►
From YouTube: WG API Expression Bi-Weekly Meeting for 20200721
Description
WG API Expression Bi-Weekly Meeting for 20200721
A
There
we
go
welcome
everyone
to
working
group,
api
expression.
We
are
tuesday
july
21st
2020
and
we're
gonna
talk
about
what
we're
doing
do.
We
have
any
topic
in
the
agenda.
I
can
only
see
two
and
I
can't
see
that
person
I
wouldn't
try
to
pronounce
that
name.
B
A
B
A
I
think
we've
done
that.
So
that's
good.
A
And
I
yeah,
I
don't
know
what
this
is.
Do
you
know
what
this.
B
Is
I
think
we
have
to
to
report
our
status
at
the
community
meeting?
At
some
point?
B
I
didn't
click
the
link
so
far,
but
that's
what
I
I
at
least
heard
they
went
to
annual
reports
all
right,
yeah.
I
think
that's
something
we
do
have
to
to
work.
A
Out
all
looking
yeah,
my
internal,
my
internal
filter,
considered
this
was
spam.
I
remember
doing
that
and
thinking
I'm
not
interested
in
that,
but
it
looks
like
I
was
wrong.
Are
we
supposed
to
do
something?
It's.
B
Yeah
now
I
don't
think
so.
I
I'll
I'll
read
over
it
and
we'll
just
discuss
I'll.
Tell
you
talk
to
you
easiest,
okay,.
B
A
Yep,
okay:
is
there
anything
else
that
we
want
to
talk
about
that
we
didn't
have
in
the
in
the
agenda.
B
A
Okay,
cool
yeah,
I
mean,
if
you,
if
anything
you
want
to
talk
about
at
any
time.
Please
interrupt
me.
I
want
to
start
by
mentioning
that
this
task
is
now
over.
So
thank
you
so
much
to
julian,
because
we
have
a
much
smoother
transition
between
client
side
and
server
side
transition
apply.
A
So
I
think
that's
a
great
progress.
It's
going
to
be
amazing
for
the
next
release.
Yeah,
that's
great,
and
I
also
want
to
acknowledge
that
we
have
done
that.
Thank
you,
joe,
because
again
this
is
going
to
be
improving.
The
experience
significantly
and
joe
has
also
done
the
documentation
for
that.
So
this
is
definitely
completely
great.
C
Yeah,
it
took
us
a
while
to
phrase
it,
so
it
might
be
worth
a
read.
A
That
learning
there
is,
there
are
a
few
things
that
I
thought
about
this
week.
Discussions
we've
had
with
with
julian,
which
are
not
in
this
document,
yet
I'm
probably
gonna
put
them
in
the
misc,
but
okay.
So
we
are
not
supposed
to
track
the
fields
on
objects
that
no
that
don't
have
any
manage
field
and
are
not
being
created.
A
So
in
other
words,
if
you,
if
you
update
an
object
and
it
doesn't
have
the
fields,
then
we're
supposed
to
keep
it
that
way,
not
not
starting
to
track
them
and
that's
important
for
people
who
are
converting
from
a
older
version
of
the
cluster
to
a
newer
version.
A
We
don't
want
to
suddenly
start
tracking
fields
and
existing
big
objects
and
deal
with
the
consequences
of
that.
But
I
do
not
know
if
this
works
today
and
I
am
not
convinced.
E
Yeah,
I
think
initially
it
it
does
seem
to
work
yeah.
I
can
link
in
that
pr.
A
Yeah
right,
yes,
absolutely,
but
it's
it's
hard,
obviously
to
create
a
test
that
is
going
to
say
what
happens
after
an
update
of
the
cluster,
because
we
don't
have
a
good
way
to
update
the
clusters,
especially
in
the
integration
test.
A
So
we
don't
have
a
test
that
shows
that
this
works
and
also
going
to
lead
to
that.
We
have
no
mechanism
to
stop
tracking
fields
and
specific
objects,
because
if
you
send
a
null
manage
field,
we're
going
to
use
the
one
from
the
live
object.
If
you
send
an
empty
one,
we're
going
to
reset
it
reset,
but
not
stop
tracking.
