►
From YouTube: Kubernetes SIG CLI 20200311
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
A
When
it
comes
to
announcement,
there's
only
one
due
to
corona
verse
spread
across
the
globe
cube
con
tu
was
postponed.
The
last
update
is
from
March
4th
from
what
I
checked
earlier
today.
It
is
proposed,
that'll,
be
held
somewhere
in
July
August,
but
no
day
has
been
posted,
as
of
yet
so
you'll
have
to
follow
the
news
on
that
topic.
A
Koukin
China
is
entirely
canceled,
but
you
called
North
America
its
plan
to
take
place
as
initially
planned
in
November.
If
I
remember
correctly,
that's
pretty
much
it
for
the
announcements,
maybe
one
another
important
thing
is
we're
in
the
code:
freeze
mode,
which
means
only
critical
bug
fixes
for
118,
which
are
stabilizing
wine
and
118
release
or
are
fixing
the
tests.
So
if
you
got
pinged
about
fixing,
some
tests
make
sure
to
handle
those
with
as
your
top
priority
and
other
than
that,
I
think
we
can
move
on
to
the
main
topics.
A
C
A
A
B
A
B
B
C
So
a
substitution
replacement
kind
of
solution
on
top
of
everything
else
that
customized
does
and
it
was
well-received
and
it's
it's
github
issue,
and
there
was
an
initial
implementation
of
it.
That
was
done
that
we
played
with
a
little
bit
and
but
it
was,
it
was
done
as
a
as
an
example
plug-in,
and
so
it
was
put
into
the
example
plugins
directory,
and
it
clearly
has
some
you
know
boilerplate
organization
kind
of
stuff
in
there
like
geared
toward
being
an
example.
So
I
just
wanted
to
see.
C
If
anyone
here
is
aware
of
what
the
the
plans
are
for
that
replacement
transformer
either
in
terms
of,
are
there
plans
to
take
it
forward
as
a
as
a
built
in
plugin
for
customized,
instead
of
just
an
example,
or
will
that
will
that
example
be
maintained
as
something
with
with
a
stable
interface?
Or
is
the
intention
just
for
for
operators
to
sort
of
clone
down
that
example
and
roll
their
own
I?.
B
Can
hazard
a
guess,
you'd
want
Jeff
on
the
call
he
would
be
able
to
give
a
more
canonical
answer
because
he
typically
drives
a
lot
of
the
customized
API
pieces,
but
I've
been
talking
with
him
quite
a
bit
lately
about
this
sort
of
thing
and
I
can
definitely
answer
the
question
after
this
one,
the
replacement
transformer
a
little
less
so,
but
what
I
my
you
know,
given
that
I'd
say
the
what
there
was
a
sort
of
class
of
problems
around
bars
and
substitutions
and
replacement
transformer
seems
like
it
kind
of
falls
into
that
category.
B
B
This
this
this
particular
use
case,
has
been
asked
for
a
lot
I.
Think
being
able
to
do
substitutions
is
something
that
definitely
a
lot
of
folks
are
asking
for.
It
was
originally
out
of
scope
of
customized
I
think
we
are
a
pretty
clear
upfront,
or
at
least
there's
a
number
of
documents,
I
think
where
we
say
that
that
is
not
within
within
the
original
kind
of
in
scope,
I
guess
I'd
say,
although
I
don't
think
it's
like
a
hard
scope,
but
it
definitely
wasn't
what
we
were
truly
original
II
trying
to
solve.
B
That
said,
it
seems
like
it's
part
of
the
problem,
and
one
of
the
things
Jeff
and
I
have
been
going
back
and
forth
on
is,
as
we
see
more
folks,
developing
their
whole
system
around
customizing
kind
of.
So
this
is
a
piece
where
it's
like
your
your
customized
is
becoming
kind
of
like
your
system
or
a
key
part
of
your
system
to
do
deployment.
I.
B
Setters
were
actually
also
targeted
in
a
different
use
case.
Then
this
particular
one
but
I
think
it's
an
example
of
like
a
different
philosophy.
