►
From YouTube: Kubernetes SIG CLI 20190828
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
Okay
good
morning,
good
afternoon,
good
evening,
depending
on
where
you
are
it's
August
28,
and
this
is
another
six-year
line
meeting,
my
name
is
marcin
and
I'll
be
your
host.
We
have
short
lists
of
announcements
for
today,
code
freeze
for
gonna
tease,
1.16
is
due
tomorrow.
I
haven't
heard
about
any
changes
with
that
regards,
so
make
sure
that
all
the
PRS
that
you
care
about
are
tagged
by
tomorrow
afternoon,
Pacific
time,
if
I
remember
correctly,.
A
B
A
I
offer
view
during
the
main
part
of
the
cube
for
North
America
and
during
which
we
are
hoping
to
basically
touch
base
with
our
developers
and
end-users
and
we'll
be
hosting
some
kind
of
a
live
event
and,
depending
on
what
people
will
be
interested
to
hear
about
we'll
be
covering
the
topic,
whether
that's
a
direction.
Six
Eli
is
going.
If
people
will
be
interested
in
how
to
sew
the
develop
plug-ins
will
cover
that
one.
We
will
have
90
minutes
that
sufficient
amount
of
time
to
cover
at
least
couple
of
topics.
A
Okay
with
that
I
think
we
can
move
to
the
topics
section
and
the
first
topic
is
cube.
Cto
copying
shoes.
So
let
me
quickly
introduce
you
the
discussion
that
we
are
having
so
basically
over
the
past
six
months
or
so
but
differently.
Since
the
security
audit
started
in
communities
at
the
beginning
of
this
year,
we
have
noticed
a
significant
amount
of
CVE
great
issues
with
cube
CTO
copy,
not
of
the
other
commands
had
any
issues
during
that
period,
cube
CTL
copy,
head
numbers
issues
and.
A
A
A
In
the
meantime,
though,
we
still
would
like
to
improve
the
current
the
current
problems
that
we
have
with
cube,
CTO
and
ensure
that
no
further
CDE
issues
are
happening
against
cube
city
or
copy,
so
Jordan
listed
down
alternatives
for
the
current
process.
These
include
that
we
will
go
through
a
regular
deprecation
process,
which
is
either
twelve
months
or
two
releases.
Whichever
is
longer
in
our
case
as
we're.
Releasing
every
quarter
that
usually
means
an
entire
an
entire
year
of
supporting
cube,
CTO
copy
command.
C
E
To
clarify
a
couple
things
we,
this
is
a
discussion
that
we've
been
trying
to
have
for
a
while.
It
was.
We
were
kind
of
perpetually
in
a
state
of
addressing
a
security
issue
and
the
thought
was
like
as
soon
as
as
soon
as
this
one
gets
resolved
and
the
releases
get
pushed
out.
Then
we'll
start
the
discussion
about
like
how
can
we
deal
with
this
in
a
long-term
way
and
then
before
that
discussion
could
get
started?
Another
security
issue
would
get
reported
and
we'd
be
back
to
trying
to
figure
out.
E
E
E
That
would
disrupt
a
lot
of
people's
use
of
the
command,
who
don't
actually
care
about
these,
like
advanced
simile,
edge
cases
they're
just
using
it
to
pull
like
a
log
file
out
and
that's
not
unreasonable
thing
to
want
to
do,
and
so
I
I'm
in
favor
of
the
fix
for
the
current
security
issue
being
dropped.
Symlink
support
I
think
we
tried
really
hard
to
support
some
links
in
a
safe
way
and
have
failed
over
the
last
six
months,
so
I
think
droppings
and
link
support.
E
E
Is
this
sort
of
half
half
functional
command,
a
a
reasonable
user
experience
or
you
people
running
this
command,
expect
it
to
work
anywhere.
Kubernetes
works
like?
Are
they
expecting
the
container
runtime
to
provide
these
capabilities
so
those
things
weren't
really
considered
when
the
command
was
avid?
It
was
just
sort
of
a
this
is
clever.
You
can
do
this
and
it
kind
of
works
and
so
I
think
those.
E
Those
are
the
types
of
questions
that
we
need
to
ask
when
we're
deciding
whether
to
support
this
long
term
or
start
with
the
deprecation
period
and
the
urgency
of
removing
it
immediately
is
lowered
if
we
drop
out
if
we
gonna
define
clearly,
this
is
what
the
screen
is
for
and
what
it
does.
This
is
not
what
it's
for,
if
you
want
these
advanced
use,
cases
use
tar
directly,
which
you
can
do
today
like
going
back
many
many
versions
of
controlled,
so
there's.
E
E
So
things
like
wild
card
expansion
and
permission
preserving,
so
you
could
create
executable
files
locally,
so
getting
feedback
on
those
in
particular
and
kind
of
clarifying
in
the
command.