So
julian
and
I
were
discussing
the
idea
of
having
a
third
mechanism
to
say,
I
want
to
stop
tracking
the
field
which
would
allow
us
to
make
tests,
because
now
we
would
say
hey.
A
I
want
to
start
tracking,
and
now
we
have
the
stop.
The
we
don't
have
manage
fields
anymore
and
we
can
see
how
grabbing
back
the
fields
is
going
to
work
or
starting
to
manage.
The
fields
again
is
going
to
work,
and
this
is
because
we
have
a
back
somewhere
with
someone
complaining
about
the
b4
file,
supply
manager
and
again
it's
hard
to
test,
because
we
don't
have
a
mechanism
to
stop
tracking
and
then
stop
tracking
again.
A
A
A
B
B
Along
with
us,
but
and
I'm
cooking
options.
A
B
B
A
And
I
suspect
we
could
write
the
test
in
a
different
way
so
that
we
can
test
no
matter
what,
while
still
having
the
thing.
Okay,
yeah,
that's
a
good
question.
E
Yeah,
I
think
that
makes
sense.
I
can
give
that
a
try.
A
Okay,
and
basically,
where
am
I
going
with
that
so
today,
when
you
started
playing
the
first
time,
you're
gonna
get
conflict
with
whatever
was
applied
or
whatever
was
set
before
you,
but
because
we
have
this
client
side
to
serve
side
migration
flow.
Now
we
can
actually
avoid
the
conflicts.
The
first
time
you
apply
an
object,
which
I
think
is
great
we're
pretty
much
using
the
last
applied
annotation
for
cook
at
all
to
decide
whether
or
not
you
should
have.
A
Cool,
let's
start
from
scratch.
We
have
jenny
here.
A
No,
we
don't
have
jenny,
I
hey
julian,
do
you
know
how
much
time
he'll
have
to
work
with
us.
E
Yeah,
I
could
do
you
need
help
taking
a
look
at
this
or.
A
And
I
talk
to
jenny
and
see
if
she
has
anything
that
she
can
share
with
you
or
hand
off
or
whatever.
Oh,
if
you
want
to
work
with
you
on
that,
I
I'll
talk
to
her
okay
status.
B
Wiping
what's
up
with
that
kevin
yeah,
I'm
I'm
continuing
to
add
more
strategies
and
search
implementations.
I
had
a
weird
one
today
with
demon
sets
where
I
have
a
strategy
and
it
didn't
get.
It's
still
got
a
conflict
with
status.
So
I'm
looking
into
that
weird
I
don't
know
what's
going
on,
but
in
general.
B
The
list
of
remaining
strategies
that
make
it
make
the
test
fail
are
fairly
is
is
very
small.
I'm
still
not
sure
if
I
should
add
strategies
for
all
internal
types,
or
only
if
we
have
a
saving
test,
feel
free
to
continue
the
direction
there,
but
in
general,
I'm
adding
strategies,
and
it's.
A
That
we
do
checkpoints
on
this
as
often
as
possible,
like
especially
when
you
have
a
doubt
like
whenever
you
have
a
doubt,
just
create
the
vr,
make
sure
that
we
all
agree
on
how
to
solve
this
specific
problem
before
you
have
to
generalize
or
like
apply
the
same
pattern
in
like
15
different
places.
Just
so.
A
Waste
any
of
our
time,
I
I
feel
deeply
sorry
that
you've
had
to
spend
so
much
time
iterating
on
these
100
times,
yeah.
B
A
I
just
want
to
avoid
that
as
much
as
possible.
From
now.
E
B
One
specifically
is
the
the
pot
template
out
of
the
disable
feed
for
parts
from
the
pod
util
yeah.
It's
also
called
a
pot
template
and
I
I'm
I'm
not
really
sure
if
it
works
that
away
using
the
field
set
with
prefix
stuff
to
to
I'll
I'll
I'll
push
it
in
after
the
meeting,
and
I
continue
but
yeah
there's
a
few
few
parts
that
I'm
still
not
sure
if
you
want
to
do
them
this
way.
So
I'll
keep
in
touch.
A
Yeah
yeah,
thank
you.
Kevin
remove
manage
fields,
I'm
not
gonna
lie.
This
is
backlog
now,
because
we've
done
what
we
believed
was
the
most
important
okay
and
extending
schema.