Right
which
would
be,
should
another
program
be
replacing
things
in
customization
files
ahead
of
time
and
should
customize
facilitate
people
in
there
I
don't
know.
The
answer
is
any
of
that
context.
Helpful
yeah.
C
Yeah
yeah,
it
is
very
much
and
appreciate
it
and
you
totally
I
totally
understand
and
don't
disagree
with
it
all
the
the
idea
of
having
you
know
wanting
to
have
a
limited
philosophy
and
so
I.
If
this
isn't,
if
this
kind
of
thing,
if
the
decision
is
made
for
it
not
for
customized,
not
to
be
a
good
home
for
it,
then
then
that's
fine.
C
Then
then
we
would
need
to
you
know,
create
something
you
know
as
a
standalone
kind
of
tool,
but,
like
you
said,
is
something
that
a
lot
of
people
need,
so
it
it
would
still
probably
make
sense
for
it
to
be
it's
a
standalone
tool
or
something
that
that
many
people
could
use
just
whether
or
not
it
lives.
Whether
or
not
it's
in
scope
for
for
customized,
proper
I
think
would
would
be
a
different
question
right
totally.
B
B
B
What
people
have
asked
for
what
problems
people
are
trying
to
solve
it
with
it
and
then
figuring
out
like
what
is
the
right
way
to
evolve
customized
in
order
to
address
all
those
issues
that
maybe
weren't
originally
in
scope,
right
and
which
philosophy,
and
so
maybe
we
could
have
a
it-
might
make
sense
to
schedule
like
an
open
discussion
on
that
yeah.
C
That
customized
focuses
on
as
well
as
substitution
and
replacement
kind
of
tasks
and
we're
working
on
the
the
next
version
of
airship,
which
we
want
to
be
much
smaller
and
which
we
want
to
leverage
a
lot
of
the
sort
of
common
community
projects
that
have
grown
up
in
the
meantime.
And
so
we've
a
we've.
C
C
As
a
tree,
even
the
hardest
thing
that
the
least
commonly
implemented
things
seems
to
be
the
replacement
and
substitution,
which
is
important
for
our
use
case,
at
least
as
well,
so
so
yeah
we're
we're
happy
to
collaborate
on
whatever
the
right
solution
is.
But
since
we
already
have
our
own
granite
open-source,
but
you
know
maintained
by
us
solution,
we're
hoping
that
we
can.
We
can
get
a
little
bit
more.
You
know
buddies
involved
in
in
developing
and
maintaining
this
thing
with.
C
I
think
so
I
think
so,
and
we're
also
not
blocked
we're
gonna
at
least
start
with
our
own
replacement
transformer
solution
and
and
probably
probably
hack
a
few
more
features
into
it
at
least
to
get
us
started,
but
I.
You
know
I'm
we're
not
necessarily
going
to
try
to
make
it
a
a
long
term
or
prod
grade
kind
of
solution
yet,
but
just
something
to
help
us
keep
getting
movement
going
forward,
but
but
yeah.
So
we
can
keep
working
on
that,
and
so
we
all
kind
of
agree
what
the
the
best
path
forward
is.
A
A
See
any
objection,
okay
wing
this
one
I
was
at
the
scheduled
for
118,
there's
no
schedule
for
119
and,
and
basically
the
118
schedule
ends
pretty
much
around
that
date.
If
I
remember
correctly,
26
Thursday
is
that
they
are
planning
to
release
one
118,
which
basically
means
we
would
not
start
running
119
sooner
than
that
date
or
close
to
that
date
or
which
basically
means
at
the
beginning
of
April,
we'll
be
doing
119
planning
for
6
e,
Li
and
I'm.
A
A
C
A
B
Touch
on
the
quiz
I
can
quickly
answer
the
functions
piece.
Yes,
they
serve
similar
roles.
There
is
a
plan
to
unify
these
things.
I,
don't
think,
there's
any
urgency
around
it.
It's
mostly
going
to
be
around
formats
functions.
