►
From YouTube: Kubernetes SIG CLI 20210616
Description
Kubernetes SIG CLI bi-weekly meeting on June 16th, moderated by Sean Sullivan (co-chair). The first part of the meeting was accidentally missed.
B
In
situations
where
it
would
break
yeah
compatibility
right,
let's
I
mean
I
think,
that's
like
a
good
starting
point.
Let's
document
it,
let's
there's
that
api
conventions
folder.
That
seems
like
the
right
place
to
put
this
as
a
new
file.
It
doesn't
have
to
be
a
long
thing,
but
should
just
say:
like
hey,
here's
what's
already
happening,
and
here's
how
we
should
do
things
in
the
future.
I
know
in
a
way
that
is
the
most
consistent
with
what
we've
already
done
and
you
know
meets
our
general
philosophy
and
goals.
A
D
B
B
E
F
Yeah,
it's
not
maintained
by
us.
It
was
donated
by
a
company
called
replicated,
but
the
tutorials
are
old
and
out
of
date
and
people
file
issues
in
the
customized
repo
for
it.
So
we
were
wondering
if
there's
anyone
who
who
knows
someone
that
we
can
reach
out
to
to
like
either
get
it
fixed
or
redirect
to
our
documentation.
B
B
G
B
E
So
I
think
we're
we're
coming
to
the
end
of
what
we
had
for
topics,
and
I
think
that
we
should
move
on
to
actually
maybe
have
some
some
quick
stand
up,
since
we
do
have
so
much
time.
I
know
that
that
nick
mitchell
is
here.
At
least
I
saw
his
I
saw
him
earlier.
Would
anyone
like
to
give
a
quick
update
on
how
on
the
status
of
kui.
D
Sure,
yes,
sorry,
I
missed
the
last
meeting.
I
had
a
doctor's
appointment,
howdy
everyone
so
cool.
Yes,
we're
we're
going
pretty
well
we're
in
the
process
of
rolling
out
or
finishing
up
the
10.4
release.
Good
news.
This
release
has
behind
it
a
red
hat
release,
so
we
got
a
lot
of
bug.
Fakes
help
from
their
side
to
help
stabilization,
so
that
should
be
released
soon.
D
Also
in
this
release,
there
will
be
initial
support
for
talking
to
s3
via
a
lightweight
vfs,
so
you
can
ls
slash,
s3,
aws
and
so
on,
so
just
some
initial
support
for
for
the
aws
and
ibm
and
we
can
add
azure
and
gcloud
as
well.
If
needed,
you
know
I
just
that's
what
support
supported
me
in
any
of
this
initially
so
and
of
course,
a
bunch
more
bug
fixes,
so
that
should
all
be
coming
out
open
a
few
days
we're
just
facing
a
few
regressions
now.
E
C
Customize,
I
can
just
speak
to
a
couple
points.
One
thing
is
that
we
merged
a
markdown
copy
of
the
roadmap
that
we
brought
to
a
previous
6cli
meeting
right
into
the
root
of
the
repo.
So
if
anyone's
interested
in
reviewing
that
is
that
they're
easy
to
access,
I
can
also
add
a
link
to
that
to
the
meeting
notes.
C
There,
where
go,
yaml
had
a
breaking
white
space
change
and
basically
it
changes
the
indentation
of
lists,
and
it's
not
configurable
and
the
upstream
author
is
super
against
making
it
configurable
or
undoing
that
change.
That
happened
very
randomly
within
the
v3
release
and
the
reason
this
is
an
issue
for
us
is
that
this
actually
happened
a
while
ago
and
customize
bumped
the
version
of
go
yaml
that
they
use
a
few
months
ago.
C
I
believe,
and
because
sorry
not
customized
cube
cuddle
did,
and
that
means
that
the
version
of
customize
that
is
vendored
into
cute
cuddle.
If
the
release
happened
today
say,
would
contain
that
breaking
white
space
change
in
customize
a
side
we're
able
to
pin
the
old
version
so
that
we
don't
actually
pick
it
up
and
we
don't
break
our
users.
But
we
can't
do
the
same
in
cube
cuddle
it
because
of
the
way
that
the
the
depths
are
in
in
kubernetes
kubernetes.
C
C
So
jeff
was
basically
working
on
making
a
pr
to
go
yaml
with
that
change
in
it,
and
then
I
was
going
to
vendor
that
into
kml
turns
out.
The
change
is
extremely
difficult
to
make
and
jeff
is
now,
as
we
were
discussing
on,
leave
and
the
code
freezes
impending,
so
that
this
is
a
very
important
issue
that
we're
going
to
need
to
tackle,
and
if
anyone
is
interested
in
helping
out
with
that,
we
could
really
use
it.
B
What's
the
impact
of
the
white
space
changes
to
consumers
of
customize?
Is
it
like?
Is
it
only?
Is
it
concerned
mostly
like?