I
I
don't
know.
If
you've
I
mean
you,
I
know
you've.
C
I'm
still
interested
in
it.
Let's
keep
it
on
the
list.
I
was
also
going
to
ask
if
daniel
had
mentioned
that
there
might.
It
might
be
good
for
me
to
spend
some
time
looking
at
how
clients
go
could
better
support
apply?
I
haven't
looked
at
that
space
at
all,
so
I
just
had
a
general
question
for
the
group,
which
is:
has
anybody
already
been
working
in
that
space?
Is
there
any
thoughts
or
anything
written
down?
I
should
be
reading
about
the
kind
of
the
client
go
support
for
apply.
C
C
A
C
Let's
move
that
up
if
that's
okay,
yeah
I'm
going
to
start
banging
on
it
and
seeing
if
I
can
come
up
with
another,
I
know
it's
a
little
bit
of
a
hard
nut
to
crack,
but.
A
So
the
biggest
fun
with
the
the
the
group
legends
especially,
is
that
it's
going
to
keep
some
fields
and
it's
going
to
send
the
default.
The
zero
value,
sometimes
right
now
it's
empty,
and
it's
going
to
pretend
that
this
is
part
of
the
intent
which
we
don't
want.
C
Yeah,
that's
that
we
gotta
fix
that
okay
yeah.
So
that's
one
of
the
one
of
our
key
requirements.
Is
you
actually
have
to
be
able
to
scope?
You
actually
have
to
be
able
to
define
a
manifest,
which
is
these
are
the
fields
I
care
about.
Somehow
we
have
to
do
that
using
as
much
of
the
existing
client
go
infrastructure
as
we
can.
B
Yeah,
I
think,
while
discussing
it,
I
I
think
I
think
it
was
this
one
that
daniel
or
anthony
suggested
do
you
have
a
kind
of
a
builder
interface
for
for
the
types,
because
we
can't
use
the
go
types
to
encode,
json
or
yaml
ads
spec,
or
something
like
that
that
operates
on
on
maps.
But
I
don't
remember
exactly
it
should
be
in
there
somewhere.
C
I
hope
yeah
I
worked
on
a
previous
system
and
granted
it
was
in
a
different
language.
It
was
in
java,
but
the
way
that
you
could
do
all
this
is
that
in
job
it
was
always
getters
and
setters
anyway,
so
it
kind
of
looked
a
little
more
normal
in
that
language,
but
basically
you
had
to
guard
everything
behind
an
interface.
C
A
Yeah,
I
think
the
the
analogy
is,
with
the
met,
the
meta
accessor.
You
know
how
the
meta
acceptor
can
work
on
like
a
typed
object.
I
can
work
on
a
on
a
structural
object
and
I
think
that
you
use
the
the
same
pattern
as
a
meta
accessor,
so
you
have
a
method
that
you
can
call.
So
you
have
a
complete.
You
know
what
the
types
are
and
what
what
exists,
but
it's
actually
feeding
a
map.
So
the
things
that
you
don't
touch
are
never
going
to
be
set
in
the
map.
C
Right
that
makes
more
sense,
yeah,
which
kind
of
kind
of
maps
with
like
how
we
really
do
apply
right.
Everything
is
everything
is
carried
around
in
unstructured
representation,
but
we
know
that
there
is
types
so
you
have
to
like
in
a
in
a
type
safe
language.
You
want
to
provide
them
with
a
type
safe,
client,
exactly
okay,
okay,
I'll
start.
C
Looking
at
that,
the
only
other
thing
I
had
is
there's
a
leftover
from
the
unfit
setting
fields,
work
where
right
now,
we
don't
have
full
feature,
parity
and
applies
like
like
with
structured
like
the
other
merged
or
the
patch
operations.
In
that
you
can't
express,
I
want,
I
expect
a
optional
field
to
be
sent,
or
I
expect
a
map
entry
to
be
absent,
and
so
one
proposal
yeah
no
way
to
declare
this.
Yes,
this
should
not
be
there
or
this
should
not
exist.
C
One
way
to
one
way
to
kind
of
band-aid
over.
