►
From YouTube: Kubernetes SIG CLI 20161221
Description
Kubernetes SIG-CLI planned the Roadmap for 2017 and priorities for Q1.
B
A
Yeah
in
terms
of
ramping
up
I
think
we
have
a
github
label.
Hulk
wanted,
or
something
like
that.
I
could
I
could
make
sure
we
have
a
number
of
our
issues
tagged
appropriately,
14
flat,
like
would
be
nice
starting
points
for
people
ramping
up.
So
that's
one
of
the
ideas
in
case
you
want
to,
you
know,
give
them
some
issues.
Using
label
would
be
nice.
B
B
B
A
B
The
first
part
is
around
just
documentation:
I
have
an
item
here,
I
listed
as
just
getting
the
commands
logically
grouped
on
how
we
want
them,
and
there's
I've
been
thinking
about
this
more
since
we
first
discussed
it
and
we're
finding
division.
I
think
this
is
pretty
small
task.
Actually,
it's
more
sense
of
examples
in
productivity
hacks.
One
thing
I've
seen
on
issues
is
a
request
for
a
feature
that
that
is
like
I
wish.
Could
control
could
do
this
and
then
we
get
it
either
it
approved
or
halfway
through
min
implementation
and
someone
who's.
B
Just
a
really
good
hacker
on
coop
controls
says:
oh,
did
you
know
you
can
combine
these
two
flags
and
pipe
it
to
this
other
command,
and
then
you
get
that
same
creature.
These.
These
are
things
I'd
wasn't
even
aware
of
or
hadn't
considered,
so
having
someone
who's
familiar
with
all
those
hacks,
or
maybe
it's
just
collectively,
sharing
some
of
our
our
productivity
things
we
do
and
documenting
them
might
be
a
way
to
get
them
into
people's
hands.
B
B
This
this
is
mainly
derived
I
was
thinking.
Documentation
is
thinking,
documentation
more
in
the
form
of
like
tutorial
focused
unless
reference
focused,
we're
assuming
user
comes
to
ku
control
without
any
knowledge
about
what
the
tool
does
or
how
to
manage
your
configs.
We
have
some
documentation
around
this
on
the
website
already,
but
I
I
think
as
the
tools
kind
of
evolved
in
our
vision
of
it
is
evolved.
Maybe
we
want
to
restructure
a
backyard,
especially
with
the
apply
stuff,
since
that
didn't
exist.
B
Originally
in
the
first
iteration
of
KU
control,
the
onboarding
documentation
doesn't
do
a
great
job
of
essentially
taking
that
into
account
and
explaining
the
differences.
Someone
that's
stuck
in
design,
developer,
Docs,
I
think
this
is
really
just
like
reevaluating
kind
of
all,
how
we
introduce
users
the
tool
and
then
yes,
something
about
linking
to
online
docs.
We
need
to
think
more
about
how
that
might
be
done.
If
we
do
it
I'll,
actually
I'll
try
and
stick
to
is
just
the
top-level
stop
in
the
comments.
B
Well
documented
trade-offs
when
multiple
approaches
to
solving
a
problem.
This
stuff
here
is
like:
how
can
how
can
users
factor
their
config
right?
Do
they
do
the
storm
in
github?
Do
they
storm
locally?
Do
they
just
copy
and
paste
the
same
thing
over
and
over
again
or
do
they
use
like
Jason
net,
to
try
and
use
inheritance
which
box
is
used
well,
but
there's
a
lot
of
manages
to
it.
A
Yeah
in
terms
of
in
the
tank
team
of
user
onboarding,
and
things
like
that,
it's
related
to
D,
provide
holistic,
well
true
item,
but
I
see
that
not
only
as
a
point
in
terms
of
documentation,
but
also
probably
in
line
walk
to
write
because
sometimes
sometimes
I
feel
like
well
keep
CTL
is,
is
always
related
to
a
context
right
when
you
are
using
Kip's
TTL.
You
are
in
a
namespace
in
a
given
server
with
a
given
session,
so
everything
you
ran
and
cube.
A
Ctl
is
a
little
related
to
a
can,
take
context
and
I
feel
that
sometimes
we
we
don't
do
a
great
job.
You
know
for
people
to
understand
exactly
where
they
are
at
some
point
and
where
they
can
go
next
right.
