►
From YouTube: Kubernetes SIG CLI 20221116
Description
Kubernetes SIG CLI bi-weekly meeting on November 16th, 2022.
Agenda and Notes: https://docs.google.com/document/d/1r0YElcXt6G5mOWxwZiXgGu_X6he3F--wKwg-9UBc29I/edit#bookmark=kix.vejp7jzgg2y
A
A
A
I
think
we
have
some
very
aggressive
deadlines
for
some
enhancements,
including
for
prune
and
so
I
just
want
to
make
sure
that
that's
at
the
top
of
everybody's
mind,
we
have
the
holiday
season
coming
up,
and
so
there's
going
to
be
quite
a
bit
of
of
non-work
going
on.
So
we
we
want
to
keep
our
eye
on
the
next
day,
which
is
the
enhancements
freeze.
We
don't
know
what
that
date
is
yet,
but
we
should
distribute
that
when
we
do
get
that
data.
A
A
Okay,
why
don't
we
move
on
to
introductions
then?
So
this
is
the
part
of
our
meeting
where
those
who
are
new
to
the
6cli
or
haven't
been
here
for
a
while
can
introduce
themselves
to
your
colleagues.
I
think
we
do
have
a
couple
of
new
people.
This
is
completely
voluntary.
So
if
you,
if
you
don't
want
to
do
that,
that's
fine
as
well.
Would
anybody
like
to
introduce
themselves.
A
A
B
A
C
A
C
B
B
B
But
then
the
contributor
reasonably
asked
for
guidance
on
on
how
to
go
about
it,
because
this
command
is
under
tested
in
general
and
looking
into
it,
it
seems
like
the
reviewer
on
the
original
Cube
control,
cppr
way
back
when
asked
for
the
same
coverage
that
I'm
asking
for-
and
the
answer
was
basically
it's
impossible,
because
RCI
doesn't
actually
run
the
containers.
So
you
can't
possibly
test
things
that
depend
on
responses
effectively
from
the
container
runtime
when
it
comes
down
to
it,
and
that
seems
to
be
true
today.
B
As
far
as
I
can
tell
so
I
guess,
first
question:
does
anyone
know
of
that
not
being
true
or
a
way
that
we
have
to
test
features
that
depend
on
on
containers
and
then
the
second
question
is:
if
that's
not
true,
how
the
heck
do
we
go
about
making
improvements
to
these
areas
that
actually
can't
be
tested,
but
like
even
when
it?
You
can
see?
B
There's
like
a
big
back
and
forth
between
me
and
the
contributor,
because
I
mean
the
contributors
are
using
different
container
runtimes
in
our
local
environments
and
the
feature
that
they're
adding
works
for
them
and
it
doesn't
work
for
me.
So
it's
there's
also
that
complexity
there
that,
like
it,
actually
depends
on
the
container,
runtime
and
I.
Think
the
protocol
that
gets
selected
automatically
in
negotiation
also
changes
something
just
from
digging
around
the
code
in
this
case
about
the
error
handling.
E
Maybe
the
offer
was
looking
either
at
the
unit
test
which,
yes,
they
don't
on
containers,
or
they
were
looking
into
integration,
which
again
they
don't
run
containers
because
integration
usually
starts
the
API
server
controller
managers.
They
don't
they
don't
have
cubelets,
but
e2e's
are
executed
against
a
full-blown
kubernetes
cluster,
so
you
should
get
whatever
you
need
from
there.
That's
the
first
thing,
a
general
guideline
for
the
cases
like
that.
F
E
B
Yeah,
the
to
for
that
last
part,
the
what
they're
trying
to
do
is
enrich
the
error,
but
the
problem
is
that
some
container
from
some
container
run
times
we're
getting
back
a
structured
status
where
we
can
actually
see
the
exit
code
as
an
exit
code
and,
in
other
cases
we're
just
getting
back
strength.
That
does
happen
to
mention
the
exit
code
in
there,
but
it's
not
structured.
So
we
have
to
be
string
scanning
and
our
original
thought
on
the
pr
was
like.
Well,
we
don't
want
to
string
scan.
B
E
B
E
B
G
I
was
gonna
Echo,
some
of
the
same
stuff,
I
think
this
I
wouldn't
even
try
to
mess
around
catching
with
this
unless
we
can
stub
out
or
mock
things
inside
of
our
unit
tests,
the
integration,
u2e
bash
tests
and
the
ginkgo
test,
and
all
that
should
be
testing
for
this.