That
might
be
to
add
some
kind
of
tombstoning
where,
like
there
is
a
special,
a
special
way
to
represent
in
a
configuration
or
a
manifest
in
yaml
that
you're
saying
that
a
map
item
or
a
optional
field
is
you
have
an
opinion
about
it,
but
your
opinion
is
that
it
should
be
absent,
yeah,
and
so
I
was
going
to
put
out.
I
have
a
rough
sketch
of
a
design
for
that,
so
I'll
probably
be
sending
that
around.
C
So
that's
awesome.
Videos
like
that.
C
I
do
have
some
notes
about
it.
If
anybody
starts
it,
but
yeah.
B
I
think
clint
go
client
would
be
more
valuable
right
now:
okay,
cool.
I
agree
with
that.
A
A
We
need
to
figure
it
out.
Okay,
let's
okay,
mike
you
want
to
talk
about
the
the
api.
B
A
B
B
A
A
Yeah,
I
finally
removed
all
the
tasks
from
from
the
project
here,
especially
in
the
topology.
We
have
a
ton
of
things
to
document
that
are
table
and
I've
had
a
email
exchange
with
somebody
at
google
this
week
about
how
the
documentation
is
not
good
enough
and
confusing.
So
I
I
know
I'm
I'm
gonna
have
to
work
on
this
this
week.
B
Yeah
I
mean
we're
at
co,
incorporates
now
with
the
release.
I
I
missed
my
time,
but
if
that's
more
important
before
we
release,
I
can
help
with
that
and
continue
with
the
status
wiping
afterwards.
C
Okay,
we
could
still
go
in
post,
119.
A
Okay,
so
noodle
I
am,
I
was
looking
at
the
pr
when
the
meeting
started
so
and
I
know
we've
been
through
a
couple
of
iterations
and
I
think
we
are
very
close
to
being
able
to
merge.
C
I
don't
yeah,
I
think
jenny's
was
I
need
to
do.
I
need
to
make
sure
that
we
got
all
the
that
we're
unblocking
her
in
every
way
possible,
but
I
know
she
had
a
pr
out
for
that.
I
think
she
was
working
to
get
that
refreshed
this
one
specifically
yeah.
I
think
that's
the
one
I
think
so
well.
No,
no!
No!
No!
That's
something
else!
Sorry!
Okay!
It's
about
these!
It
hurts
us
about
the
defaulting,
yeah,
yeah,
okay,.
A
A
I
think
I
so
I
see
a
couple
of
a
few
people
on
the
on
the
call
that
haven't
talked
yet.
I'm
not
gonna
put
them
on
the
spot
if
they
don't
want
to.
But
if
you
want
to
stop
contributing,
I'm
I'm
happy
to
help
you
find
tests
in
this
update.
G
G
Which
is
when
you
do
a
force
apply.
Is
it
possible
to
also
provide
the
fields
that
you
the
values
of
the
fields
that
you're
forcing
so
that
we
can
avoid
like?
So
if
we
have
a
scenario
where
we're
continuously
deploying
and
then
without
force
and
a
conflict
is
detected,
then
we
can
asynchronously
follow
up
for
human
judgment
on
whether
we
should
force
and
we
can
provide
the
human
a
like
a
list
of
the
things
that
will
be
forcibly
updated.
G
But
in
the
meantime,
if
some
other
controller
changes
a
field
we'd
like
to
expire,
that
manual
judgment
in
the
before
we
apply
it.
So
if
we
go
and
forcibly
apply
after
an
approval,
it
should
still
error
out
something.
C
Yeah
so
you're
thinking
about
the
case
where,
like
you've,
got
a
conflict,
you
want
to
resolve
that
conflict,
but
only
that
conflict
and
not
any
new
conflicts
that
come
in
the
meantime
exactly
yeah.
I
wondered
about
this
too.
I
wondered
if
optimistic,
concurrency
control
might
be
enough
ant.
One
of
you
thought
about
it
at
all.
A
We
did
talk
about
it
a
little
bit.
This
is
under
this
item.
It's
not
in
it's
in
the
backlog
right
now,
but
if
so,
first
of
all
welcome
jacker
and
yeah.
So
we've
thought
about
this.
So
can
you
please
look
at
this
issue
and
tell
us
if
this
would,
if
this
is
exactly
what
you
have
in
mind
and.
