►
From YouTube: Kubernetes SIG CLI 20210505
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
As
my
two
co-hosts
for
six
eli,
our
agenda
is
pretty
packed
so
I'll
just
jump
right
into
I
put
in
the
announcements.
122
release
dates
the
one
important
day
that
is
very
crucial
for
everyone
is
the
enhancement
freeze,
which
is
next
thursday,
which
is
may
13th,
so
any
new
features
updates
changes
from
alpha
to
beta
and
whatnot
those
has
to
land
those
have
to
land
by
that
the
next
thursday.
A
If
you
need
help
pushing
some
of
the
enhancement
through
reach
out
to
any
of
the
leads,
which
will
be
myself,
sean,
eddy
and
enfield,
we
will
help
you
to
go
through
those
by
next
thursday
and
we
will
be
talking
about
enhancements
in
a
while
as
well.
A
That's
one
thing:
the
other
thing
is
something
that
I
sent
last
week
to
6
july,
but
I
haven't
put
it
here
I'll
I'll
I'll.
Do
that,
after
the
call
is
we've
changed
slightly
a
bit
how
we
are
going
to
organize
and
cancel
the
meetings?
A
We
would
like
to
see
agenda
being
pre-populated
24
hours
prior
to
the
call,
and
if
the
agenda
for
the
next
call
is
empty,
24
hours
prior
to
call
which
it
will
be
usually
around
6
p.m.
Central
european
time
noon
for
us,
eastern
and
9
00
a.m
for
west
west
coast.
If,
by
that
time,
the
agenda
is
empty,
we
will
be
cancelling
the
call.
This
will
help
all
of
us
to
better
prepare
review
the
topics
if
there
are
something
filled
in
perfectly
and
cancel.
A
B
A
Okay,
cool
with
that,
I
see
new
faces
in
the
goal.
So
now
is
your
time
to
introduce
yourself
tell
a
little
bit
about
yourself.
What
do
you
do?
What
do
you
expect
from
sexy
li.
D
Hi,
oh
sorry,
did
I
speak
over
someone
no
go
ahead.
Go
ahead,
yeah
hi,
my
name
is
janis.
I'm
a
software
engineer
at
arito
and
I
was
the
release
manager
for
kubeflow.
1.3
qflo
is
a
small
project
about
kubernetes
and
machine
learning,
and
I'm
hoping
to
tell
you
a
bit
about
how
we
use
customize
some
of
the
problems
we
bump
into,
and
I
hope
that
we
can
work
more
closely
with
upstream
customize
provide
the
use
case,
provide
even
contributions
and
you
know
have
everyone
benefit
from
that.
A
Awesome
we're
always
open
for
any
feedback,
good
or
bad
about
any
of
our
self
projects.
E
Yeah
yeah
go
ahead
and
slate
it
yeah.
So
I'm
one
of
the
release
notes,
shadow
for
1.22,
and
so
I
was
looking
to
find
out
other
more
stuff
in
in
the
kubernetes
ecosystem
to
contribute
to,
and
I
bumped
across
six
cli,
and
I
was
look
like
it
caught
my
eye
like
maybe
I
can
make
some
meaningful
contributions
here,
so
yeah,
I'm
looking
for
something
you
know
big
enough
friendly,
because
I'm
not
very
experienced
in
the
entire
ecosystem
and
kubernetes
requires
a
lot
of
context
so
hoping
to
start
small.
E
So
let
me
know
if
there
are
any
beginner
issues
or
any
kept
that
require.
You
know
like
the
good
first
issues
that
I
can
get
hold
of.
I'd
love
to.
You
know,
look
around
actually
so
yeah
cheers.
A
Right,
that's
that's
something
that
I
completely
forgot.
So
we
have
two
enhancements
liaisons
if
I
said
it
properly,
which
will
be
basically
two
people.
A
A
Okay,
slide
in
cool,
so
island
is
one
of
them.
Then
he
will
help
us
with
gathering
the
enhancements
and
pushing
them
through
to
the
secret
release,
because
starting
from
121
sick
release
is
not
reaching
out
to
six
for
enhancements,
but
it's
more
about.
We
need
to
explicitly
tell
them
that
this
is
something
that
we
will
be
working
on.
A
There
is
a
a
document
where
we
need
to
put
the
enhancements,
so
we
will
be
working
closely
with
saturn
and
the
other
person.
I
remember.
G
I
am
hi,
I'm
vega,
I
I
we
are
talk
before
in
the
slack
like
slayden,
I,
we
attended
the
enhancement,
kip's
improvement
and
habit.
A
A
I
think
I
recognize
the
rest
of
the
name.
Okay,
with
that
we
can
move
on
to
the
main
topics
janes.
I
assume
you
want
to
be
presenting
I'll
make
a
co-host
and
I'll
stop
sharing.
H
D
C
Know
how
do
you
mean?
Well,
you
say
you're
going
to
provide
some
feedback
on
customize.
Should
we
be
writing
this?