If
we
want
to
catch
it,
there
I
think
some
of
those
do
run
on
pre-submit,
though
we'll
have
to
check
into
which
ones
do,
but
the
greater
thing
about
the
copy
command.
G
I've
talked
to
the
CRI
folks
and
Phil
Estes
and
container
defaults,
and
a
bunch
of
different
people
at
kubecons
and
Micah
Hassler
too
Micah
Hasler
at
this
recent
one.
He
had
some
ideas
but
I
think
the
the
biggest
thing
that
we'd
all
like
to
see
is
this
API
should
just
be
part
of
container
runtime
interface
to
copy
files
from
and
move
between,
a
container
and
not,
and
it
sounds
like
everyone's
on
board.
E
I
I.
There
was
also
a
discussion
as
far
as
the
dashboard
group,
where
they
would
want
to
see
a
much
more
tenderized
exit
codes
from
being
returned
from
the
CRI,
because
in
in
The
Bachelor
group
case,
they
wanted
to
act
upon
particular
failures.
The
problem
is,
the
current
failures
are
not
standardized,
so
that's
probably
what
Eddie
was
talking
about.
There
are
efforts
to
standardize
that
and
let's
see
right
but
it'll
be
it'll,
be
some
time
before
we
get
to
that
point
where,
where
this
is
standardized
first
and
then
implement
it
across
the
board,.
G
B
F
B
F
E
F
A
F
F
F
H
H
A
E
There
is
an
enhancement
where
we
wrote
it
down.
I'll
try
to
find
the
link
in
a
moment
about
the
decisions
about
the
consideration.
The
general
consensus
was
that
copy
is
pretty
broadly
used
and
we
would
not
want
to
force
people
off
of
it
or
eventually
force
people
to
start
using
a
plugin
to
do
that.
But
we
will
significantly
limit
the
capabilities,
especially
those
around
blobbing
and
recursiveness,
because
those
were
the
ones
that
were
causing
the
majority
of
the
problems.
E
We
decided
to
only
allow
copying
single
files
instead,
I'm,
obviously
not
sure
the
directories
are
supported.
So
this
way
it's
just
simpler
and
we
can
have
a
full
control,
because
a
lot
of
the
cves
were
related
with
how
we
are
dealing
with
with
links
and
all
those
edge
cases
where
you
can
actually
Escape.
D
B
Sir
definitely
good
to
look
into
I,
just
I'm,
still
not
totally
clear
on
what
I
should
say
to
the
contributor
should
I
tell
them
actually
just
remove
the
tutu,
because
we
can't
provide
it
like.
We
can't
provide
a
consistent
message
or
should
I
tell
them
to
look
into
the
ete
tests,
because
we
and
and
like
except
a
mix
of
string
scanning
and
exit
code
checking
or
yeah
like
what?
What
is
the
final?
Does
anyone
feel
strongly
about
what
we
do
with
the
particular
PR
in
light
of
what
we
just
discussed,
I.
E
Would
probably
just
suggest
dropping
it
to
do
and
if
Eddie
has
some
links
towards
standardization
and
the
CRI
interface,
or
there
are
some
discussions,
I
can
probably
look
up
the
ones
that
happen
with
regards
to
batch.
We
can
point
to
those
places
and
then
just
close
were
eventually
a
limit
to
only
dropping
the
the
to-do's
for
now,
because
that's
that's
not
something
that
we
we
would
be
willing
to
fix
at
this
point
in
time.
G
A
E
F
E
E
Reinvent
this
mechanism,
because
that's
something
that
is
that
is
proven
stable
and
we
wouldn't
have
to
deal
with
the
conversions
and
all
those
problems
that
might
appear.
We
would
be
just
generating
new
types
and
we
would
ensure
that
they
convert
properly
between
different
versions
and
one
of
the
reasons
when
we
were
top
when
we
were
talking
about
it
during
kubecon
was
that
we
will
most
likely
be
heavily
changing
the
QRC
and
the
alpha,
especially
at
the
alpha
stage.
E
The
main
goal
is
to
have
something
usable
with
minimal
set
of
features,
but
out
there
as
soon
as
possible.
The
reason
for
that
is
the
sooner
we
ship
it
to
users.
