yeah.
B
I
mean
I
think
in
particular
with
those
two.
It's
really
useful
to
tell
people
like
when
you
should
use
which
one
right,
yeah
that'd
be
a
really
good
message
get
across
is,
like
you
know,
we're
not
gonna
like
show
you
every
detail
about
mini
cube,
but
we're
gonna
tell
you
when
you
should
be
using
it.
We're
gonna
tell
you,
you
know
why
you
should
be
using
in
those
cases
and
like
maybe
hear
some
like
tips
and
tricks
that
you
didn't
know
and
like
kind
of
same
thing
with
country
right.
E
And,
and
in
that
time
you
you
have
time
to
answer
questions,
there
are
specific
thrown
ins
and
you
have
time
to
tell
people
how
to
contribute
in
more
detail
and
then
us
doing
the
intro,
which
is
mostly
like
okay
check
out
this
github
page.
You
can
say
that
well,
this
is
like
the
flow
of
how
to
contribute,
if,
like
a
typical
flow
to
mean
a
Cuban,
want
the
kind.
B
F
A
I
think
that's
more
beneficial,
because
the
primary
focus
is
not
the
innards.
The
primary
focus
for
most
people
is
the
end
user
consumer
consumption
model
yeah.
We
are
t-minus
like
13
minutes.
We
kind
of
spent
a
lot
of
time
chatting
about
this
and
we
were
probably
have
to
federate
out
some
of
this
to
see
who
could
potentially
be
presenters
for
some
of
this
stuff
doing?
What
is
there
any
other
items
we
want
to
discuss
with
regards
to
Akane
you
and
sessions
I
just.
B
A
Okay,
any
last
minute
updates
one
two
three,
the
last
thing
I
wanted
to
do
today
was
just
to
give
you
a
brief
status
readout,
because
this
is
a
very
large
signal
and
maybe
just
do
a
quick
round
robin
even
if
you're,
not
part
of
the
sig
I
know,
a
lot
of
people
are
very
interested
in
in
kind
and
people
from
the
sig
or
working
on
kind.
So
if
we
can
get
a
brief
update
there,
that
would
be
useful
because
it's
it's
gonna
affect
some
of
the
other
things
that
we've
been
planning
as
p04
the
cycle.
A
B
I
just
felt
like
at
what
I
said
two
weeks
ago,
the
last
time
we
did
this
so
last
time
I
mentioned
we
did
a
few
triage,
so
we've
been
keeping
up
with
our
a
few
triage
and
we
are
now
trying
to
burn
down
the
set
of
issues
tied
for
the
v1
alpha
one
milestone.
It
is
slowly
sort
of
creeping
down
we're
trying
to
assign
outstanding
issues
and
get
things
resolved.
B
Vince
and
I
started
working
on
a
dock
for
proposing
some
somewhat
radical
restructuring
of
the
API
that
I
accidentally
shared
with
the
entire
States
mailing
list
when
I
when
I
shared
it
with
Vince.
So
you
probably
solving
for
that
go
by,
but
if
you're
interested
in
sort
of
where
we
think
that
the
API
might
go
in
the
future
and
sort
of
doing
some
refactoring,
please
go
ahead
and
check
that.
B
Gonna
be
discussing
it
in
more
detail
during
the
epoch,
happiness
I
think
we're
also
we're
also
seeing
sort
of
good
momentum
on
multiple
different
providers.
You
know
keeping
up
with
the
main
repo
and
sort
of
providing
a
baseline
for
us
to
sort
of
validate
that
the
things
are
building
actually
work.
So
it's
been
great
to
see.
A
I'll,
do
a
brief
update
for
convenien.
Pretty
much
114
is
kind
of
a
semi
quiescent
release.
We
did
a
lot
of.
We
did
a
hard
push
to
get
to
GA.
There
are
some
major
feature
enhancements
that
we
are
addressing,
one
of
which
is
the
H,
a
secret
story
that
we
have
had
ongoing
for
perpetuity
for
several
cycles
that
keps
been
approved.
We
are
going
to
try
and
execute
on
that
so
get
a
chai
out
of
experimental
and
kind
of
make
the
user
story
even
cleaner.
A
So,
like
we
have
a
user
story
for
AJ,
we
consider
it
experimental
it's
not
as
clean
as
we
would
hope.