So
for
me,
when
I'm
running
a
number
of
cubes
CTL
command,
sometimes
I
get
lost
well,
where
exactly
I
am
I
am
in
this
given
name
space
and
if
I'm
a
new
user,
I
just
did
my
first
keep
see
children
or
my
first
keeps
tell
create
okay.
C
B
Right
great
unya,
as
they
go
through
this
just
think
about
that
I
guess
the
plan
is
going
to
be
like.
Why
didn't
you
just
know
any
that
you
think
are
important
for
q1
or
maybe
like
we
can
each
rank
or
top
three
at
the
end,
and
then
that
way
after
I
go
through
all
these,
we
can
just
quickly
kind
of
figure
out
which
ones
we
want
to
nail
for
16
all
right,
so
theme,
engineering
velocity.
B
This
is
essentially
technical
debt,
but
also
includes
stuff
that
maybe
isn't
like
technical
debt,
but
just
things
we
could
do
to
make
us
a
faster
and
more
effective
Karen.
B
B
Yeah,
so
one
of
these
one
of
the
things
I
think,
is
a
great
there's.
A
proposal
about
this
I
think
I.
Think
a
lot
of
this
stuff
is
just
deciding
what
we
want
to
do:
I
mean
within
the
whole
set
of
requirements
or
whole
set
of
goals.
I
think
a
broader
meta
theme
here
is
like
dr.,
not
necessarily
fixing
anything
or
adding
anything
but
just
deciding
what
we
want
people
to
do,
because
there's
a
couple
ways
of
solving
a
lot
of
problems.
B
So
yeah
distribution
distributing
multiple
tools
for
developing
a
set
of
client
tools:
khoob
admin
to
control,
there's
a
federation,
client
tool.
I
think
this
list
is
just
going
to
grow.
You
know
linearly
over
the
set
of
releases,
so
we
need
to
figure
out
a
packaging
mechanism,
and
that
may
include
stuff
like
how
do
we
package
and
allow
users
to
install
tools
that
may
not
be
distributed
with
core
but
are
considered
tied
to
the
project
very
closely,
such
as
hell
more
composed
without
stripping.
B
Moving
some
of
the
main
repo
is
what
I
think
is
really
important
from
the
perspective
that
it's
going
to
allow
us
a
lot
more
control
over
how
we
develop
the
product.
But
it's
a
lot
of
work.
You're
doing
some
research
on
good
kid,
sub-modules
right
now
on
sub
trees,
which
I
don't
know
that
much
about
but
might
be
able
to
help
us
or
not.
B
Okay,
common
client
libraries
factored
out
client
tools
for
shared
infrastructure,
so
this
could
be
a
couple
different
things.
Ideas
I
was
thinking
here
include.
So
what
is
flags
I
know?
This
seems
like
kind
of
a
silly
thing,
but
just
like
having
the
set
of
flags
that
we
think
might
be
common
for
tooling,
back
room
doubts
you
can
just
inherit
them
might
be
an
interesting
idea
that
way
as
you
switch
between
tools,
the
dry
run
flag
is
always
consistent
and
it
like
always
dry
dash,
run
not
like
no
dash
or
something
like
that
right.
B
That
one's,
like
seems
really
minor
and
easy
to
do,
but
might
be
helpful
for
folks
being
started.
All
there
is
like
we
have
infrastructure
built
up
around
like
visiting
all
the
files,
for
instance,
as
we,
if
you
do,
directory,
apply
on
a
directory
or
something
like
that.
If
infrastructure
around
just
how
you
are
busy
config
files,
how
you
parse
them,
we
have
library
set
up
around
authorization
and
authentication
config
stuff,
we're
probably
going
to
have
some
stuff
set
up
around.
B
Can
preferences
and
preference
files,
so
I
imagine
kind
of
what
I'd
like
to
see
is
if
we
develop
clients
like
khoob
admin
and
the
Federation
stuff,
that
the
way
that
you
manage
preferences
for
those
guys
is
the
same
to
the
extent
possible
same
syntax,
maybe
even
like
similar
same
directory
in
just
a
different
dot
preferences
file
or
perfect,
like
cube
admin,
dog
preferences,
or
something
like
that.
I.