I'm
taking.
D
Oh
yeah,
absolutely
so
this
is,
I
will
provide,
be
providing
the
slides
as
well.
Everything
that
I
will
show
I
will
make
available
right
after
the
meeting,
and
I
think
the
recording
will
also
be
available
right.
D
Okay,
so
can
you
see
my
screen.
A
Kip
slow
cube
flows
use
up.
D
D
So
let
me
get
started
so.
First
of
all,
I
want
to
talk
a
bit
about
keyflow.
What
keepflow
is
in
one
sentence:
keepflow
is
a
muslim
platform
that
is
built
on
top
of
kubernetes
and
because
of
that,
we
try
to
keep
everything
simple,
portable
and
scalable.
D
I'm
trying
to
give
a
small
idea
of
the
problem.
We're
trying
to
solve.
You've,
probably
seen
this
slide
a
million
times
before,
but
I'm
going
to
say
it
again.
The
the
idea
behind
the
slide
is
that
many
people
think
most
of
the
time
in
machine
learning.
Projects
is
spent
on
writing
ml
code
and
models
when,
in
reality,
most
of
the
time
and
problems
are
in
the
infrastructure
that
support
and
surround
the
machine
learning
processes.
D
And
finally,
kubeflow
is
an
open
source
project.
We
have
a
lot
of
contributors
and
we
would
love
to
welcome
new
contributors
in
this
area.
D
1.3
is
what
we
released
right
now
we
had
the
custom
some
custom
tooling,
which
we
called
kfctl,
which
was
very
painful
to
maintain
and
in
general
we
didn't
want.
D
We
wanted
to
integrate
with
standard
tools
like
customize,
so
in
1.3
we
try
to
move
everything
to
plain
customize
and
use
only
that
for
our
manifest
and
we've
been
pretty
successful
with
that
right
now,
a
user
can
deploy
just
with
customizing
cube
ctl
and
we've
seen
that
people
can
also
integrate
much
easily
much
more
easily
with
other
systems
like
cargo,
when
we
use
just
customize,
which
is
pretty
standard
now,
while
working
with
so
many
customizations
we've
bumped
into
a
number
of
issues
that
we
feel
are
probably
issues
for
other
projects
as
well.
D
So
I
wanted
to
make
a
slide
and
list
them
all
here
just
to
let
you
know
that
you
know
these.
These
are
things
that
are
bugging
us
and
would
love
you
to
work
together
with
you
solve
them.
So
that's
another
thing.
D
You
know
this
use
case
doesn't
work
as
we
expected
to
or
something
similar
and
then
also
be
able
to
contribute
to
devote
some
time
to
contribute,
as
well
like
we'd
like
to
develop
this
type
of
relationship
and
trust
slow
with
time.
Of
course,
this
doesn't
happen
overnight.
D
Then,
of
course
we
have
a
small
problem
with
vars
in
customized,
which
I
know
there
are
a
lot
of
upstream
efforts
to
remove,
and
then
in
the
new
customize
there
is
a
library
that
doesn't
support
utf-8
text
in
comments
which
breaks
basically
anyone
who
uses
utf-com
a
text
and
comments
which
are
mostly
people
from
non-english
speaking
languages,
for
example.
I
know
we
had
some.
D
D
D
The
patching
becomes
pretty
weird,
I
believe
we
haven't
opened
an
upstream
issue
yet,
but
this
is
what
we're
going
to
do
very
soon.
So
again,
I
don't
want
to
get
into
the
technical
detail.
It's
just
a
quick
rundown
of
things
we've
bumped
into
and
of
course,
for
anything
that
does
not
have
an
upstream
issue
right
now.
We
will
open
one
very
soon.
D
And
finally,
to
wrap
up.
I
want
this
very,
very
quick.
I
think
that
quickly
is
probably
one
of
the
biggest
open
source
adopters
of
customize.
I
know
big
open
source
projects,
other
big
open
source
projects
have
adopted
customers
as
well,
but
I
think
we
use
it
a
lot.
I
don't,
I
don't
think
anyone
else
has
this
many
customizations
in
one
project,
but
for
moving
on.
D
J
To
do
okay,
all
right,
great,
hey,
first
off!
Thank
you
so
much!
Those
are
that's
a
great
summary,
I'm
familiar
with
all
those
issues
and,
as
you
know,
we've
been
chatting
especially
on
issue
three,
seven,
nine
four,
which
is
the
ordering
issue
and
we've
got
while
we've
got
sean
and
matcha
here.
This
would
be
a
good
time
to
talk
about
the
ordering
issue,
so
I
can.
J
I
can
set
them
up
this
way
and
when
customized
started
a
couple
years
ago,
there
was
some
desire
to
have
customized
do
some
ordering,
so
it
did
do
some
ordering.
So
we
would
put
some
of
the
cluster
scoped
resources
first,
particularly
namespace,
because
if
you
apply
a
bunch
of
resources
and
some
of
them
are
in
a
namespace
and
that
namespace
doesn't
exist,
you're
going
to
have
an
error,
so
customize
did
some
sort
of
rudimentary
sortering
at
the
beginning.