We
are
turning
on
it
and
hopefully
it'll
be
cleaner
again
and
again
and
again
as
we
iterate
on
it.
The
next
one
is
like
we
did
could
be
a
minute
phases
where
we
put
a
lot
of
these
sub
commands
that
we
had
in
alpha
phases
into
an
it
for
how
the
entire
workflow
works.
Convenient
phases
is
being
worked
on
the
cycle.
There's
a
number
of
other
bug
fixes.
Anything.
A
That's
kind
of
major
feature
adds,
especially
that
would
be
API
affecting
have
been
punted.
We
do
not
plan
to
address
those
things
in
this
cycle,
otherwise
the
v1
14
milestone
has
been
very
curated
and
is
remarkably
stable
for
this
time
and
the
cycle,
usually
it
expands
wildly
and
then
then
I
have
to
boot
a
bunch
of
stuff
out,
but
we
are
we
it's
it's
remarkably
stable.
E
Well,
I
can
do
a
quick
one
for
the
component
standard
workgroup
we've
got
component
based
component
base
repo
emerged,
which
is
a
good
first
step.
We
have
a
case
like
proposal
up
for
Mike.
Pelton
is
finalizing
that
at
the
moment
and
making
it
cleaner,
so
we
can
publish
to
github
soon
I
kept
for
basically
an
enhanced
PFLAG
for
use
with
kubernetes
or
like
anything
that
has
the
specific
requirements,
basically
merging
flags
on
top
of
a
config
in
a
structured
manner,
when
I'm
gonna,
open
enhancement,
issues
for
component
base
and
all
the
components
configs.
E
But
it's
unclear
whether
any
components
configs
I
can't
be
graduated.
This
cycle
either.
Yes,
well,
whatever
like
it,
take
that
that
requires
component
owners
to
actually
go
through
their
IP
ison,
see
that's
things
like
ok,
which
regards
to
that
structure.
But
the
structuring
restructuring
component
bases
preceding.
E
D
I
can
give
a
quick
cops
up.
They'd
have
think
the
the
most
interesting
thing,
I
think
that's
happening
in
cups
with
broader
relevance
is
we
are
finally
making
our
move
to
Etsy
d3
and
making
that
mandatory
yay,
but
as
part
of
that,
we're
actually
making
a
TD
secure
by
sort
of
secure
by
default.
I'm
I
think
we'll
make
it
impossible
to
opt
out
of,
and
then
we're
also
stopping
we're
trying
to
stop
external
add-ons
from
using
the
same
ED,
CD.
So
calico
and
psyllium.
D
A
A
D
E
D
And
calico
has
done
their
can.
What
I
call
it
tyfa
I
think
to
address
some
of
this,
but
I'm
sure
that
we
will
find
more
such
issues.
Thank
you
for
calling
pointing
that
out,
but
I
thought
that
was
the
most
that's
sort
of
I
think
the
most
interesting
things
this
for
this
Seguin
yeah,
hopefully
we're
going
to
like
cluster
it
go,
and
things
like
that
soon
that
that's
their
this
is
that
this
week's
update.
F
L
H
A
question
on
that
you
can
support
multi-master
in
kind.
We
do
wheat.
Thank
you.
Did
that
documented
I'm.
F
B
Mini
cube,
so
we
picked
up
a
couple
of
new,
maintain,
errs
and
documented
our
issue
triage
process
based
on
the
kubernetes
community
processes.
We
pushed
out
a
new
bug,
fix
release
and
will
follow
up
with
another
one
next
week,
but,
more
importantly,
we
defined
which
issues
we
will
resolve
for
1.0
1.0
is
focused
around
documentation
and
debug
ability,
and
we
also
released
our
2019
roadmap
and
our
priority
is
post
1.0
hour
localization
in
the
ease
of
installation.
E
E
E
E
On
the
setup
ducks
tab,
there's
a
lot
of
a
lot
of
really
old,
getting
started
guides
like
things
that
existed
in
serie
preventive,
zero,
not
something
they
are
still
there,
and
also
the
getting
started
from
scratch
thing
which
has
been
updated
in
like
years,
yeah,
etcetera,
etcetera,
etcetera
and
we
were
actually
kind
of
like
we
have
some
coverage
and
our
sub
project,
but
not
as
well
as
would
like
to
so
so
for
a
new
user.
Getting
into
this
page,
they
have
actually
no
idea,
and
then
they
have
no
like
wow.