Some
background
functions
are
effectively
inspired
by
customized
plugins,
but
also
inspired
by
medic
controller
and
where
functions
come
from
is
a
decoupling
of
the
logic
to
trance
apply
a
set
of
operations
to
a
collection
of
resources
from
the
particular
environment
that
that
happens
to
be
needed
so
like,
for
example,
in
customize.
B
However,
in
meta
controller,
meta
controller
is
the
orchestrator
and
the
source
is
not
a
customized
resource
target,
but
instead
is
the
API
server
right
and
they're
written
back
to
the
API
server,
whereas
customized
plugins
are
written
to
standard
out
right,
but
in
both
cases
the
architecture
is
the
same
right,
functional
here's,
a
list
of
resources
that
need
to
have
be
evaluated
and
either
deleted
or
added
or
modified
in
some
way.
Here's
the
configuration
that
defines
how
to
do
those
operations
and
then
write
them
out
to
some
place,
so
functions
are
there's
two
pieces
right
now.
B
There's
like
a
spec,
a
function
spec,
which
is:
how
can
we
define
a
sort
of
input
output
that
one
could
begin
to
write
a
program
in
in
that
we
could
then
drop
into
different
environments
VI.
The
vision
is,
this
would
be
customized.
This
would
also
be
meta
control,
or
this
would
also
be
qu
builder
right.
It
would
be
insert
your
system
here
and
then
drop
in
the
function
and
it
runs
CI
CD
systems
tacked
on
whatever
jenkins.
B
The
other
part
is
the
runtime
right
now.
The
only
supported
runtime
for
functions
is
containers.
We
started
out
with
this
one
because
has
a
lot
of
lot
of
nice
properties.
It
supports
jinjo
supports
Telmex,
panning
sports
coupe
val
supports,
essentially
whatever
system
you're
using
today,
you
can
implement
as
a
function
means
they're
a
little
harder
to
write,
but
a
lot
of
flexibility.
It
means
that
when
they're
run
locally,
they
can
be
run
sandboxed.
So
we
can
disable
the
network.
We
can
make
sure
you
don't
have
write
access
to
disk.
B
We
can
make
sure
that
you
don't
have
the
ability
to
escalate
privileges
and
that
you
run
is
nobody
user
and
now
your
function.
Now
you
can
download
a
function
and
run
it,
and
it's
not
like
downloading
a
binary
and
running
it
right.
It's
a
lot
more
strict
than
that
containers
support
a
distribution
mechanism
which
is
kind
of
like
docker.
Well,
I
might
fetch
it
for
you.
The
support,
versioning
via
tags,
there's
a
lot
of
nice
properties
about
containers
as
a
publishing
mechanism
for
these
sorts
of
things,
other
considering
other
runtimes.
B
You
could
consider
a
endpoint
like
it
should
be
endpoint
that
you
send
the
resources
to
and
then
get
the
resource
back.
That's
what
meta
controller
does
so
that's
what
functions
are?
Yes,
they
are
basically
trying
to
standardize
the
interface
for
plugins
and
then
also
enable
plugins
to
be
used
in
other
environments.
So
if
you
wrote
a
Kumail
plug-in
to
do
validation,
you
could
use
it
in
customize.
You
can
also
use
it
in
a
CI
CD
pipeline
without
customize
these
sorts
of
things
yeah.
C
And
let
that
makes
perfect
sense
and
appreciate
the
the
background
there
too,
and
and
I'm
definitely
sold
on
the
benefits
of
the
container
based
approach.
So
would
you
would
you
sort
of
foresee
long
term
and
I
know
this
probably
delves
into
the
the
are
the
architectural
discussion
you
want
to
have
next
week.
C
B
The
vision,
there's
no
timeline,
okay,
you
can
hesitant
to
come
out
and
say
it
because
if
we
said
it,
then
people
would
start
asking
for
timeline
and
didn't
that
it's
like
eminent
right,
we
expect
function.
Functions
are
much
less
mature.