J
At
some
point,
we
thought
a
better
approach
would
be
to
just
honor
the
ordering
that
people
specified
in
their
customization
files
and
not
necessarily
do
the
sort.
So
now
customize
can
do
two
things
depending
on
a
flag
it
can
by
default.
It's
going
to
continue
doing
this
legacy
sorting
and
with
the
flag.
It
will
not
do
the
sort
you'll
just
respect
whatever
order
you
specify
in
your
customization
files
and
doing
a
depth
first
search.
J
J
They're
supposed
to
be
able
to
handle,
you
know
in
an
ideal
sense,
arbitrary
ordering
now
I
think
we
do
want
to
do
cluster
scope
first
and
and
then
name,
spaced,
scoped,
second
and
a
user
can
of
course
order
things
that
way
in
their
customization
files.
But
I
think
you've
got
a
problem
now.
You've
got
like
400
files
and-
and
you
don't
necessarily
want
to
go
back
and
edit
all
those.
J
So
the
here's,
the
here's,
the
thing
the
question
I
have
for
like
manche
and
sean
and
for
discussion
is
like
how
much
should
clients
worry
about
how
they
order
things
in
and
apply.
C
So
I
have
a
couple
of
points
so
so
the
general
way
that
this
so
up
until
now,
there
has
been
really
no
tool
to
order
resources
and
wait
for
resources
to
come
up
before
another
resource
is
applied
and
generally
what
we
do
is
you
apply.
Resources
say,
like
you
had
a
crd
and
then
you
have
an
instance
of
that
custom
resource
in
the
same
apply
it's
a
lot
all
of
most
of
them,
or
many
of
them
are
going
to
fail.
And
then
you
apply
it
again.
C
It's
kind
of
like
an
item,
potent
principle
where
you
just
keep
running
it
multiple
times,
and
then
it
eventually
works
which
is
not
which
doesn't
work
for
for
everybody
and
which
is
why
I
actually
have
a
branch
in
the
cli
utils,
where
I've
created
a
a
an
ordering
and
a
basically
it
depends
on
annotation
so
that
you
could
say
that
one
particular
resource
depends
on
another
resource
and
it
will
apply
them.
C
It
it'll
create
a
graph,
it'll,
do
a
topological
sort
and
it
will
order
them
in
the
correct
it'll
sort
them
in
the
correct
order
in
order
to
apply
them
in
the
correct
order.
C
You
have
to
wait
until
it's
established
before
you
can
apply
a
custom
resource
of
that
definition,
and
and
so
actually,
in
this
branch
that
I
have
in
cli
utils,
that
I've
been
experimenting
with
it.
It
has
certain
tasks,
including
a
weight
task.
It'll
wait
on
a
particular
status
for
this
resource
in
the
cluster
until
it
tries
to
apply
the
next
set
of
resources,
and
so
so
there
are
efforts,
actually
substantial
efforts
to
to.
Actually
you
know,
experiment
with
this,
and
you
know
to
to
approach
this
problem.
C
It's
it's
not
just
ordering
it's
ordering
and
then
waiting
on
a
particular
status
before
the
next
resource
can
be
applied,
and
so
it
ends
up
being
a
a
significant,
a
significant
effort.
Katrina
has
a
question.
K
Yeah,
I
just
have
a
point
on
that.
I
think
it's
great,
that
six
cli
is
is
working
on
that
in
in
cli
youtube
tillerson.
That
will
be
able
to
provide
maybe
a
more
canonical
solution,
but
I
wanted
to
mention
that
there
are
already
a
whole
series
of
mature
tools
out
there
that
that
do
this
and
have
their
own
opinions
about
how
it
should
be
done.
I
used
to
maintain
one
of
them.
K
That's
why
I'm
pretty
keenly
aware
of
it,
but
I
know
it's
also
was
not
the
only
the
one
out
there,
so
the
one
I
maintained
was
called
crane.
K
There's
another
one
called
k
up
and
there's
others
that
I
don't
know
the
names
offhand
and
I
think,
there's
kind
of
a
matter
of
a
separation
of
concerns
here
where,
like
customize,
is
a
configuration
generation
tool
and
I
think,
like
the
way
that
you're
using
it
is
the
canonical
way
where
you're
committing
the
customizations
to
a
repo
and
then
and
then
the
deploy
is
a
separate
step,
and
then
the
other
set
of
concerns
is
the
like
actual
deploy
step.
Where
you
have
those
orchestration
concerns
and
those
tools
care
about
ordering.
K
They
care
about
dependencies
and
that
sort
of
thing,
so
that's
my
take
on
it.
A
A
But
if
you
create
a
crd
that
is
somehow
bigger
than
the
namespace
or
you
come
up
with
some
primitive
to
organize
several
namespace
into
a
bigger
namespace,
there's
no
way
customize
will
be
able
to
tell
that
this
crd
should
be
or
the
cr
should
be
created
before
namespace
and
only
then
follow
up.