G
A
Yeah
plus
one
these
ex
explain
if
this
is
matching
exactly
what
you
want
or
how
you
would
like
the
solution,
maybe
to
look
like
if
you
have
an
opinion.
It's
it's
interesting
that
you're
mentioning
this,
because
I
we
didn't
know
how
important
this
was
gonna
be,
but
I
realized
that
it's
probably
going
to
be.
C
So
I'm
I'm
gonna,
try
and
move.
You
know
yeah,
even
when
I
was
just
like
theoretically
kind
of
going
through
all
the
stuff
that
you
could
do
with
the
pious.
I
was
learning
that
I
felt
like
it
was
a
gap
I
was
like.
Oh,
we
have
currency
control
primitives
here
that
are
really
nice.
Oh,
but
there's
this
gap.
I
can't
there's
this
thing.
You
can't
possibly
do.
G
Yes,
I'm
also
looking
at
I'm
a
home
maintainer,
I'm
looking
at
refactoring
the
ownership
model
in
home,
which
is
pretty
it's
based
on
releases,
and
it's
just
you
own
the
entire
object.
Basically,
I
think
we
can
make
it
a
little
bit
more
nuanced
and
allow
cross-release
ownership
of
the
same
resource,
potentially
with
server
side,
apply
support.
If
we
had
this
feature.
G
G
Yes,
so
if
I'm
overwriting
a
field
with
value
foo
to
value
bar,
if
the
field
in
the
meantime
has
been
updated
to
baz,
then
I
want
to
prompt
for
manual
judgment
again
or
you
know,
error
out.
C
G
A
Yeah,
I
think
we're
looking
for
the
right
semantics
of
how
do
you
specify
that
and
I'm
yeah
I'm
curious
how
we
could
do
that.
It
sounds
like
the
resource
version
is
not
gonna,
be
good
enough,
because
it's
it's
changing
too
quick
and
especially
in
the
cicd
workflow,
where
you're,
like
somebody
you're
going
to
test
and
see
if
you
have
conflicts
and
then
maybe
you
want
some
human
to
say,
yeah
I
want
to
the
resource
will
have
most
likely
changed
in
between
so
the
resource
version
wouldn't
work.
Yeah.
C
A
What
I
have
in
mind
right
now
is
that
we
could
literally
hash
something
maybe
like
hash,
the
the
set
of
conflicts
and
return
this
hash,
as
we
display
the
list
of
conflicts
and
possibly
have
an
option.
When
you
apply
and
say
I
want
to
follow
this
hash,
the
hash
has
changed
because
the
conflicts
have
changed.
Then
you
get
an
error.
Otherwise
you
just
go
to.
I
don't
know.
G
That
sounds
pretty
reasonable.
I
can
chime
in
on
the
the
ticket
that
was
linked,
yeah
an.
B
Extension
of
that,
I
would
have
said
hash
every
single
conflict
like
the
the
conflict
and
its
valley,
and
then
that
sounds
more
like
what
jake
was
asking
for.
Like
you
can
say,
I
want
to
force
this
conflict,
it's
an
often
pieces
out
of
a
conflict
but
yeah.
I
think
the
ticket
is
a
good
way
to
outline
all
use
cases.
A
A
So
this
is
typically
a
routing
thing
in
the
open
api
solo.
Well
like
the
way
we
set
up
the
api
service
is
completely
different
from
the
way
we
set
up
anything
else,
any
other
resource,
and
so
it
doesn't
have
the
field
manager
properly
set
up
when
it's
set
up
when
this
is
installed,
and
so
we
can
track
the
manage
field
and
we
can
do
so
outside
apply
for
the
api
service.
A
This
you
need
to
know
the
internal
of
the
api
server
to
do
that.
I'm
I'm
definitely
not
the
right
person
to
do
that,
because
I
don't
know
enough
about
the
setup
of
the
api
server
okay.
So
it's
not
an
easy
first
task.
It
sounds
like
it
could
be
an
easy
first
task,
but
it's
it's.
I
don't
think
it
is
unless
you
want
to
learn
about
the
the
internals
of
the
api
server.
A
It's
pretty
much
new,
we
need
jordan
to
do
it.
Empty
bird
selector
doesn't
mean
empty.
