►
From YouTube: Kubernetes SIG CLI 20220223 - KRM Functions
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
Welcome
to
this
meeting
of
the
paragram
function.
Sub
project
today
is
february
23rd.
My
name
is
katrina
biry,
and
I
will
be
your
host
for
this
meeting.
We
have
an
agenda
today
and
that
that
is
being
presented
here
and
please
add
your
name
under
attendees,
with
with
your
company
that
you're
representing
to
all
those
present,
and
since
I
see
some
faces
that
I
recognize
on
this
call,
but
that
are
new
to
the
subproject
I'd
like
to
invite
carlos
and
ifan
to
introduce
yourselves.
B
C
Hi,
I'm
fun
carlos
and
I
work
on
the
same
team
and
then
we
know
katrina,
because
we
were
all
working
together
before
so
good
to
see
you
katrina.
A
So
we
have
a
couple
topics
here
today.
The
first
one
is
a
feature
request
from
carlos
I've
had
a
look
at
this
and
I
think
munchie
and
natasha
also
did
so
actually
unless
yuan
are
you
familiar
with
this
as
well,
or
should
we
give
an
overview
before
we
dive
into
discussion.
D
A
B
Sure
so
I
was
playing
with
the
results
part
of
the
krm
spec,
one
of
the
things
that
I
found
is
like:
let's
it
has
a
proposed
field
and
a
current
value
proposed
value
and
current
value.
Well,
it
is
not
explicitly
mentioned.
I
believe
the
intention
for
these
two
values
are
so
we
could
say
as
a
con
as
a
user
of
the
krm
spec
caren
function
spec,
I
may
want
to
build
a
tool
that
could
do
some
modifications
on
my
results,
but
maybe
not
directly.
B
B
Right
now,
I
only
have
one
possible
volume
and
what,
if
I
want
to
what,
if
I
run
one
five
and
the
way
to
do
it
right
now,
I'm
not
sure
if
this
is
the
expected
way,
but
one
way
of
doing
it
is,
I
could
generate
multiple
result
entries
multiple
results
where
one
will
say
proposed
value
saturday
and
the
other
one
will
say
proposed
value
is
sunday.
B
Essentially,
I
will
need
to
repeat
the
entire
lock
off
of
the
result,
and
if
I
have
five
options,
I
will
need
to
repeat
these
five
times
so.
My
proposal
here
is
to
have
essentially
multiple
options,
so,
instead
of
proposed
value,
it
will
be
proposed,
values
so
plural
and
then
it'll
be
multiple
fields.
Sorry,
multiple
entries,
where
each
entry
is
going
to
be
one
option,
although
this
will
be
a
breaking
change
for
the
spec,
because
that'll
be
dropping
one
of
the
fields.
B
B
A
Thank
you
yeah.
I
think
I
think
that
does
make
sense
to
not
force
the
consumer
to
dedupe.
Essentially,
what
errors
are
actually,
but
what
results
are
actually
a
single
result.
A
The
one
question
that
I
personally
have
about
the
suggestion
to
introduce
it
as
a
second
field,
duplicate
the
old
one
and
then
move
forward
since
we're
not
on
the
server
side.
Here
we
don't
have
a
way
like
we
don't
have
a
defaulting
stack
that
we
can
use
to
make
it
seamless
for
folks
consuming
these
apis.
So
you
could
theoretically
end
up
in
a
situation
where
you
have
one
system
that
implements
the
specification
that
emits
the
old
field
and
a
new
one.
A
That
is
aware
of
both,
and
we
have
to
say
what
they're
supposed
to
do
with
that
and
if
the
new
one
decides
well,
I'm
going
to
emit
the
the
newer
field,
because
because
that
is
not
deprecated
right,
but
then
it
goes
back
to
a
second
processor
that
is
only
aware
of
the
old
field.
Then
we
have.
We
have
a
problem,
so
I'm
not
sure
even
by
supporting
both
fields,
because
we
lack
that
defaulting
stack
component,
it
is
actually
compatible
between
tools
implementing
the
spec.
A
B
A
Yes,
that
said,
like,
I
wonder
how
is
it
very
expensive
to
make
a
v2,
given
the
small
number
of
consumers,
we
have
right
now
that
that
has
for
quotes
fields.