We
expect
them
to
evolve
a
bit.
We
have
to
figure
out
when
the
right
time
of
integrating
them
into
customized
would
be,
and
in
making
sure
that
we're
happy
with
how
they
are,
and
then
we
also
need
to
figure
out.
Are
we
get
how
we're
gonna
support
existing
customized,
plugins
I'd?
B
Imagine
we
want
to
support
both,
which
you
know
like
we
wouldn't
want
to
deprecated.
We
want
to
make
sure
folks
who
are
writing
plug-ins
today.
Those
continue
to
work
then
may
be
considering
a
migration
path
like
with
a
dock
of
like
okay.
Here's,
if
you
wrote
your
customized
function,
just
wrap
it
in
this
binary
executive,
this
binary
and
then
now
it's
converted
to
the
new
format.
Yeah,
there's
a
lot
of
stuff
to
be
thought
out.
B
B
C
E
A
Yeah,
ok,
does
anyone
have
any
bigger
topics
that
I
want
to
bring
up
to
the
attention
of
the
entire
group?
Okay,
hearing,
none!
Let's
move
on
to
the
send
ups!
In
that
case,
I
see
that
crew
has
no
updates
and
I'm
guessing
that
either
Cornelius,
Oh,
AK
or
Ahmed
put
it
here,
but
I'm
not
seeing
any
of
them
on
the
call
so
cuy,
because
I
see
that
there
are
some
good
news
from
NIC,
so
go
ahead.
Yeah.
D
Thanks
all
to
y'all
we've
merged
the
initial
kit
for
GUI
we
got
a
small
bump
in
stars.
They
grew
up
till
1100
now
after
that
I,
so
we're
making
pretty
good
progress.
I
think
we're
we're
in
pretty
good
shape.
Now
we
have
a
pretty
solid
react
based
component
foundation,
we're
going
to
work
over
the
next
couple
of
days,
if
not
flesh
out
the
document
so
that
we
can
get
additional
developers
to
easily
contribute
new
components
or
new
commands
will
flush
out
the
documentation
in
all
respects
there.
D
D
So
if
I
disassembling
the
election,
executable
cuckoo,
foo
and
then
I
say
cocoa
tofu,
the
electron
window
opens
up,
but
oddly
it
doesn't
have
a
dock
icon,
Mac
OS,
it
doesn't
receive
keyboard
input,
but
as
when
I
launched
my
launch
the
very
same
executable
just
from
bash,
everything
is
fine.
So
I
was
wondering
whether
the
coo
cuddle
authors
had
any
insight
into.
Are
you
spawning
the
plugins
in
any
particular
way
that
might
think
might
lead
to
there.
A
Is
a
I
can
point
you
to
a
particular
line
and
the
code,
but
we
are
invoking
at
least
on
UNIX
based
system
were
invoking
exactly
which
creates,
which
replaces
the
current
binary
with
the
binary
from
from
from
from
cuba
caudal.
So
in
my
big
culprit,
maybe
we're
missing
some
settings.
I
said
that
it
works
properly
with
with
cuy
agree.
A
D
B
D
That's
mostly
working
on
the
docks
and
the
finish
couple
aspects
of
the
plugin.
The
hope
is
that,
once
if
we
can
iron
out
this
exact
verses,
torque
issue,
then
the
coop
cut
will
plug
in
for
Kui
will
be
the
electronics
cutable
itself
without
any
modification,
so
there
won't
be
any
need
for
additional
rapper
scripts
or
whatever.
There
must
be
the
binary
Punk
in
place.
I
know,
there's
still
the
outstanding
issue
with
how
we,
how
we
get
the
coop,
how
we
get
that
via
crew,
via
with
all
the
Similan
case,
use
off.
B
A
D
A
D
Actually
one
one
thing
that
we've
been
working
on:
maybe
we
can
demo
it
next
time
we've
been
able
to
do
a
pretty
good
job,
I
think
it's
pretty
nice
integration
of
the
KU
cuddle
command
line,
help
into
a
graphical
form.