The
sooner
we
will
be
able
to
gather
feedback
and
eventually
start
building
something
else
following
that
I
was
talking
with
Eddie
on
during
kubecon
and
I
promised
him
that
I'll
put
together
a
PR
that
actually
implements
the
Alia
command,
the
LDS
command
fulfills
at
least
50
percent
of
the
requirements
that
we
have
for
command.
E
E
E
The
second
comment:
if
you
could
open
it
up,
I
could
actually
I
was
actually
able
to
build
on
top
of
the
current
aliasing
mechanism.
That
Cobra,
already
provides
to
adding
an
alias
to
commands
is
as
simple
as
just
injecting
those
when
you
are
going
when
we
are
building
the
tree
of
commands
and
in
a
similar
fashion
like
in
inject
additional
Flags,
a
fun
fact
when
I
was
trying
to
work
with
the
QRC
Alias
command.
E
F
E
B
F
E
F
E
F
E
Only
have
to
wire
these
two
together
and
the
aliases
will
be
working
like
a
term
like
I
said
the
only
two
things
that
I
noticed
when
I
was
playing
with
it
was
one
audiences
are
actually
reported
in
the
help
commands.
So
if
you
do
I,
don't
know
Cube
cuddle
delete
help.
It
actually
gave
me
the
scenario.
Yes
I'm,
not
sure,
if
it's
something
that
we
do
want
to
do,
because
theoretically
you'll
get
that
information,
but
that
information
will
be
dependent
on
the
of
your
QRC.
E
Secondly,
the
audience
is
not
all,
because
the
Alias
will
not
show
the
information
about
the
the
flag
rejected.
So
that's
something
that
I
noticed
immediately.
The
second
thing
is
I:
did
I
didn't
play
with
it?
Maybe
it's
possible
with
the
completions
will
not
work
out
of
the
box.
I'm,
pretty
sure
that
if
we
reach
out
to
Mark
Mark
will
be
able
to
fix
it
for
us.
F
E
I
think
that
that
would
be
still
a
good
starting
point
to
to
get
the
ball
rolling,
although
if
I
remember
correctly
through
an
hour,
yes,
the
completions
will
also
should
apply,
because
when
I
was
working
with
that
Elite
and
playing
with
it,
it
actually
worked.
So
there's
a
high
chance
that,
just
because
we
are
registering
an
alias
for
a
command
through
a
cobra
built-in
mechanism
that
completion
should
work.
E
The
thing
that
I
did
not
test
is
specifying
multiple
commands,
multiple
arguments
to
a
single
command
or
the
priorities,
because
that's
the
other,
that's
the
other
thing
and
we'll
probably
have
to
figure
out
what
will
happen
if
someone
decides
to
have
different
aliases
for
for
the
same
command,
but
that's
just
because
the
the
structure
that
I
picked
is
created
this
way,
but
that's
that's
probably
the
least
of
the
problems
that
we
have
give
it
a
go
play
with
it.
I
I
put
it
together
on
Monday
or
yes,
Monday.
E
B
Sorry
two
questions
the
first
one
is
like.
Is
this
actually
what
is
in
the
cap
currently,
and
the
second
question
is
when,
in
this
example,
that's
like
showing
on
the
screen
right
now,
is
the
delete
being
replaced
with
blah
so
like
we're
allowing
built-in
commands
to
actually
be
completely
overwritten?
Is
that
correct.
E
E
This
way
when
you're
a
regular
user
RM
is
just
RM
and
it
won't
ask
you,
but
if
you
log
in
as
a
root
admin
as
a
root
user,
it
will
always
ask
you
for
for
the
confirmation,
so
I
figure
out
that
probably
the
ability
to
override
the
default
commands
would
be
the
bad
thing
that
we
could
do.
I'm,
not
saying
that
people
have
to
use
it.
But
this
way
you
could
be
protected,
especially
in
the
light
of
the
discussion
that
we
had
for
the
interactive
and
all
that
being
able.
E
Of
course
it
will,
it
will
cause
problems
in
some
Fields,
because
when
you
will
be
moving
between
different
environments,
where
on
one
you
will
have
the
QRC
file
and
on
the
other
one
you
want,
and
in
the
one
where
you
have
QRC,
you
will
have
a
default
Dash
Interactive
and
on
the
other
you
want.
Yes,
you
will
run
into
problems
and
we
will
not
save
you,
but
maybe
over
time
we
will.
We
will
teach
people.