Our
proposed
values,
plural,.
B
Are
there
other
users
so,
like
you
said
there,
there's
like
very
few
users
currently
for
appearance
back?
I
wonder
if
there
have
been
other
users
stumbling
up
on
proposed
value
and
putting
like
oh,
this
is
fine,
so
mostly
just
wondering
do
we
have
enough
users
to
get
feedback
on
whether
this
will
be
like
a
good
idea
or
or
not.
A
E
So,
as
far
as
I
know,
we
we
we
are
not
using
this
feature,
a
lot
in
cab.
E
B
I
I
guess
so
if
I
want
to
modify
a
field
directly
to
a
known,
correct
value,
that's
only
one
value.
I
know
that
the
I
can
use
the
filters
to
make
that
notification
automatically
the
tool
I'm
thinking
about
is
something
that
will
ask
the
user
to
choose.
Do
you
want
to
make
this
change?
Is
it
okay
to
make
this
change,
and
sometimes
there
could
be
multiple
options
for
the
correct
value?
B
I
do
have
a
tool,
although
I
cannot
speak
too
much
about
it,
but
it's
it's
essentially
a
tool
that
will
ask
the
user
for
proposals
in
case
there
are
multiple
changes.
One
one
thing
that
will
be
similar
to
the
use
case
is
when
you're
working
on
say
bs
code,
and
if
you
have
a
spell
spelling,
corrector
extension,
when
you
click
modify,
you
will
get
a
list
of
multiple
options
to
choose
for
the
board.
You
want
to
correct
it's
essentially
that's
the
same
use
case
that
we
want
to
implement,
but
for
configurations.
C
I
don't
know
if
what
I'm
going
to
add
is,
as
give
you
a
little
more
information,
but
I
think
how
we're
going
to
use
this
field
like,
for
example,
I
don't
know
from
the
other
perspective
like
when,
when
the
user
or
when
krm
returns
the
results
with
a
list
of
proposed
values,
we
also
use
it
from
the
other
end,
not
from
the
analyzer's
end,
but
from
the
receiver's
end.
C
A
I
I
guess
we
still
have
the
two
options
on
the
table
like
if
we
think
that
we
have
a
decent
sense
of
the
consumers
of
this,
and
none
of
them
are
actually
using
this
field
anyway,
then,
the
the
concern
that
I
raised
about
incompatible
incompatibility
between
them
is
not
a
big
deal,
so
we
can
still
go
with
munchie's
suggestion.
A
Otherwise,
if
if
we
want
to
make
a
v2
of
the
spec
considering
there
aren't
very
many
consumers
and
none
of
them
are
using
the
field
anyway,
it
still
wouldn't
make
a
very
big
difference
to
anybody
to
cut
that
v2.
I
don't
think.
B
E
Yeah,
I
think
if
we
want
to
cut
with
who,
so
ideally,
we
should
get
more
feedback
from
our
users
and
accumulate
more
like
propose
the
change
and
then
cut
with
two
and
the
one
question
I
think
we
we
should
make
it
clear.
Is
that
so,
if
we
cut
way
to,
how
long
are
we
going
to
maintain?
We
want,
for
example,
like
the
sdk,
how
we
maintain
both
and
for
how
long.
A
A
I
guess
it
comes
back
to
the
question
of
what
does
it
mean
like
to
be
v1
or
v2
of
the
specification?
There
are
a
couple
things
like
we're
talking
about
field
of
freezers
list
specifically,
so
the
resource
list
object
would
get
a
v2
version
and
as
long
as
like
the
sdk
could
probably
flip
straight
to
using
it.
And
if
you
need
the
old
version,
you
can
use
the
whole
version.
A
We
can
for
sure
review,
what's
in
here,
the
rest
of
it
besides.
Here.
A
Since
resource
list
is
otherwise
very
simple,
like
it's
a
function
config
in
the
items
which
that's,
that
is
what
it
is
like,
I
don't
think
there's
any
possibility
for
this
to
change.
The
result
is
the
most
complex
part,
and
we
were
just
talking
about
how
neither
of
us
actually
uses
that
personally.