So
what
we
did
also
in
our
internal
tooling
is
that
we
have
numbers
and
the
files
and
the
files
are
processed
in
order
of
defining
them.
A
That's
one
possible
take
and
it's
pretty
explicit,
because
the
file
name
tells
you
the
order,
and
you
can
easily
tell
how
this
should
be
organized
and,
like
jeff
said
earlier,
that
capability
already
exists.
I'm
I'm
pretty
sure
that
it'll
take
some
more
time
to
rework
the
manifest
for
kubeflow.
Given
that
there
are
this
many,
but
that
will
be
the
general
direction
that
we
will
be
pushing
people
towards.
J
Yeah,
the
flip
side
of
all
that
is
that
it's
completely
theoretically,
true
that
you
can
always
make
a
custom
resource.
That's
going
to
be
really
slow
to
get
ready,
and
you
know
it
doesn't
even
matter,
maybe
where
you
order
it.
It
may
just
be
always
really
too
slow
and
everything
else
is
going
to
come
in
and
fail
unless
you
break
it
up
into
two
separate
applies
and
you
do
a
weight.
J
But
that
said,
you
know
it's
possible
that,
with
for
kubeflow's
particular
application
space,
these
kinds
of
that
kind
of
stuff
just
doesn't
happen
most
of
the
time
and
you
might
be
able
to
just
get
what
you
want.
99
of
the
time
by
you
know
ordering
them
in
the
customization
files.
D
To
clarify,
we
do
get
these
types
of
errors,
it's
not
that
we
don't
get
them.
We
do
get
their
errors
when
you
apply
crd
and
then
you
play
a
custom
resource
and
it's
not
ready.
This
is
normal.
This
is
how
kubren
this
works
and
the
way
that
we
get
around
it,
because
we
try
to
be
as
standard
as
possible
and
let
people
use
their
own
tooling
is
that
we
simply
apply
again.
D
D
We
and
what
happens,
is
in
the
older
customized
releases.
These
were
applied
first,
so
so,
when
you
applied
the
packets
first,
you
registered
the
new
logic,
and
then
you
applied
the
custom
resource
and
the
new
logic
run,
which
is,
I
guess
why
you
applied
that
new
logic
in
the
first
place.
But
then
what
happened
is
that
someone
created
the
customization
which
had
a
mutating
web
a
web
hook
and
the
custom
resource
in
it.
They
tried
to
apply
it
and
it
failed
because
the
web
hook
was
not
ready
yet,
which
is
normal.
D
It
takes
some
time
to
be
ready
and
the
way
they
fix
that
is
they
reorder
the
thing.
So
when
you
apply
new
logic
and
the
new
resource,
the
new
logic
is
registered
after
the
resource
is
created,
so
it
never
runs.
Essentially
they
override
it.
This
way,
and
what
happened
in
our
case
is
that
it
placed
the
mutating
wave
hook
for
istio,
which
injects
side
cars
at
the
very
end.
So
everything
was
created,
nothing
got
an
issue
sidecar
and
it
failed
which
didn't
happen
in
previous
versions.
D
J
So
we
can,
we
can
do
something
like
that.
I
just
want
to
you
know,
make
sure
everybody
understands
it
for
the
record
like
if
something
doesn't
work
because
of
an
order
and
you
change
the
ordering
and
it
does
work
effectively.
You
just
kind
of
got
lucky
right.
It's
not
going
to
work
long
term
because,
because
of
the
reconciling
nature
of
kubernetes
so
like
it
might
work
on
monday,
but
then
on
tuesday
it
won't,
but
again
just
for
practical
matters.
J
A
point
too
that
we
it's
it's
maybe
sets
a
dangerous
precedent
to
like
have
clients
do
the
sort
of
ordering
without
having
the
without
educating
people,
that
they
also
need
to
think
about
waiting.
C
Yeah,
so
that
was
that
would
be.
My
point
is
that
you
could
get
the
correct
ordering
and
that's
not
enough
right.
So
you
could
you
could
get
where
it's
creating
the
crd
before
the
custom
resource
and
there's
a
race
condition
there,
because
it
could
be
that
the
crd
is
not
yet
established
when
you
apply
the
custom
resource.
C
D
I'm
not
I'm
not
talking
about
the
waiting
part,
though
I
understand
that
is
out
of
scope,
of
course,
and
the
problem
is
that
if
you
try
the
reverse
order
in
your
example,
if
you
try
to
apply
the
custom
resource
first
and
then
the
crd,
it
will
fail,
it
won't
succeed
with
in
their
own
state.
This
is
what
yeah.
J
I
understand
I
understand
your
argument
is
that
it
will
guarantee
to
always
fail.
It
doesn't
have
a
chance
to
succeed
at
all
yeah
yeah
right.