So
when
you're
using
Kui,
you
like
see
the
cuckoo
clock
man
because
it
could
get
that
shakes.
B
Are
you
aware
of
the
Li?
You
tell
us
what's
being
done?
Oh
so,
there's
the
next
generation
set
to
control
libraries
in
commands
being
developed
in
CLI
utils,
and
so
this
has
stopped,
but
I
think
we'll
relapsed
a
bit
with
information.
You'd
want
so,
for
instance,
one
of
the
things
being
developed
is
a
rich
notion
of
status.
Yeah
like
propagating
set
us
up
from
a
pod.
If
you
create
a
deployment
like
knowing
one
that
deployment
isn't
ready
and
two
it's,
because
the
image
can't
be
polled
or
something
like
that.
B
Those
seems
like
things
that
would
you'd
probably
want,
because
I
know
you
do
some
walking
of
structure
anyway
yep.
So
you
those
tool,
you
probably
you're
gonna.
We
probably
want
to
think
about
what
role
those
tools
might
be,
because
I
can
imagine
you
wanting
to
delegate
to
some
of
that
logic.
Absolutely.
D
Yeah
another
thing
that
came
to
mind
over
the
past
couple
of
weeks
is
this
notion
of
tapping
in
I
know
that
there's
a
danger
issue
here,
but
tapping
into
the
event
stream
flowing
out
of
the
API
server
and
and
so
that
people
have
a
more
live
notion
of.
What's
going
on,
you
know
from
the
set
point
of
of
a
learning
or
a
general
situational
awareness,
so
maybe
next
time
we
can
also
discuss
the
danger
issues.
They're
like
at
what
level
are
we
tapping
into
too
much
information
to
the
fire
hose?
D
How
can
we
get
just
some
more
situational
awareness
with
a
much
on
app?
We
can
put
things,
for
example,
in
the
tray
menu
you
thought
in
Windows
or
Mac
or
Linux
just
to
get
some
general
sense
of
what's
happening
behind
the
scenes.
So
if
you
guys
have
any
insight
there
as
well
that'd
be
great.
That's.
B
D
D
Yeah,
so
one
of
the
things
that
now
that
we're
a
good
react
citizen,
we
can
very
easily
pull
in
any
form
based
componentry.
That
makes
sense
so
we're
thinking
about
for
whether
it's
via
customize
or
just
generally
presenting
even
as
read-only
views
being
able
to
summarize,
for
example,
a
resource
in
a
more
form
based
approach
and
I.
Think
this
may
require
some
user
studies.
I
think
that
some
people,
you
know
just
appreciate
information
in
different
forums
and
people
just
like
to
see
the
raw
information.
D
Some
people
will
be
frustrated
by
forums
and
some
people
need
the
forms
to
make
any
progress.
So
I
think
it
require
a
bit
of
a
bit
of
consideration,
but
yeah
forms
is
definitely
on
the
short
list,
especially
anything
in
light
of
all
the
great
work
guys
doing
with
customized
I.
Think
that's
mentioning
overlaps
there
threading
information
through
the
lifecycle,
some
customized
back
and
forwards,
but
then
also
being
able
to
reflect
that
distinction
between
the
template
and
the
values.
I
think
the
forms
might
fit
there
nicely.
Yeah.
B
We're
starting
to
think
about
how
we
can
attach
more
of
that
information
to
the
reasonable
selves
with
customized
like
well.
One
thing
with
customized:
now
is
it's
a
black
box
and
so
like
you
can't
spit
out
customized
like
you
can't
spit
out
of
customize
something
that
then
some
other
tool.
You
didn't
understand
how
customized
structured
it
that.
C
D
A
E
Yeah,
okay,
so
I'm
I'm
y'all
from
China,
and
this
is
first
time
I
attend
the
meeting,
alright
and
and
so
on.
I
submitted
several
pr's
to
control
and
now
there's
two
hanging:
PR
QPR's,
which
related
to
you
could
control,
apply
and
cope
control
expose
so
there's
generally.