Oh,
their
gifts
are
going
to
look
different
and
maybe
their
tests
are
going
to
break
because
their
tests,
like
have
you,
know,
static,
results,
golden
files,
or
is
it
like
this
changes
the
semantic
meaning
of
the
contents
in
such
a
way
that
could
impact
their
workloads
or
something
like
that.
A
I'm
glad
you
asked
because
so
sorry
like
as
a
consumer
of
customized,
did
that
actually
affected
us.
So
formatting
is
one
of
those
things
is
very
prominent
part
of
user
experience.
Right
so
imagine,
for
example,
in
go
land,
you
suddenly
confirm
changes
and
you
have
millions
of
lines
of
gold
written.
It's
just
very
disruptive
change
and
the
fact
that
this
a
library
doesn't
allow
configuration
of
formatting.
It's
just
bizarre,
in
my
opinion,
so
forking
it
or
even
going
to
the
extreme
of
choosing
a
different
library.
C
C
A
B
A
I
can
talk
about
the
consumer
user
experience
so
like
someone
who's
using
customized
libraries
and
you
create
a
tools
based
on
that.
So
one
day
you
you
upgrade
that
tool
and
you
run
it
and
then
all
your
configuration
all
your
manifest
suddenly
have
this
superfluous
formatting.
So
it
is
disruptive.
It
masks
real
changes,
the
real
configuration
changes
that
has
some
side
effects
that
could
lead
to.
So
it's
it's,
it's
pretty
bad
ux.
I
don't
know
how
you
would
categorize
the
severity
of
it.
I
think.
B
A
C
Yeah,
just
from
the
customized
perspective,
I
wanted
to
say
that
we
do
think
that
this
is
a
really
big
deal.
If
we
were
to
accept
this
change,
we
would
cut
up
version
5.0
for
it
and
any
similar
change.
We
would
consider
a
major
breaking
change
for
our
users,
because
we
expect
that
people
are
using
this
and
get
ops
workflows,
and
there
are
various
ways
that
this
can
be
extremely
disruptive
from
them.
Does
it
change
the
semantics?
C
Repo
itself
has
a
huge
amount
of
tests
that
are
intentionally
brittle
so
that
we
realize,
if
we
make
this
kind
of,
if
we
introduce
this
kind
of
regression.
So,
even
within
the
repo
itself,
you
can
see
the
the
huge
impact
that
something
like
this
would
have,
even
though,
on
the
surface,
it
might
seem
like
a
minor
styling
issue.
C
So
to
your
point,
like
I
think,
since
we
were
thinking,
we
had
plenty
of
time
and
the
fix
was
not
actually
as
difficult,
so
we
thought
let's
move
forward
and
try
to
get
this
solved,
but
given
that
it's
more
difficult
and
time-consuming
than
we
thought,
I
think
your
point
that
we
could
investigate
rolling
back
the
problem
is
is
good.
I.
B
Just
reach
out
and
like
I
know,
maybe
that
got
updated
because
someone
ran
go
mod,
tidy
right
and
it's
like
you
know.
No
one
cares
right
so
or
maybe
it's
like
critical,
because
there's
like
a
bug
in
the
old
version,
that
is
like
security,
vulnerability
right
and
therefore
it
has
to
be
upgraded
right.
C
F
C
F
E
E
I
have
talked
to
daniel
smith
of
the
api
machinery
and
we
we
will
have
to
perform
other
steps
in
order
to
surface
these
headers
and
interface
with
the
api
machinery
to
to
make
this.
These
headers
consumable,
because
they're
not
stored
in
the
they're,
not
logged
by
default
and
they're,
also
not
sent
to
validating
or
mutating
web
hooks.
It
sounded
like
daniel
was
interested
in
allowing
that
so
yeah.
So
that's
where
coop
cuddle
headers
stands
right
now.
B
B
B
Phil,
well
that's
great
work
like
this.
This
is
one
of
those
features
that
I
think
has
been
something
we
knew
we
wanted
to
do
for
a
while
and
and
it's
great
to
see
it
come
to
fruition
and
what
couldn't
potentially
help
us.
As
a
group,
you
know
make
decisions
with
like
more
data
about
how
the
tool
is
actually
being
used
rather
than
speculation
and
anecdotal
evidence.
D
On
that
note,
actually
we
actually
have
a
summer
intern
starting
up
who's
going
to
maybe
us
looking
into
some
usability
ux
studies
with
kui.
Is
there
any
sort
of
should
we
should
be
crosstalking
with
zig
ux?
Is
there
any
sort
of
do
they
have?
Should
we
get
them
involved,
or
should
we
just
treat
this
as
a
summer
insurance
project
and
let
the
intern
go
crazy?
Do
you
think
there's
any
need
to
really
benefit
from
involving
a
larger
audience
on
this.
B
D
No,
no.
This
is
just
a
separate
question
on
the
topic
of
of
getting
feedback
from
you
know,
on
usability
on
on
our
tooling.