An
argument
is
that
when
it
does
succeed,
it's
kind
of
you're
just
getting
lucky
anyway,
because
fundamentally
kubernetes
is
a
giant
race
condition
right,
yeah.
K
Yeah,
I
think
the
point
for
us
for
customize
is
that
it
was
not
really
advisable
in
the
first
place
to
try
to
have
an
opinion
about
order.
That
seems
deployable.
That
was
arguably
a
mistake.
We
can't
take
it
back,
so
we
have
the
legacy
flag
and
preserves
the
old
behavior,
but
we
don't
want
to
continue
to
be
in
the
business
of
trying
to
determine
an
order.
K
That
seems
like
it's
going
to
work
because
that's
kind
of
a
doomed
effort,
so
something
in
the
middle,
like
jeff,
said
we're
enabling
the
user
to
open.
K
That
is
not
going
to
work
for
somebody
in
the
long
run,
but
would
still
have
some
injection
flexibility
for
for
folks
who,
who
are
I'm.
D
Yes,
I
I
think
katrina
that
you
dropped
for
like
10
seconds.
I
saw
someone
else
in
the
chat
as
well,
so
it
wasn't
only
me.
A
We're
we're
in
agreement
with
regards
to
the
ordering
and
there
that
it's
something
out
of
this
cult
of
customized
and
I'm
gonna
cut
this
further
discussion
about
cupolo
use
of
customize.
We
can
continue
the
topic,
though
next
time
we
still
have
about
25
minutes
before
top
of
the
hour
and
then
a
call,
and
I
would
like
to
go
through
our
120
120
planning
attempts
so
janice.
A
F
D
A
Cool
awesome,
let
me
share
back
my
screen,
because
I
have
all
the
all
the
things
over
there.
Of
course,
I
lost
everything
else,
so
122
planning
my
target
is
being
all
over
the
place.
First
of
all,
there
are
two
actual
enhancements
that
we
are
working
on.
First,
one
is
the
sub
resource
support
for
cube
cuddle,
which
nikita
presented
about
four
weeks
ago.
If
I
remember
correctly,
this
was
basically
about
adding
a
sub
resource
flag
to
support,
getting
editing
and.
A
Patching
and
applying
sub
resources
like
scale
or
status,
are
there
any
particular
questions?
I
remember
and
nikki
recommended
earlier
today,
in
the
slack
thread
that
I
created
last
week
for
all
those
topics
that
I
put
also
in
the
agenda.
If
anyone
had
any
questions,
I'm
quickly
checking
the
slack.
A
A
Okay,
moving
on
to
the
next
one,
this
one
is
actually
from
sean,
so
sean
in
121
introduced
something
like
http
request,
headers,
where
we
send
information
about
the
command
and
argument
you're
passing
to
a
cubecado
command.
A
This
is
useful
for
monitoring
what
people
are
using
cubecardo
for
what
are
the
most
frequently
executed
commands,
what
kind
of
flags
and
so
forth.
A
C
That's
correct
so
moving
it
to
beta.
It's
already
alpha
the
code
for
alpha
and
a
feature
flag
already
exists,
so
I've
started
working
on
the
cap
to
bring
it
to
beta,
and
I
think
one
of
the
bigger
one
of
the
bigger
items
we
have
to
or
hurdles
we
have
to
get
past
and
I
don't
think
we
have
to
do
it
in
the
cap.
C
But
we
we
need
to
actually
have
some
documentation
in
the
real
coop
cuddle,
because
it'll
be
once
we
move
it
to
beta,
it
will
be
default
and-
and
so
there's
a
couple
of
hurdles
the
I
know
that
the
enhancement
freeze
is
the
13th
next
week
and
I
I
think
that
a
lot
of
what
I'll
be
implementing
with
the
the
cap
is
like
is
saying
what
we
are
going
to
do,
because
the
code
freeze
is
going
to
be
after
after
that,
but
but
yeah
I've.
C
I've
started
along
this
process
and
really
glad
to
see
that
we
had
a
couple
of
people
from
the
enhancement
liaison
here
today.
How
you
doing
I'm
sean
I'll
be
trying
to
move
this
to
beta.
A
I
had
one
question
with
regards
to
this:
I
remember
a
discussion
a
couple
weeks
back
that
the
x
headers
were
not
passed
into
the
cube
api
server.
Is
that
something
that
requires
further
work
or
eventual
discussion
with
sig
api
machinery.
C
So
I
haven't
actually
that's
a
good
point.
That's
something
that
I
need
to
bring
up.
What
I
thought
I
was
answering
there
was
that
the
ietf
has
asked
that
has
has
mentioned
that
x,
headers
are
no
longer
supported,
and
so
we're
going
to
change
our
headers
to
just
basically
cuddle
command
or
whatever
the
name
of
the
head.
Our
headers
are
they're
not
going
to
be
prefixed
by
x.