The
first
question
is,
we
know,
keep
control
apply,
don't
support
an
object
with
generates
name
right.
E
B
Well,
the
thing
with
generate
name
essentially
means
create
a
new
instance.
Each
spawn
yeah
and
apply
is
intended
to
not
do
that.
Like
imagine,
the
situation
where
apply
applies
to
be
able
to
run
continuously
in
a
loop
and
if
there's
no
changes
necessary,
it
shouldn't
do
anything
right.
If
you
have
generate
aim,
what
does
it
do?
Does
it
if
you
just
run
it
in
a
loop?
B
Does
it
just
like
hot
loop,
creating
objects,
or
does
it
not
create
object
the
first
time
and
then
try
and
figure
out
which
one
it
created,
but
not
generate
names
after
that,
like
it's
not
really
clear,
what
generate
name
should
do
and
that
will
apply
in
that
case,
create
a
new
object
or
use
the
same
one.
No
other
one
seems
to
make
much
sense.
E
B
A
E
Yeah,
so
it's
just
a
little
bit
strange
when
we
use
the
encounter
problem
like
this,
because
they
thought
create
and
it
applied
generate,
do
the
same
sign
but
actually
apply
Kong
to
generate
name
where
it
can
so
I
just
assembly,
Pete
I
submit
a
PR
about
this
and
I.
Don't
know
if
this
works,
because
I
in
that
PR
give
the
information
that
apply.
Can
support
generate
name
to
give
this
information
directly
to
the
user.
E
Yes,
the
PR
just
updates
their
message,
yeah
and
yeah,
and
the
second
is
about
coop
contra
exposed
so
coop
contra
exposed
and
if
that's
designed
to
not
support
the
external
name,
because
we
know
external
name
will
not
rely
on
the
selectors,
but
the
coops
CTO
or
could
control
expose
will
try
to
find
that
the
selectors
from
the
object
so
actually
could
conquer
contra
exposed
thousands
what's
the
external
name,
but
it's
not
retaining
a
dog.
Oh
it's
not
to
retain
the
help.
Information.
B
That's
that's
possible,
I'd,
say
generally.
The
direction
that
we
are
headed
is
oh
well
definitely
away
from
like
to
control
run.
You
could
control
riding
to
control,
expose
or
probably
sibling
commands
and
they're.
Both
they
both
were
introduced
early
on
in
the
project
to
give
folks
a
sort
of
docker
like
very
familiar
experience.
B
Neither
one
is
really
intended
for
people
to
use
beyond
just
learning
about
the
system
and
the
way
we
recommend
people
using
the
system
beyond
like
if
they're
actually
running
workloads
for
real
is
to
use,
apply
and
declarative
configuration
or
some
approach
like
that,
rather
than
directly
editing
objects
in
the
cluster.
That
said,
I
think
if
you
needed
something
like
this
there's
great
service
as
well
or
there
should
be,
which
would
also
be
very
limited
and
probably
not
have
a
lot
of
options.
So
here's
my
question
is:
why
are
using
to
control
expose.
E
Because,
in
the
help
information
the
coop
contra
expose
state
like
and
class
type
II,
like
external
name
and
other,
are
supported.
Its
is
written
in
helping
for
so
some
user
might
use,
and
he
definitely
use
this
country
expose
external
name
and
found
no
matter
how
you
do
that.
You
will
never
create
external
name
by
cube
if
control
expose
it's
impossible.
Oh
so.
B
E
B
B
So
it's
really
great
that
you're
bringing
them
here
and
I
think
on
both
the
answer
is
that
these
are
aren't
areas
that
we
need
to
support,
but
we
do
need
to
explain
why
they're
not
supported,
and
so,
if
you
wouldn't
mind,
if
you
said
you
filed
some
PRS,
if
you
wouldn't
mind
linking
those
in
the
dock
just
so
that
they're
easily
accessible
and
don't
get
lost
in
the
noise,
that
would
be
really
helpful.
Yeah.