A
A
I
think
we're
kind
of
not
seeing
these,
and
I
don't
know
if
we
in
server
side
of
play
can
make
a
difference
between
a
piner
and
an
empty
thing.
Do
we
joe?
Can
you
remember.
C
You
know
we
right
now
right
now.
What
will
happen
is
whatever
you
put
in
your
configuration
will
get
carried
along
until
it's
converted
into
native
objects.
If
that
happens
right,
so
if
it's
a
crd,
it'll
just
store
the
json,
so
it
can
tell
totally
tell
the
difference
between
you
know
a
nil
thing
and
an
absent
thing
and
an
empty
map.
C
An
empty
list
can
tell
the
difference
between
a
lot
of
things
that
are
kind
of
sometimes
coalesce
other
places
if
it
gets
converted
to
a
nato
type
that
will
just
like
disappear
right
like
if
the
native
type
is
a
nullable
field,
then
it
can't
it
can't
hold
the
difference
between
an
absolute
field
and
a
null
value,
so
it'll
it'll
get
like
massage
down
or
it
kind
of
like
coalesce
down
to
whatever
the
representation
is
so
it
depends
which
is
kind
of
annoying
right,
because
that
means
you
have
to
actually
know
as
a
user
if
you're
working
with
a
crd
or
a
native
type,
and
if
the
native
type
is
this
kind,
what
it's
going
to
do.
A
Yeah
the
problem
here
is
that
you
have
either
an
empty
map
or
a
map
that
has
fields
and
the
way
we
merge.
That
is
that
we
check
the
nothing
plus
the
thing
that
other
thing
and
when
we
melt
we
end
up
with
the
thing
right,
and
so
the
cement
is
completely
wrong,
because
we
melt
nothing
which
means
everything
and
something,
and
we
end
up
with
something,
of
course,
while
merging
everything
plus
something
should
obviously
be
everything,
but
I
mean
we.
D
Have
no
way
of
knowing
that
there
is
no
chance.
We're
gonna
know
that
yeah,
there's,
not
it's
just
not
available
to
us
solution-
is
to
make
that
atomic,
which
actually
makes
sense
according
to
jalan,
and
I
think
I
agree
with
him.
C
I
saw
his
comment
about
that
and
I
agreed
yeah
and
then
I
also
saw
your
comment
where
it's
like.
I'm
not
disagreeing,
but
note
that
this
is
happens
at
the
call
site
not
on
the
object
itself,
yeah,
which
makes
it
harder
to
to
actually
implement
and
ensure
that
we
do
everywhere.
F
So
this
sounds
a
little
nasty
in
terms
of
implementation
and
if
the
crd
author
wants
to
use
it
in
his.
C
Right,
you
guys
can
assign
this
to
me
anton
because
it
feels
really
close
to
what
I
was
just
working
on.
I'm
sorry,
I
haven't
gotten
more
time
with
it
yet,
but
let
me
let
me
just
go
through
all
the
cases
and
see
what
we
can
start
working
on,
because
that
might
be
something
I
can
do
in
the
background.
A
I
remember
having
long
discussions
about
that
when
we
prototyped
that
and
like
is
the
least
type
of
property
of
the
of
the
structure
itself,
of
how
it's
used,
and
I
think
arguably,
it's
going
to
be
both
depending
on
the
situation
yeah.
I
think
that
we
have
to
go
because
it
can
be
in
two
places
now.
C
Yeah
I've
been
wondering
for
a
while
and
sorry.
This
is
a
bit
of
a
tangent
if
we
need
more
winters
on
the
native
types
to
make
sure
that
we
follow
conventions
that
we
expect
things
to
have,
and
if
this
would
be
like
that
kind
of
case,
we're
like
you
would
you
would
say
well,
if
it's
already
defined
on
the
type,
then
you
really
shouldn't
define
it
on
the
call
side
or.
D
A
Okay,
all
right
so
like
today,
if
we
change
this,
these,
these
types
to
be
atomic,
what's
gonna
happen,
so
we
have
a
section
in
where
is
it
gone
here?
A
I
can't
remember
exactly
well,
but
I've.