C
But,
yes,
that's
correct,
that's
something
that
had
dropped
off
my
my
radar,
which
is
to
talk
to
the
api
machinery
to
ask
about
how
these
headers
are
being
sent
to
mutating
and
validating
web
hooks,
but
also
how
they're
going
to
be
processed
to
extension.
Api
servers,
which
is
kind
of
a
esoteric
topic
itself,
how
these
headers
are
moving
around
the
the
api
machinery
and
maybe
even
how
they're
being
logged.
A
C
This
that
that
was
my
assumption
was
that
there's
going
to
be
a
significant
amount
of
value,
even
if
they
aren't
being
sent
to,
for
instance,
validating
and
mutating
web
hooks.
Our
biggest
use
case
that
we've
identified
is
basically
just
telemetry.
Just
you
know,
storing
these
headers
in
logs
that
somebody
else
can
consume
and
see
how
people
are
using
coop
control
in
our
in
the
system.
A
We
can
push
that
one
over
before
ga.
It
would
be
good
to
make
sure
this.
This
gets
reflected
in
the
enhancements
europe.
You
will
be
updating
good
point.
Cool
thanks,
yeah
go
ahead.
I.
B
Wonder
if
sean
do
you
have
any
thoughts
around
how
we
as
a
sig
can
get
some
of
that
data?
Maybe
we
can
put
out
a
call
to
the
end
user
community
and
give
them
a
query
to
run
that
they
can
share
that
with
us
and
make
sure
they
obfuscate
anything
private.
C
C
I
think
that
there's
gonna
be
you
know
dozens
and
hundreds
of
different
ways
in
which
this
data
gets
surfaced,
but
we
would
like
to
know
at
least
some
of
the
common
ways,
and
I
yeah
I
don't
think
so.
I
guess
the
answer
is.
C
I
don't
think
that
there's
just
one
way
it's
going
to
be
done,
I
think
it's
you
know
like
for
aws,
there's,
probably
a
monitoring
solution
that
goes
through
logs
and
surfaces
stuff,
and
I
think
that
you
know,
maybe
microsoft
has
their
own
tools
to
do
that,
and
and
this
I
guess
one
of
the
standard
tools
is
going
to
be
what
prometheus,
or
so.
I
think
that
there's
going
to
be
many
different
ways
that
this
happens,
but
we
we
should
have
our
yeah.
C
A
Yeah,
I
think
we
can
probably
put
together
a
blog
post-
maybe
not
necessarily
right
now,
but
we
can
target
that
for
let's
say
123.
A
A
C
And
the
most
important
concept
is
that
we
have
a
session
id
for
each
command
so
like
each
command
can
end
up
with
dozens
of
rest
calls,
and
we
want
people
to
know
that
they
should
take
those
and
each
one
of
them
will
have
the
the
same
uuid
the
same,
unique
identifier,
so
that
you
can
combine
these
and
say:
okay.
Well,
these
seven
rest
calls
were
actually
just
one
coupe
control
call
and
combine
them
correctly.
C
So
so
we're
gonna
actually
have
to
do
some,
some
education,
so
that
whoever
is
gathering
this
telemetry
knows
how
to
to
do
that
process.
A
Okay,
moving
on
to
the
next
topic,
so
we've
been
talking
about
this
topic
for
quite
a
while,
and
after
last
week's
or
even
that
was
before
the
box
crop.
I've
opened
an
umbrella
issue
where
I
described
the
end
goal,
which
is
that
we
want
to
refactor
all
the
commands
and
to
be
able
to
split
flags
from
options
eddie
and
myself.
We
dug
into
it
for
a
little
bit
and
we
found
that
there's
a
weight
command.
A
That
clearly
does
that
already.
This
will
be
something
like,
and
we
already
have
a
bunch
of
folks
that
accepted
and
picked
up
different
commands.
How
I
started
looking
into
logs
and
get
eddie
eddie
said
he
will
look
into
some
simple
conflict.
Commands
katrina
mentioned,
train
or
describe
ricardo
mentioned
basic,
create
commands,
and
I
see
other
people
already
are
looking
into
this
topic.
So
I'm
pretty
sure
that
we
can
somehow
slowly
spread
the
love.
A
What
I
would
like
to
achieve
from
this-
and
that's
not
definitely
not
something
that
we
want
to
see
happening
in
around
122,
but
maybe
in
the
long
run
without
any
specific
date,
is
a
description
of
what
a
perfect
command.
Maybe
perfect.
Yeah
something
along
those
lines.
What
a
typical
cube
cuddle
command
looks
like
I've
been
and
myself
sean
eddie
and
pretty
much
all
of
us.
We've
been
giving
several
talks
already
about
the
topic.
A
What
a
cube
cuddle
command
looks
like
what
I
would
like
to
see
is
formalize
what
we
come
up
with
after
working
through
several
examples
and
put
those
thoughts
into
a
cap
and
then
when
there
will
be
new
people
coming
in,
we
will
be
able
to
point
them
to
this
as
their
starting
point
and
as
a
reference
for
working
with
cube
cuddle
so
feel
free
to
look
at
it
and
comment
and
pick
up
other
commands,
which
you
are
interested
in
working
on.