F
B
So
the
the
two
concerns
that
I
would
have
with
this
approach
would
be
the
one
kind
of
religious.
What
you're
talking
about
about
the
help
test
text
that
built-in
commands
that
are
defined
and
well
documented
online
and
help
Etc
will
not
behave
as
as
documented
and
I
feel
like
making
issues
open
about
that.
B
E
I'm
open
to
to
say
that
initial
we
will
not
allow
because
there's
nothing
stopping
us
from
saying.
Oh,
we
initially,
we
will
not
allow
overriding
the
default
commands
and
we
will
only
allow
creating
aliases
which
are
not
an
existing
command.
That's
a
simple
check
in
the
in
the
inject
all
command
and
inject
function,
so
I'm
open
to
that
I.
E
Don't
have
any
strong
preferences
one
way
or
the
other
I
just
came
up
with
that
when
I
was
thinking
about
that
RM
hell,
yes
in
in
the
Linux
environment,
but
yeah
I'm
I'll,
probably
leave
it
tomorrow
again
or
that
to
figure
out
which
way
they
want
to
go.
I
I
promise
the
part
of
that
to
Eddie
that
I'll.
Do
it
this
week
and
I
did
everything
to
keep
the
promise.
B
Yeah
I
think
it's
a
really
neat
idea.
I
guess
that's
why
I
was
wondering
if
it
was
in
the
kept
like
that
because
like
if
we
were
relying
on
providing
the
functionality
through
an
overwrite
like
the
ability
to
default
the
ARG
through
an
overwrite,
then
saying
like
well
we're
not
going
to
allow
override
at
first.
B
E
E
B
I
guess
my
question
is
like
what
is
the
API
look,
the
and
API
look
like
because
if
it
looks
like
overriding
aliases
for
entire
commands,
then
we
wouldn't
have
instead
of
like
overwriting
arguments
or
something
like
that,
and
we
decided
it's
actually
not
safe
to
override
entire
commands.
Then
we'll
need
to
change
it
to
accommodate
overwriting
arguments
and
is
that
going
to
fit
nicely
in
in
the
plan,
but
I
just
don't
have
a
good
mental
picture
of
what
that
looks.
Like
I'll
stop
talking,
though,
there
are
two
other
race
hands,
yeah.
E
I
Oh
that's
right.
It
went
into
a
separate
so
there's
the
in
the
kept
there's
two
proposed
Edition
or
two
proposed
like
fields
in
the
in
the
cube,
RC
config
file.
There
they
go
yeah,
it's
aliases
and
overrides,
and
so
I.
When
I
did
my
implementation.
I
did
nothing
to
check
to
make
sure
that
aliases
weren't
overriding
built-ins,
because
it
was
just
a
simple
POC
but
I
I
changed
it
or
sorry.
I
I
implemented
the
delete
confirmation
using
the
overrides
part
so
that
that
basically
just
looks
to
see
like
what.
What
command
are
you
running
and
then
like?
If
there's
more
commands
like
it
checks
where
the
delineation
between
the
list
of
sub
commands,
you're
providing
and
the
flags
actually
start,
and
so
it
finds
what
command
what
command
or
what
sub
command
you're
running
and
then
applies
flags
as
as
a
defined
on
the
cube
RC.
I
Basically,
the
main
thing
I
didn't
figure
out
how
to
do-
and
this
is
what
my
question
to
the
group
is
is
some
Flags
I
think
can
take
like
be
defined
multiple
times
and
I.
Don't
know
how
to
solve
that
really
I
guess,
because
what
I'm
doing
is
the
way
that
the
precedence
is
set
in
the
execution
as
it's
like
gosh
I.
Forget
it,
it's
like
backwards
from
how
I
assumed
it
would
be
like
the
ones
that
are
at
the
beginning.
E
E
So
it
is
not
easy
for
you
to
inject
them
directly
because
you
can't
override
the
osrs
flat,
so
you
have
to
inject
into
the
park.
Now
you
have
to
be
able
to
hook
into
the
parsing
mechanism.
I
think
that's
a
kind
of
work
in
my
case
through
through
that
so
yeah
I,
don't
know,
probably
something
that
you'll
have
to
double
check
with
what
you're
saying
I
yeah
the
the
flux
are
the
the
Flack
in
general,
where
the
more
are
the
most
tricky
part
of
the
entire
implementation,
foreign.