I
just
brought
out
that
we
have
an
intern
starting
who's
going
to
be
looking
at
the
other
that
from
a
kooi
angle,
so
thought
maybe
there
might
be
some
some
larger
topics
that
we
could
that
we
could
help
out
with
there.
B
You
know
ux
I'd
say:
there's
two
two
challenges.
The
first
is
that
you
know
the
the
survey
based
sort
of
how
much
do
you
like
this
or
that
sort
of
thing
tends
to
be
noisy
it's
hard
to
write
the
questions
so
that
you're
not
right,
like
you
get
one
chance
to
ask
the
question
and
then
you're
like.
Oh,
that
isn't
actually
what
I
wanted
to
ask
now.
E
B
The
thing
we
could
do
to
most
improve
the
ux
actually
would
be
to
make
the
ap
the
api.
If
you
will
configurable
on
the
client
side,
and
so
you
can
opt
in
to,
for
instance,
a
you
know,
the
the
like
the
dev
or
the
user
centric
api
or
the
which
has
fewer
backwards
compatibility,
guarantees
right
that
you're
opting
into
you're
saying
I'm
running
this
on
my
laptop
and
exiting
xero
and
giving
me
an
error
message
when
we
decide
we
don't
like
that
particular
command
anymore.
B
D
That's
great
in
terms
of
like,
should
we
be
considering
injecting
any
like,
like
with
your
headers
in
the
in
coop
cuddles?
Should
we
be
doing
something
similar
in
kuia
like
we're
doing
no
usage
tracking
at
all
I'd
do
that
intentionally?
In
fact,
I
stripped
out
a
lot
of
that
stuff
when
we
did
the
initial
contribution,
because
I
figured
let's,
you
know,
err
on
the
side
of
caution,
but
is
there
something
we
should
be
doing
there
to
help
with
getting
feedback
in
terms
of
you
know
all
the
standard
task
accomplishments?
D
B
I
think
one
thing
to
ask
first
is
who
is
getting
the
data
like
in
the
case
of
the
headers
right,
we're
sending
it
to
the
api
server
as
part
of
the
request
write-in,
so
the
data
isn't
being
exported
to
6cli,
for
instance,
it's
being
exported
to
your
api
server
and
hopefully
we
can
use
that
to
have
customers,
maybe
like
we
can
build
a
tool,
then
that
allows
customers
to
publish
or
something
like
that
right,
but
it's
still
staying
within
their
system.
Yeah-
and
it's
also,
I
think
another
thing
to
consider
is
like.
B
F
B
And
I
just
like
you
know,
destroyed
my
livelihood
or
something
like
that,
and
you
could
imagine
saying:
okay
well
knowing
that,
because
control
deletes
each
namespace
as
a
separate
request,
it's
very
difficult
to
differentiate
between.
I
want
to
delete
this
one
namespace,
which
may
be
perfectly
fine,
and
I
want
to
delete
all
the
namespaces
which
isn't
fine
so
that
context
in
the
header,
for
instance,
you
could
imagine
someone
puts
a
something
in
in
front
of
their
cluster
like
either
a
proxy
or
they
put.
B
Maybe
it
would
work
with
validating
web
hooks,
or
you
know
these
sorts
of
things,
but
you
could
say:
okay
fail,
delete
all
namespaces,
and
in
that
case
it
has
nothing
to
do
with
us
gathering
data
it
has
to
do
with
giving
customers
the
rights
of
information
about
the
use,
what's
the
user's
doing
yeah.
So
so
I
guess
I'd
start
with
those
two.
B
Is
it
in
the
is
the
user,
the
owner
of
the
data,
and
then
is
it
like
data
that
is
generally
just
useful,
so
like
mouse
movements
which
would
not
qualify
for
the
latter
one
right,
but
but
like
from
a
high
level?
What
I
want
to
do
is
display,
you
know
all
broken,
all
pods
that
are
broken
in
my
system,
and
I
do
that
by
doing
five,
different
requests
and-
and
now
it's
it's
almost
like
a
trace
header
like
it's
almost
like
cross
stack
stuff
and
that
and
tying
together
stuff
with
context.
D
D
D
B
D
Have
there
been
any
studies
at
the
coop
cuddle
level
in
terms
of
like?
Do
we
have
a
notion
of
what
common
scenarios
that
we
should
be
tracking?
You
know,
like
you,
know,
creates
weights
inspect
delete
you
know.
Are
there
sets
of
tasks
that
you've
established
as
being
sort
of
meaningful
common
problematic?
Do
we
have
any
kind
of
repository
of
those
kinds
of
scenarios.
B
I
don't
think
we
have
if
we
do,
it
is
not
based
on
a
holistic
view
of
observed
patterns
from
a
broad
set
of
customers
across
cloud
providers.
I
could
imagine
like,
for.
I
could
imagine
most
likely
what
we
would
have
if
we
had
anything
at
all
would
be
like.
I
talked
to
customers
at
a
booth
at
a
kubecon
and
the
customers
that
I
talked
to
said
they
did
this
right.