A
Was
thinking
am
even
something
related
to
validations,
because
we
currently
don't
don't
see
them
and
have
that
inside
keeps
itself
right,
because
every
command
does,
for
example,
flags
and
arguments
validation
a
different
way,
so
Jose
shin
should
also
be
a
target.
We
could
start
by
having
something
side
cubes,
to
tell
some
library
that
validates
like
mutually
exclusive
Flags
require
arguments.
Things
like
that
in
make
the
actual
validation
and
also
our
message
messages
consistent,
and
then
we
strike
that
as
the
library.
B
Is
like
we
going
back
to
the
theme
of
just
documenting
how
we
recommend
people
do
things
even
within
coop
control
itself?
We
have
this
notion
of
subcommands
versus
top-level
commands
and
you
can
either
create
a
sub
command
or
you
can
create
from
a
top
level
command
where
you
can
add
a
flag
to
a
top-level
command
to
get
the
same
thing
right,
which
is
to
control
behavior
of
that
command.
B
I've
seen
some
e
example
of
this,
where
we
do
something
the
same
thing,
two
different
ways
as
we
have
could
control
run
then
dash
dash,
like
generator,
which
allows
you
to
do
like
job
versus
deployment
versus
pod
or
something
else,
and
then
each
one
of
those
has
their
own
set
of
mutually
exclusive
flags.
That's
how
I
personally,
this
is
tough,
where
you
can
all
these
job
flags.
B
If
using
the
job
generator
sort
of
thing,
then
you
have
could
control,
create
job
to
control,
create
deployment
which
allows
so
commands
instead
of
place
to
control
them,
and
then
now
you
actually
don't
have
that
mutually
exclusive
thing
and
all
the
flags
become
much
more
discoverable,
because
you
know
what
all
the
job
flags
are.
When
you
do,
control
help
create
job
or
whatever
it
is
so
again,
just
just
having
better
guidance
around
when
you
sub
transverses,
adding
plagues.
B
Cool
Tony,
that's
cool
that
this
is
something
freaking
cute.
One
I
think
I
think
this
would
actually
be
really
helpful
and
powerful
updating
the
roadmap
yeah.
Oh
my
gosh
yeah.
How
do
we
do
this
stuff?
There's?
Also
there's
also
all
projects
github
has
like
projects
for
managing
these
things.
Do
we
want
to
move
the
roadmap
some
of
the
stuff
into
a
project,
get
a
project
which
allows
you
to
like
mark
items
off
automatically
but
doesn't
allow
you
to
do
cross
repo
tracking.
D
B
You,
oh
man,
that
was
something
I
just
I've
been
wishing
for
every
morning
when
I
wake
up.
So
that's
awesome.
Okay,
yeah
santa
came
early
with
your
apparently
a
couple
months
ago
and
I
just
didn't
realize
it.
Oh
great
thanks,
Erin,
that's
huge
I
have
some
work
to
do
later
today:
okay,
moving
on
to
Application
Lifecycle
Management.
So
this
is
actually
probably
the
more
feature-rich
piece
see.
Also
clue.
Control
does
number
different
things
that
are
in
its
wheelhouse.
It
sets
up
configuration
stuff
for
you.
B
Think
part
of
this.
So
the
first
item
I
have
here
is
actually
not
about
to
control
itself
but
like
how
does
coop
control
fit
in
with
the
other
tools
that
are
needed
right,
because
coop
we
can
all
use
to
control
to
go
launch
a
deployment
on
keroon
a
days,
but
that's
typically
not
sufficient
to
run
a
organization
you
need
like
see.
B
Icd
may
possibly
write
some
folks
do
that
you
need
a
dev
ops
team
that
is
managing
this,
so
I
think
some
of
the
things
here
are,
if
you're
using
coop
control
to
like
you,
control
with
a
config
based
mode.
How
do
you
initialize
those
configs
right?
Do
you
just
write
them
from
the
ground
up
I
know
like
internally
at
Google
the
way
we
initialize
configs?
B
Typically
as
we
go
by
someone
who
solved
the
problem
before,
like
my
have
a
Java
app
I
find
someone
who's
written
a
config
for
java
app
the
copy
and
paste
the
whole
thing
into
my
directory.
Half
of
the
stuff
I
don't
understand
what
it
does
so
I
just
leave
it
there,
and
then
you
know
bang
on
it
with
a
hammer
enough
to
make
it
work.