G
So
I
I'll
be
short,
and
then
we
should
move
on
to
our
guest
topic,
so
we
don't
make
them
come
back.
The
I
haven't
had
a
chance
to
look
through
either
PR,
yet
I'm.
So
sorry,
the
the
idea
of
I
think
your
PR
looks
good
mache,
but
I
think
the
way
you
combine
the
Alias
and
the
overrides
things
is
a
foot
gun
like
Katrina,
said
so.
I
think
it's
a
good
proof
of
concept
for
us
to
move
on
there
and
we
can
improve
on
it.
G
I
I,
don't
every
y'all
are
running
into
the
same
thing
that
I
ran
into
when
I
prototyped
This
months
ago,
and
it's
that
Cobra
isn't
set
up
for
us
to
do
this,
and
instead
of
hacking
around
all
of
that
look,
John
McBride
said
that
whatever
changes
we
want
to
get
into
Cobra,
he
will
most
likely
approve
and
get
in.
So
we
don't
have
to
do
this
hackie.
We
should
maybe
add
hooks
to
the
Cobra
command
struct
that
you
can
add
a
hook
to
intercept
and
like
change
the
flags
or
something
but
think
outside
the
box.
G
E
Yes,
I
at
some
point
in
time,
I
was
like
desperate
enough
to
I
need
to
talk
to
John,
because
I
can't
figure
it
out,
but
then
the
the
parsing
was
there.
So
probably
more
testing
is
required
if
the
parsing
works,
then
we're
good,
if
not
definitely
talking
to
John
and
the
sooner
the
better,
because
it
would
be
good
to
get
it
on
in
127
as
a
simple
Alpha
and
let
people
test
it
and
Report.
C
So
hi
so
yeah
so
like
I've,
been
working
on
a
very
simple
tool.
But
basically
you
know,
like
you,
run
a
command
and
then
you,
you
know
it
gives
a
bunch
of
log
outputs
and
you
just
kind
of
check
whether
the
command
has.
Actually
you
know,
like
you
check,
you
have
a
reference.
I
mean
like
and
you
check
it
out
yourself.
It's.
C
You
know
like
the
sample
markdown
file
right
I
mean
like
the
idea
is,
like
you
run
a
command
and
you
get
an
output
and
you
check
it
against
the
code
of
preference,
I
I.
The
reason
this
came
about
was
because
we
wanted
to
I
think
there
was
one
there
were
a
couple
of
requirements
where
we
want
to
check
that
command
line
interfaces
I
mean
like
so
that
you
know.
If
you
make
a
change,
you
know
like
you,
don't
you
don't
break
the
command
line
interface?
C
C
F
E
We
left
that
quite
some
time
ago,
we
left
ourselves
quite
a
a
little
bit
of
freehand
to
be
able
to
change
the
output
in
some
ways.
Of
course,
there
are
a
bunch
of
tests
where
we
are
looking
at
the
output,
but
just
skimming
through
through
your
examples.
The
MD
examples
I
have
a
couple
questions
like.
E
How
are
you
planning
to?
How
are
you
planning
to
deal
with
the
variable
elements
of
the
output?
Your
and
Sample
MD,
for
example,
has
timeline
or
or
some
IDs
or
stuff
that
will
be
inevitably
changing
during
the
execution
in
a
similar
way,
certain
parts
of
the
cube
cuddle
output
will
be
will
be
variable
and
I
would
not
want
to
have
very
strict
output
batching,
but
rather
something
closer
to
brexit,
Reg,
X
or
something
that
would
tell
me.
Oh
now,
here's
a
little
bit
of
change
and
then
there's
this
trick
part.
C
Yeah
so
currently
we
I
mean
like
we
have
we
offer
I
mean
I,
basically
use
regexs
to
you,
know,
take
out
the
headers
or
like
because
the
log
headers
are
going
to
keep
changing
so
I
use
regex
to
take
up
the
log
headers
and
only
look
at
the
output
of
the
log.
We
have
a
couple
of
you
know
like
projects
as
for
klog
and
then
Json
I
mean
like
you
could
add
more.
So
that's
that's
one
of
the
ways
we
kind
of
prevent.
C
You
know
the
parts
that
mutate
I
mean
like
yeah
I
mean
like
we,
you
know
like.
We
only
get
to
compare
the
stuff
we
want
so,
but
the
problem,
one
of
the
issues,
is
all
based
on
a
markdown
file.
So
you
can't
go
and
change
the
golden
reference.