B
Yeah,
I'm
gonna
toss
a
disclaimer
on
that
thread
too,
so
so
lao
actually
already
did
a
pr
shout
out
to
lao
great
job.
I
I
took
a
look
at
it
last
night
I
tossed
a
hold
on
there.
I
think
he
pulled
out
the
validate
and
complete
commands,
which
I
don't
think
the
weight
had.
So
I
think
we're
just
basically
the
tldrs.
B
I
think
we're
going
to
have
to
see
a
few
of
these
before
we
say
like
yes,
this
is
the
way
forward
for
sure
and
I'm
sorry,
I
just
I'm
going
to
toss
a
disclaimer
on
there.
That's
just
like
hey,
like
you
know,
just
this
one,
we'll
we'll
sit
for
a
bit
while
we
gather
and
wait
so
if
you're
taking
this
on,
just
be
mindful,
you
might
have
to
redo
a
bunch
of
work.
If
we
change
your
mind
so.
A
Yes,
and
it's
definitely
not
something
that
we
are
oh,
it
has
to
be
done
for
122.
I'll,
be
prepared
that
it
might
take
a
little
bit
a
few
iterations
where
we
will
be.
We
are
in
parallel.
We
are
also
looking
into
this
from
different
commands.
Different
commands
have
different
specifics
and
we're
trying
to
wrap
the
head.
A
Our
heads
around
wait,
which
art,
which
I
linked
there
is
very
simplistic
and
it
was
written
from
start
as
a
command
that
can
be
embedded
in
different
commands,
so
it
was
created
such
that
it
was
very
from
start.
It
was
created,
as
with
the
separation
in
mind,
a
lot
of
the
existing
commands
were
not.
Some
are
simpler.
A
Some
are
a
little
bit
more
complicated,
so,
while
you're
working,
it
will
depend
on
which
commands
if
you
pick
up,
for
example,
api
resources
or
api
versions.
These
were
available
simple.
If
I
remember
correctly,
one
of
those
I
can't
remember,
which
does
not
even
have
flags,
so
in
that
case
the
split
won't
even
happen.
A
A
Okay,
the
next
topic
is
something
that
is
open
for
way
too
long
since
november
1.
and
mark
put
together
and
rewrote
all
the
completions
to
go
from
what
I've
heard
from
people.
It
is
lightning
fast.
A
It
is
working
amazingly
well,
I
haven't
personally
got
a
chance
to
look
into
it
and
I
just
realized
that
it
can
help
just
change
their
colors,
because
the
links
are
not
blue
anymore,
which
is
a
little
bit
irritating.
They
are
just
bold
as
a
side
off
topic,
so
we
would
want
to
see
a
little
bit
more
folks
looking
into
this
pr.
This
is
something
that
I
do
want
to
see
merged
for
122..
A
We
will
definitely
won't
be
able
to
bring
this
back
because
that's
a
new
feature
and
we
we
don't
bring
back
features,
but
from
the
initial
discussion
that
I
had
with
eddie,
I
think
sean
also
mentioned
looking
into
it.
It's
pretty
good
in
brian
as
well.
It's
pretty
good,
it's
pretty
bad,
unfortunately,
so
that
requires
a
little
bit
of
more
eyes
looking
into
this.
A
So
if
you
have
some
time,
have
a
look
at
it
and
give
it
a
try
locally
that
that
might
be
a
reasonable
feedback.
If
you
just
say
there
that
oh,
I
give
it
a
try.
I've
tried
this
that
and
that,
and
it
all
worked
and
you
know,
put
their
ear
thumbs
up
and
then
we
will
know
that
the
more
people
tried
it
out.
There
won't
be
any
major
problems
with
this.
C
C
So
so,
if
you're
going
to
work
on
a
particular
few
commands,
just
I
would
say,
throw
a
comment
on
the
pr
and
say:
I'm
I'm
validating
these
x
commands
these
n
commands,
and
so
we
can
all
we
all
don't
pile
on.
Just
like
one
part
of
this
pr,
one
one
or
two
commands
we
can
actually
get
all
all
of
them.
L
G
L
Of
the
suggestions
I
made
was
refactoring
that
into
a
reusable
completions
component,
because
you
know
if
you're
fetching
a
list
of
pods
in
one
command
and
you're
doing
it
again,
another
command.
You
don't
have
to
like
test
that
in
both
commands.
You
can
test
that
in
one
place
and
I
so
I
think
if
he
refactors
it
a
little
bit,
it's
going
to
be
a
little
less
intimidating
to
review.
A
Sure
and
to
add
a
little
bit
more
color,
this
particular
issue
will
help
us
solve.
I
remember
the
two
most
frequently
upvoted
issues
for
cub
cuddle,
which
also
goes
in
line
with
what
I
stated
in
the
email
sent
out
about
a
week
ago
that
we
would
be
trying
to
solve
first,
the
most
voted
issues,
so
yeah
eddie,
just
saying
that
powershell
and
fish
was
the
two
most
frequently
requested
completions
for
campground
okay.