What
is
what
is
this
intended
for
and
limiting
it
to
that
so
I
agree
with
the
mailing
list.
Comments
that
any
exceptions
to
deprecation
policies
should
be
captured.
Benetcam
like
the
the
deprecation
policy
predated
caps,
and
so
it
said
if
we're
gonna
make
an
exception
for
reasons
that
this
policy
doesn't
foresee.
E
That'll
be
done
like
in
discussion
with
sig
leaders
and
people
maintained
errs
of
that
area,
and
so
that
discussion
needs
to
happen.
The
place
that
happens
today
is
a
cap,
so
that
makes
sense
to
me-
and
I
wouldn't
I
wouldn't
announce
any
kind
of
deprecation
in
116
before
we
figure
out
whether
we
actually
need
to
deprecate
it
or
want
to
deprecated
it.
So,
let's
not
disrupt
users
more
than
we
need
to
before
we
figure
out.
What's
the
right
thing
to
do
was.
E
Well,
I
think
the
first
step
would
be
just
to
direct
them
to
tar
for
use
cases
that
the
built-in
one
doesn't
support
things
like
siblings
information,
preserving
and
wildcard
crazy,
wildcard,
globbing
thing
if
I
think,
if
this
isn't,
if
what
you
want
is
tar
behavior
and
this
isn't
getting
it
to
you-
use
tar.
The
goal
of
this
command
is
not
to
expand
to
you
right.
C
C
E
I-I-I
would
probably
start
slow
with
what
we
direct
users
to
do
so,
let's
figure
out
if
this
functionality
makes
sense
or
like
Tim
talking
on
the
thread
was
talking
about
like
it's
just
something
we
want
to
teach.
You
contain
a
runtime
to
be
able
to
do
like
copy
a
file
out
until
we
actually
decide
those
things.
I
wouldn't
like
push
people
away
from
using
the
command
prematurely
so
saying.
E
E
Least
yeah
I
mean
we
have
a
responsibility
to
users
to
if
we
say
this
is
a
reasonable
thing
for
them
to
run
on
their
local
machine
like
make
it
reasonable
and,
like
I
said,
we've
tried
to
preserve
the
same
link
following
for
like
a
year
now
and
failed
repeatedly,
so
yeah
I
think
Tim
all
Claire's
PR
to
drop
someone
support
is
the
responsible
thing
to
do
and
likely
does
not
impact
many
use
cases.
I
have
never
actually
seen
someone
using
control-c
P
to
pull
out
things
depending
on
some
link
structure.
A
A
We
can
obviously
we
can
extend
the
help
command.
Currently
the
help
command
for
cube
City
okapi
curly
states
that
you
need
to
have
thar
in
the
container
right.
Next
to
it,
we
could
say
that
the
current
functionality
of
keep
CTO
supports
only
regular
files.
If
you
want
to
work
with
something
more,
if
you
need,
if
you
need
to
cover
advanced
use,
cases
here
are
cubes
ETL,
exact
alternatives
that
you
might
consider.
D
C
C
A
E
C
E
A
A
B
Some
quick
context
on
that
that
particular
feature
I
asked
the
the
creator
to
to
present
it
to
us
to
actually
try
to
convince
us
that
that
was
something
that
was
needed
and
useful.
Brian
grant
gave
him
pointers
to
about
six
other
issues
and
it's
and
he
ended
up
realizing
that
it
was
a
lot
more
complicated
than
he
realized
and
and
so
it
to
be
honest,
that
sounds
like
that's
something
that
might
may
have
been
dropped
after
Brian
weighed
in
on
it.
A
Okay,
I'll
get
the
order,
one
more
chance
to
shop
on
the
next
call
and
if
he
does
not
I'm
gonna
drop.
The
topic
entirely.
I,
don't
notice
that
the
community
poll
is
adding
a
proposal,
but
the
proposal
is
should
be
turned
into
a
cap
against
enhancements
repository.
If
you
got
a
chance
to
talk
with
him,
I'll
probably
leave
a
comment
in
a
PR
asking
him
to
move
over
to
cap
so
that
we
can
continue
the
usual
enhancements
proposal
process.
A
F
It's
still
early
days
for
this
future,
but
I
wanted
to
start
a
conversation
with
six
CLI
about
how
this
could
be
used
in
coop
control
to
to
sort
of
improve
options
for
debugging.
This
specific
use
case
that
I'm
that
I'm
targeting
here
is
ability
to
troubleshoot
Australis
containers,
so
I
linked
a
little
demo
there
you
can
either
I,
don't
know
if
we'd
be
better
to
sort
of
play
it
now
or
to
just
let
everyone
take
a
look
at
it
on
their
own.
So
what
do
you
think.
F
F
F
F
F
F
F
F
You
could
use
it
to
run
an
agent
inside
the
container
right
like
if
you
wanted
to
run
at
are
listening
to
standard
in
or
something
like
like.