So
I
mean
like
it's
it.
It's
it's
more
long
lines
of.
Do
you
want
to
go
and
change
the
reference
itself
in
the
sense
that
is
it?
Is
it
possible
to
identify
a
specific
logic
and
say?
C
Okay,
so,
although
the
output
is
variable,
it's
it's
still
compatible.
You
know
like
it's.
It's
still
okay
to
go
ahead
and
Mark
it
as
fine.
So
if
that's,
if
I
have
a
couple
of
examples
to
work
with,
maybe
I
can
just
you
know
reason
out
what
would
be
the
best
way
to
do
it
so
that
would
that
would
help
too.
E
F
E
Be
klog
g-log,
whatever
it
is
always,
oh,
it's
this
app
and
that
and
then
there's
everything
else
yeah.
Unfortunately,
that
does
not
apply
to
keep
cuddle
and
oftentimes.
Since
we
are
looking
at
stuff
like
the
out
of
the
the
width
of
the
out
of
the
output
terminal
and
so
forth,
that
data
might
change
between
invocations
in
that
you
will
have
a
variable
number
of
spaces,
or
rather
tabs
between
certain
commands
or
depending
how
long
there
will
be
or
yeah.
So
that's
one
thing
for
you
to
have
a
good
overview.
D
E
C
C
E
E
I,
don't
have
you
wrong,
yeah
that
has
some
as
well,
but
I.
Remember
that
I
nerfed
a
couple
of
them,
because
in
the
past
that
we
used
to
have
some
problems
with
this
stability
of
the
output
with
the
gas.
Currently,
the
stuff
is
a
little
bit
easier,
because
the
gut
can.
What
we
can
do
is
the
entire
printing
is
being
handled
by
the
server,
as
in
the
server
tells
the
keypad
of
how
to
print
a
particular
resource.
E
C
E
G
E
So
yeah
he's
aware
of
that.
I
asked
if
he
would
be
able
to
join,
but
I
remember,
press
I
mentioned
earlier
today
that
he
won't
be
but
he'll.
He
will
show
up
today
to
talk
about
it
earlier
anymore,
so,
based
on
what
I
understood,
because
I
could
probably
prove
me
wrong,
he's
working
closely
with
Patrick
on
it.
C
H
H
You
know
how
do
we
translate
things
that
come
from
the
API
or
how
do
we
translate
things
that
come
from
other
things,
whether
they're,
plugins
or
crds,
or
other
things
like
that
and
well?
It
might
be
possible
I
kind
of
started
thinking
that
that
sounds
more
like
a
step
too,
like
step,
one
being
hey,
let's
fix
our
infrastructure
that
we
have
currently
get
it
as
get
it
working
well
and
then
see
how
we
can
extend
and
build
on
that
so
I'm
open
to
discussion
on
that.
H
But
I
put
that
as
a
non-goal
in
the
cup
just
because
we
talked
about
it,
and
so
any
anybody
have
any
thoughts
on
that
or
you
can
chime
in
on
the
cup.
We
don't
have
to
talk
about
it
now,
because
we
only
have
a
couple
minutes,
but
you
know,
as
as
I
was
investigating
this
I
discovered
a
couple,
more
issues
that
were
affecting
coupetl's
ability
to
translate.
Even
though
we
had
translations,
they
weren't
being
out
out
hood
and
I
opened
some
issues
on
that.
H
So,
if
you're
interested,
you
can
take
a
look
at
those
issues,
yeah.
So,
for
example,
when
you
wrap
things
in
the
initializing
of
package
level
variables,
the
languages
haven't
been
loaded
yet
because
that
the
package
level
variable
initialization
occurs
before
main
runs,
and
so
in
Maine
runs
that's
when
it
loads
the
languages.
So,
even
though
it
looks
like
everything
should
be
working
right
when
you
actually
look
at
the
output,
those
things
are
not
being
translated.
H
No
I
didn't
talk
to
them.
You
mean
about
how
like
yeah,
I,
didn't
because
I
kind
of
thought
well.
This
sounds
like
a
a
a
second
step
to
this
process.
I
mean
I,
can
ask
them
or
tell
them
what
we're
trying
to
do
and
see
if
they
have
thoughts
on
on,
you
know,
ways
to
communicate.
I
guess
the
language
up
to
the
API
I
mean
I
I
yeah,
it's
probably
worth
the
conversation.