I've
already
mentioned
that
we
need
to
have
to
understand
what
the
behavior
is
when
we
change
the
types
and
we
need
to
document
that
properly,
which
we're
not
doing,
because
we
don't
even
know
how
the
code
is
going
to
behave
in
this
case.
But
I
wanted
to
have
tests
to
make
sure
that
first
of
all,
we
don't
change
that
behavior
accidentally
and
second
of
all
document
the
actual
behavior.
A
C
D
And
this
is
yells
man,
you
have
too
many
yeah.
I
might.
I
might
grab
jenny
or
something
if
I
get
overloaded,
we'll
see
how
it
goes.
D
A
We
have
a
little
bit
of
time
and,
let's,
since
the
code
freezes
now,
let's
discuss
maybe
about
other
things
that
we
want
to
do
in
the
next
cycle.
B
F
A
As
much
as
I
liked
last
time
saying:
okay,
we're
gonna,
the
plan
is
gonna,
be
these
and
we're
gonna
plan.
Accordingly,
it's
been
a
massive
failure
and
I'd
rather,
let's
let's
work
on
the
items
that
are
important
and
see
when
we
can.
B
A
B
C
D
Big,
like
that
left,
I'm
gonna
keep
poking
at
your
sentence.
What's
what's
your
use,
though,
like?
How
are
you
talking
about.
C
Yeah,
I'm
thinking
of
people
that
are
already
already
starting
to
adopt
server
side
apply
doing
things
for
it.
Maybe
they're
writing,
maybe
they're
writing
infrastructure
that
leverages
it.
You
know
anything
like
terraform
style
tools
or
whatever
are
there?
Are
there
big
outstanding
things
that
we
plan
to
change
that
would
be
like
breaking
for
them
or
otherwise.
A
Really
disruptive,
so
I'm
I'm
gonna,
think
I'm
gonna
tell
you
what
I
think
right
now.
I
I
suspect
that
changing
your
workflow
for
coop,
cuddle
from
client
side
to
server
side
is
fairly
easy,
and
so,
if
you're
using
client
side
apply
today
and
tomorrow,
you
want
to
use
source
code,
apply
you
either.
Maybe
add
the
flag,
create
an
alias
for
icd
pipeline
to
use
the
flag,
and
I
I
suspect
it's
mostly
gonna
work
the
same.
It's
maybe
not
gonna
work
exactly
the
same,
but
it's
gonna
work
the
same.
C
A
So
I'm
seeing
I'm
seeing
like
the
cube,
color
user
and
the
controller
users,
they
have
very
different
timelines.
One
is
going
to
change
and
maybe
it's
gonna
work,
the
other
has
to
plan
and
design.
Based
on
that,
so
I
suspect
we
should
tackle
the
controller
users
first
and
I'm
seeing
a
lot
more
people
asking
me
I'm
trying
to
build
this
controller.
A
D
A
C
That
makes
a
lot
of
sense
to
me,
because
I've
heard
a
couple
of
people
say
things
along
some
more
lines.
So
it
sounds
like
what
we
should
be
doing
is:
let's
land,
the
stuff
that
we've
already
got,
that
we're
already
working
on
and
then
try
and
get
as
good
of
a
story
as
we
can
for
controllers
and
kind
of
go
and
not
going.
A
Yeah
so
yeah,
so
if
you
look
at
the
controller,
this
is.
This
is
a
part
because
this
is
literally
breaking
cube
builder
right
now,.
C
B
A
B
Yeah,
I
I
think
especially
api
client.
I
thought
about
it,
so
they
apply.
I've
been
writing
a
few
controllers
over
the
last
few
weeks
and
I
always
thought
apply
would
make
it
easy,
and
then
I
thought
well.
I
do.
I
want
to
apply
a
right
json
that
I
want
to
apply
like
or
the
ghost
trucks
like.
I
think
your
api
clients
would
be
a
big
benefit.
A
C
C
Oh,
the
the
validation,
stuff,
schema,
validation,
right.
A
C
A
C
Yeah,
it
feels
like
that's
kind
of
a
separate
track,
so
if
anybody
wants
to
work
on
that
track,
that's
great,
but
it
feels
like
it's
a
little
separate
yeah
once
mike
gets
all
some
more
more
more
context
from
segarc
or
whatever
on,
unlike
how
we're
going
to
tackle
declarative
apis.