That's
probably
more
true
in
adjacent
net
world,
where
you
have
inheritance
and
things
are
a
lot
more
complicated
than
what
it
is
with
flat
configs,
but
I
still
think
we.
B
It
is
food
control
that,
like
create
deployment,
dash,
def
image,
dash,
dash
replica
count
and
then
do
dash
dash
dry,
run
and
dash
0
gamal,
pipe
it
to
a
file,
and
that
actually
creates
the
config
for
you
and
it's
not
perfect,
because
it
defaults
some
fields
that
you
might
not
want
defaulted
in
the
config,
but
it
does
a
pretty
good
job
and
it's
actually
pretty
cool
and
I'm
not
sure
if
we
tell
people
that
that's
even
an
option.
This
is
this
is
a
big
task
for
me.
B
A
I
definitely
agree
about
being
a
big
desk
and
even
for
q1
what
kind?
What
kind
of
methodology
could
you
use
in
terms
of
understanding
from
our
customers,
because
I'm
not
I'm
not
sure
if
you
guys
have
google
have
a
established
methodology,
I
already
saw
feel
doing
a
few
times
some
kind
of
fall
with
users,
but
would
that
be
something
like
that?
Or
do
you
have
anything
else
in
mind?
Oh.
B
Fantastic
question
I
mean
this
is
something
I'm
trying
to
figure
out
like
naively
what
I,
what
I'd
like
to
do
is
see
if
I
can
sit
down
with
some
of
a
diverse
set
of
customers.
I'm
thinking
between
three
and
five
over
the
quarter
for
like
days
with
dev,
ops,
team
and
I,
know
how
much
I
get
to
really
look
at
like
their
code
right,
because
you
know
it's
all
private
but
to
the
extent
possible,
I'd
like
to
sit
down
and
actually
watch
like
box,
do
a
push
yeah
I
mean,
and
I
liked
it.
B
A
Ideas:
yeah
now
that
that's
absolutely
nice
yeah.
It
would
be
awesome
if
we
could
do
that
with
unaware
of
customers.
We
here
in
openshift
we
a
couple
months
months
ago
with
these
did
something
similar
with
some
customers
in
Europe,
but
it
was
much
more
focused
on
our
web
dashboard,
but
you
know
it
would
be
something
similar
for
the
CLI
and
if
we
could,
you
know,
understand
the
process
for
a
company
like
that
yeah.
That
would
be
awesome.
A
B
I'd
love
to
know
offline
I'd
love
to
talk
to
you
just
about
how
about
you
will
you
found
that
maybe
just
impossible?
I
can
take
your
learnings
from
how
to
run
that
nah
yeah,
so
I
need
to
come
up
with
a
plan,
but
I
need
to
at
least
have
a
planning
to
one.
This
is
really
not
well
defined
right
now,
even
how
to
go
about
approaching
this
from
threats
this
stuff.
There's
a
lot
of
stuff
in
here.
B
I
skipped
a
couple
things
but,
like
you,
control
apply,
we
know
it's
broken
in
just
certain
ways.
We
have
a
list
of
issues
like
part
of
a
polite
cycle
management
is
making
sure
that
the
tools
work
and
do
what
they're
supposed
to
do
and
then
the
config
factoring
I
think
I
talk
to
me
a
little
bit
about
that
earlier.
B
I'd
like
to
do
some
research
here
as
part
of
that
first
thing
about
how
people
are
doing
that
the
command
based
management,
I.
Think
of
is
like
there's
kind
of
three
ways:
I
think
of
goop
control,
managing
apps,
there's
command
based,
that's
the
crate
set
for
you
run,
exposed
set
of
commands.
Then
there's
the
config
based
imperative
commands,
which
is
create,
delete,
update
or
replace
I'm
config
files,
and
then
the
apply
and
I
think
the
imperative
can
fake
is
pretty
worked
pretty
well.
B
Client
version
skew
retains
correct,
behavior,
Wow,
28,
q1,
yeah,
I.
Think
that
sounds
good
to
me.
Okay,
I!
Don't
know
you
probably
have
a
better
idea
of
how
much
work
this
is
than
I.
Do
getting
think.
There's
a
couple
pieces
to
this.
D
B
I
can
talk
yourself.
I
have
some
idea
on
how
to
do
some
of
that
stuff,
but
I,
probably
medi.