So
that
would
be
the
topic
that
we
will
be
actively
working
on.
A
Some
are
specifically
targeted
for
122
others
unnecessarily,
and
there
are
three
other
topics
that
I
want
to
bring
up,
which
are
more
for
long
for
longer
run.
The
first
one
is
from
eddie
about
adding
fancy
colors
but
more
like
about
user
preferences
in
general,
and
you
want
to
talk
about
a
little
bit
about
it
yeah.
So
I'll
specifically
do.
B
I
will
specifically
be
doing
a
cap
for
the
user
preferences
so
like
a
cube,
rc
or
whatever
the
hell,
we
decided
to
call
it
the
I
I
can.
I
can
paint
detailer
one
and
we
can
pick
up.
I
think
that
they
were
looking
at
doing
the
actual
kept
for
the
the
colors,
but
the
the
cube
rc
user
preferences
separating
that
from
a
actual
cube
config
for
cert.
You
know
cluster
cert
authentication.
What
not
that
that's?
A
Yeah
any
particular
questions,
if
you
feel
like
you,
have
some
input
feel
free
to
reach
out
to
eddie
and
either
ask
questions
or
provide
feedback.
I'm
pretty
sure
that
the
more
feedback
we
get,
the
better
the
the
outcome
and
the
enhancement
will
be.
A
B
To
help
too
so
I'll,
I
don't
know
if
I'll
get
that
kept
done
for
next
week,
but
I'll
definitely
start
on
it.
A
Yeah,
like
I
said,
it's
not
122
items,
so
it
doesn't
have
to
land
or
even
the
npr
right
away.
The
next
topic
is
actually
from
ricardo
and
he
already
sent
out
an
email
to
kdf
and
six
yellow
as
well
about
cue
color
return
code
normalization.
A
A
A
A
Okay,
the
last
one
we
still
have
about
two
minutes
I'll,
try
to
finish
on
time,
cube
cattle
events,
that's
something
that
I
put
together
with
himachal
from
sig
ux.
A
Basically,
the
idea
was
to
have
an
improv
version
of
cube,
cuddle,
get
events
which
would
allow
you
to
filter
and
work
with
events
in
a
smarter
way
than
what
does
what
get
does
currently,
because
it's
very
limited
being
a
generic
command
for
everything
the
proposal
itself
was
merged
last
year.
Unfortunately,
the
implementation
hasn't
moved.
If
I
remember
currently,
there
is
one
person
that
was
interested
in
he
put
together
a
proposal,
a
proposal
implementation-
I
haven't,
looked
at
it.
A
I
would
like
to
give
it
a
try,
but
I
I'm
not
sure
if
this
will
be
something
that
we
will
push
through
for
122.
Even
if
we
we
will,
that
would
be
an
alpha
command.
So
I'm
I'm
not
worried.
We
can
start
slowly
pushing
this
forward.
A
If
I
remember
correctly,
if
the
issue
the
pr
should
be
linked
towards
the
end-
yes,
so
brian
put
together
from
reworks
a
proposed
solution,
so
if
you're
interested
into
it,
feel
free
to
have
a
look
at
it,
and
you
can
have
a
discussion
in
that
vr
as
well.
A
F
No
we're
still
working
on
some
things,
but
no
no
status.
Okay,.
A
Cool
fair
jeff,
any
update,
customize
wise.
A
If
I
remember
correctly,
there
was
a
topic
with
regards
to
open
api
in
customize,
because
I
can't
remember
his
name.
He
got
rid
of
the
open
api,
go
open,
a
library
specifically
in
cube
cuddle,
and
he
had
an
open
pr
for
removing
that
same
thing
from
customized.
I
think
katrina
was
looking
at
it,
I'm
not
sure
yeah.
J
Or
aware
of
it,
that's
being
discussed
in
pull
pr
3827,
there
are
some
issues
with
that.
I
think
it's
going
to
work,
but
there
are
some
issues
with
folks
that
are
depending
on
the
old
version.
20
or
older
of
you
know
the
fundamental
case,
libraries,
because
those
depend
on
a
very
old
version
of
customize,
so
anybody
who's
depending
on
what's
old
versions
of
customized.
But
then
the
new
open
api
stuff
is
going
to
have
some
linker
errors.
J
Basically,
but
we
do
have
a
path
forward
and
that's
being
discussed
in
the
the
pr
3827
and
customize
cool
awesome.
That's
great
to
hear.
A
Okay,
we're
one
minute
past
top
of
the
hour.
I
think
we
are
done
for
today.
Thank
you
very
much
all.
It
was
very
amazing,
seeing
you
as
usual,
and
we
had
an
amazing
discussion
throughout
the
past
hour.
Thank
you
very
much
see
you
in
two
weeks,
if
I
remember
correctly,
bye
all.