If
tar
was,
it
didn't
already
exist
in
the
container,
that's
possible,
though
it
would
be
a
little
bit
tricky
because
of
the
way
that
the
amount
namespaces
work.
B
Solely
just
a
quick
update,
so
so
the
the
actual
ephemeral
containers
in
in
core
v1
type
stock
go
that
wasn't
merged
until
one.
Sixteen,
that's
correct,
okay,
so
there
there
there
may
be
an
issue,
then,
with
any
of
the
code
in
one
sixteen
for
coop
cuddle,
that's
referencing
that
are,
we
have
a
support
window,
a
vergence
queue,
a
policy
that
allows
us
to
understand,
plus
and
minus
one
release,
and
so
so
I
may
have
to
revisit.
Some
of
those
reviews.
E
B
E
F
You
know
just
the
only
thing
that
we
changed
in
in
116
for
coop
control
is
there
are
a
couple
of
sections
that
iterate
through
the
containers
like
attach
and
one
of
the
the
logs,
and
so
we
had
to
update
those
two
also
iterate
the
affirm,
all
containers
and
then
I
added
something
to
print
out.
The
ephemeral
containers
in
coop
control
can
just
described,
but
that's
all
we've
changed
so
far.
F
D
F
Talk
to
you
about
that,
so
they,
the
sick
dogs,
asked
that
I
document
this,
including
a
task
of
how
to
use
it
and
they're.
Okay,
I
just
published
this:
the
prototype,
plugin
and
they're
okay,
with
just
referring
to
that
under
my
personal
github.
If
we're
going
to
do
it
in
the
official
documentation,
I
kind
of
prefer
it
be
livin
like
a
kubernetes,
saegusa
repo
somewhere,
either
in
CLI
experimental
or
a
new
repo,
so
asking
opinions
on
that.
He.
C
F
C
C
F
C
Experimental
trying
getting
a
plug-in
first,
just
so
it's
available
to
everyone,
the
to
control
libraries
send
headers
that,
like
record,
which
command
is
being
run
so
we
can,
it
doesn't
include
any
API
or
any
values
of
flags
or
arguments,
but
it
doesn't
man.
So
we
should
check.
If
this
is
a,
you
should
check.
If
you
are
doing
that
in
your
pumpkin
as
well.
C
A
Ideally,
we
could
have
a
metric
inside
the
API
server
that
it
will
expose
that
information
that
we
can
then
collect
and
be
able
to.
You
know
provide
that
kind
of
metrics
based
on
the
actual
usage
yeah.
F
A
G
Go
with
her,
so
you
recently
added
a
security
notice
about
untrusted
plugins,
as
we
discussed
before
and
we're
getting
towards
like
his
0.3
release.
We
had
our
last
release
and
also
we're
actually
even
working
for
reliably,
since
they
need
to
make
a
release
and
no
other
real
life.
So
that's
good
and
the
centralized
and
plug-in
makes
is
growing.
You
could
like
over
50
plugins.
H
H
C
Ok
thanks
custom
ice,
so
they're
all
just
had
one
thing
something
I've
been
talking
about
is
being
able
to
do
higher
fidelity
processing
with
gamma
files.
Where
we
keep
comments,
keep
file
structure
can
write.
If
you
build
it,
it's
possible
to
write
it
back
to
a
set
of
files
and
retain
the
original
names
and
ordering
of
all
the
elements,
and
that
sort
of
things
so
does.
C
I
B
So
so
not
much
has
changed.
We,
as
I've
mentioned
before,
we've
got
about
95%
of
the
go
code
into
staging.
The
three
remaining
commands
are
get
convert
an
auth
reconcile
that
have
dependencies
on
kubernetes
core
that
still
need
to
be
worked
on
and
with
with
Jordans
help,
I've
been
able
to
continue
to
do
some
refactorings
to
get
the
two
so
that
the
git
command
can
be
the
dependency
and
the
git
command
can
be
removed,
we're
actually
pretty
close
to
that
one,
and
but
it
won't
land
for
1/16.
B
In
fact,
one
of
those
is
going
to
be
reusing,
some
of
the
coop
cuddle
code
that
modifies
the
coop
config
file,
so
there's
other
plugins,
for
example,
that
might
want
to
add
entries
within
the
coop
config
file.
We
want
to
make
that
a
clean
interface
so
that
that's
easy
to
do.
Okay,
that
was
long-winded
any.
A
Any
questions
awesome,
I'm,
I'm,
very
happy
about
the
most
with
keep
CTO
and
the
fact
that
people
are
working
towards
reusability
of
the
elements
up
to
cube
C
do
with
regards
to
config
manipulation,
I'm
inclined
to
say
that
this
kind
of
functionality
could
even
find
create
its
way
towards
CLI
runtime.
Yes,
yes,
rather
core
type
of
functionality
to
have
people
writing
C
lies
for
keep
co-op
could
have
reuse,
yeah.