This
person
will
have
to
talk
to
you.
The
other
piece
is
going
to
be
stuff
that
just
breaks
I
know
why
we
keep
seeing
this,
but
there's
some
legacy.
Logic,
that's
just
baked
into
assumptions
about
how
stuff
works
in
the
server
one
of
these
that's
going
to
break
again
for
16.
B
B
B
C
B
B
B
B
Just
might
be
a
stretch,
no
Erin,
I
I,
don't
think
so.
Man.
D
B
C
B
Yeah
so
in
Tony's
comment
there
is
a
proposal
about
the
coop
control
extension
keys.
That's
not
targeted
at
the
at
the
Federated
API
resources.
It
looks
like
both.
Those
comments
are
targeted
just
at
the
general
extensions
I.
Don't
think
I'm
not
aware
of
any
issues
about
to
control
working
with
federated
the
API.
Okay.
B
So,
for
instance,
if
you
have
like
a
Redis
master
and
the
reddest
nodes-
and
you
do
if
you
do
a
Google
get
apps
and
it
would
show
you
that
is
one
application
composed
of
multiple
controllers,
that
sort
of
thing
and
that
this
would
extend
to
the
UI,
and
so
you
have
kind
of
this
native
feel
across
both,
and
this
has
to
do
with
the
extensions
because
third-party
resources
may
so
like
for
Cassandra.
For
instance,
I
think
you
could
create
a
third-party
resource
that
then
has
a
controller
that
uses
multiple
different.
B
Other
controllers,
like
deployments
kind
of
like
I
described,
credits
and
you'd,
want
to
be
able
to
expose
that
okay
almost
done
to
control
as
a
platform.
This
goes
back
to
kind
of
some
of
the
engineering
velocity
pieces,
but
is
essentially.
This
is
automatically.
We've
talked
about
user
education,
engineering,
velocity
and
managing
the
applications,
sending
it
the
cluster
itself,
and
this
goes
to
kind
of
whatever,
like
the
infrastructure,
could
control
provides
that's
not
business
logic
for
specific
commands.
How
do
we
support
that.
A
In
terms
of
Lee,
of
removing
the
dependency
of
Cobra,
I
have
a
couple
for
requests
or
issues
from
some
external
contributors
that
some
shepherding
they're,
both
stale
forever.
Probably
I
didn't
hear
from
her
from
them
for
a
few
weeks,
but
there
there
is
some
work
started
in
that
sense
already,
so
removing
dependency
of
Cobra
in
making
the
actual
command
execution
separate
from
from
the
actual
commander.
Cobra
I'll
have
to
take
a
look
at
those
pr's
again
see
if
they
were
able
to
do
any
progress
on
that.
But
anyway,
if
somehow
started
already.
B
B
C
B
Let's
start:
is
there
a
design
doc
for
that
or
we
already
discussed?
How
are
gonna
do
that
I
can't
remember
yeah.
B
Cool
yeah
after
the
meeting
maybe
go
up
and
just
add,
clarify
like
what
specific
issues
just
on
the
dock
by
adding
printers
and
the
link
tissue
design,
technical
debt,
yay
Fabio,
actually
making
our
issues
meaningful.
That
sounds
wonderful
stabilize.
What
does
stabilize
coop
control
apply
is
that
does
that
go
back
to
that
life
cycle
management,
about
yeah.
B
A
I
added
that
one,
it's
related
to
number
la
issue
I
had
so
if
there
are
a
large
number
of
drinking
and
describers
issues,
so
I
add
that
there
in
the
sense
of
well,
we
should
probably
try
to
fix
some
of
them
because
they're
just
too
too
many
of
them.
On
the
other
hand,
I'm
not
sure
about
the
work
from
moving
printers
to
the
server
side.
If
the
actual
code
base
would
be
reused
entirely.
A
B
A
Absolutely
there
are
some
really
minor
ones.
There
are
some
that
required
decisions
like
things
like
we
not
be
completely
compatible
with
emma
1.2.
We
take
some
pieces
from
yema
1.1,
so
there
are
some
things
that
require
decisions,
but
on
the
other
hand,
there
are
a
bunch
of
issues
that
really
minor
stuff
minor
creases.
To
do.
A
B
Yeah,
that
seems
like
a
good
starting
point.
Just
getting
the
number
down
will
allow
us
to
focus
like,
even
if
they're,
like
simple
fixes
that
are
hugely
impactful.
Mr.,
the
noise
yeah.
A
Absolutely
I
added
key
one
to
that
topic
exactly
in
that
sense.
You
know
I
would
like,
for
you
want
to
see
that
number
go
down
at
least
a
little
bit.
You
know,
or
at
least
we
could
keep
the
face
of
not
having
a
lot
of
stuff
still
being
added,
and
you
know
not
come
to
the
end
of
q1
and
have
like
1000
of
them.
So
we
should
illustrate
to
improve
a
little
bit
number.
B
Okay
right
so
we've
gone
through
the
full
list.
We
have
a
number
of
different
items,
a
marked
I
guess.
The
second
piece
of
this
is
like:
how
firm
do
you
want
to
make
these
commitments?
Are
these
just
ambitions
or,
like
things
we'd
like
to
work
on
and
we'll
get
some
of
these
done,
but
not
follow
them
or
are
these
like
at
the
end
of
the
quarter?
We
fully
expect
everything
that
says
you
want
to
be
able
to
be
crossed
off
this
list.
At
least
some
section
of
it.
A
B
B
Suspect
that
kind
of
with
what
we've
called
out
on
here,
some
of
these
items,
we
won't
even
like
start
q1.
It's
kind
of
my
gut
feeling.
It
just
seems
like
there's
a
lot
of
stuff
which
might
be
fine,
and
maybe
these
are
like
some
of
these
items
are
the
things
that
we
say.
If
we
get
new
community
members
will
stick
them
on
these
right,
like
especially
big,
create
view
commands.
That
seems
like
really
straightforward
one.
B
Maybe
maybe
we
should
just
call
out
like
of
the
items
that
we
have
marked
is
q1,
which
ones
are
we
going
to
say
like
we
fully
intend
to
have
at
least
started
this
and
made
progress
on
it
in
q1
vs?
These
are
things
we're
going
to
tell
other
people
to
focus
on.
A
Yeah
personally,
I
feel
I
feel
ok
with
pretty
much
all
we
marked
as
key
one
except
probably
turbo,
researches,
stuff,
I'm,
not
sure
and
specially
curated
API
researchers,
I'm,
not
very
free,
will
be
ready
to
move.
You
know
anything
comes
up
of
those
in
q1,
but
a
part
that
I
think
it's
plea:
okay,
okay,
no.
B
C
I
think
for
third-party
resource
we
may
not
focus
on
coop
control,
only
first
quarter
when
I
first
I'd
some
suicide
implementation,
and
then
we
can
go
always
q
control
part
of
certain
resource
so
generally
as
ink.
Well,
we
can
live
a
lot
touch
with
adding
cube
control
side,
support
for
certain
resource.
B
That
sounds
good
to
me,
so
I'm
just
going
to
mark
both
of
those
two.
That's
a
really
good
point
fabio
and
I
didn't
realize
I
think
that's
what
I
was
thinking,
but
those
two
are
potentially
a
lot
of
work,
but
if
we
just
start
them
and
start
scoping
the
work
in
deciding
what
needs
to
be
done,
that's
probably
a
good
goal
for
q1
and
then
we
can
revisit
them
in
1.7
or
q2
yeah.
A
B
Cool,
ok
and
then
Aaron.
One
last
thing:
do
you
is
there
anything
on
here
or
what's
your
plan
for
q1?
Are
you
planning
on
working
on
any
of
these
items?
Would
you
like
us
to
involve
you
in
anything
specifically
I.
D
D
Mostly
I'm
coming
to
this
from
the
tpr
world,
I
want
to
see
where
I
can
pitch
in
on
getting
tprs
better
represented
in
cube,
see
Tito
at
this
point,
but
this
doc
has
illuminated
a
lot
of
other
stuff
that
I
need
to
gather
my
thoughts
on
and
and
I'll
hop
into
the
slack
room
and
dump
my
thoughts.
I
guess:
awesome!
D
B
Cool
so
then
yeah,
presumably
since
tony,
was
looking
at
the
tpr,
no
wait.
Nevermind
did
that
kid
strike
now.
It's
here
sure
apply
works
with
tpr
and
federated
api
resource
yeah.
Since
you
guys
will
probably
talk
about
that.
At
some
point
I
assume
yeah.