A
So
if
carlos
any
fun
have
more
feedback
on
what
we
could
improve
in
there,
then
that
would
be
great
to
bundle
in
with
a
v2
update
for
sure
or,
if
there's
other
folks
who
listen
to
the
recording
of
this
meeting,
perhaps
and
and
are
using
it
and
can
chime
in
with
their
feedback.
D
I
actually
have
a
question
about
what
the
call
about
the
use
cases
he
described.
So
are
you
planning
to
provide
the
users
the
functions
or
you
only
handle
the
result
on
like
filtering
the
result.
B
So
let's
say
that
the
users
will
run
a
tool
that
we
can,
that
will
consume
all
the
resources
to
make
to
collect
a
bunch
of
result,
objects
and
the
results
are
meant
to
be
something
like
linter
errors
or
warnings.
So
it'll
behave
kind
of
like
a
linter
and
the
tool.
B
D
B
B
D
So
you
provided
the
tool
that
to
write
the
instantiate,
the
result
result
object
and
in
the
current
value
you,
the
current
value
is
what
the
user
provides,
and
the
proposed
value
is
your
system's
response,
suggesting
users
how
to
fix
that
problems
and
that
result
proposed
the
value
doesn't
have
to
be
of
the
same
type
as
the
current
value
and
in
your
tool.
You
can
define
a
different
kind
than
current
value
and
expose
that,
with
a
more
I
mean,
a
more
mean,
more
complete
message
to
users.
C
D
Different
contain
attributes
and
other
attributes,
so
basically
this
is
customized.
You
provide
the
attributes
in
your
custom
struct
and
the
users.
They
can
read
that
message
either
is
rendered,
but
is
measured
by
yamu
or
json
or
in
their
client-side
tool.
They
can
get
details
from
the
proposed
value.
B
So
I
I
think,
let
me
see,
if
I
understand
correctly
so
essentially,
because
I
can
define
my
own
types
on
my
own
tool
and
I
also
own
the
tool
that
will
display
the
something
based
on
the
results.
Then
are.
B
Spec
yeah,
so,
if
understood
really
you're
saying
the
spec
doesn't
need
to
change,
because
I
can
essentially
treat
their
proposed
value
the
way
it
wants.
My
only
concern
with
that
is
that
my
understanding
is
that
the
karen
spec
needs
to
be,
let's
say
generic
or
global,
or
I
don't
know
like.
I
should
be
able
to
just
pick
some
tool,
that's
open
source
or
whatever,
that
of
the
shelf.
That
will
generate
results,
and
I
believe
that
I
should
be
able
to
use
the
same
if
this
tool
is
very
proposed
value.
B
I
believe
that
you
should
be
able
to
do
the
same
treatment
for
my
own
tools,
because
it's
implementing
a
common
spec,
and
that
will
mean
that
tools
will
behave
differently
if
I
trade
this
proposed
value
field,
especially
although,
if
that's
not
a
concern,
then
I
guess
that's
fine.
My
my
only
concern
is
that
it
will
break
the
commonness
or
global
lineage
of
the
spec,
because,
even
though
it
will
be
the
same,
spec
tools
will
take
this
value
differently.
B
D
A
Yeah,
the
the
spec
itself
is,
is
a
convention,
it's
not
a
go
thing
per
se
right,
so
I
agree
with
carlos
that
semantically
proposed
value
means
a
single
value
that
they're
proposing
to
replace
like
that.
The
author
of
the
function
is
is
saying
you
should
use
exactly
this.
A
So
if
you
go
beyond
that
and
say
well,
I'm
going
to
interpret
arrays
as
actually
a
list
of
options
instead
of
as
a
single
array
value
that
the
function
author
wants
me
to
use,
as
is
then
that
goes
beyond
the
spec
and
we
introduce
ambiguity
and
how
to
interpret
what
a
result
is,
which
I
think
is
not
ideal,
because
it
there
are
use
cases
for
this,
that
are
a
brown
field,
validation
and
where
the
type
correctly
wouldn't
match.
A
Because
say
it's
supposed
to
be
an
array,
the
field's
supposed
to
be
an
array,
and
they
gave
us
a
single,
a
single
number
and
the
what
the
result
is
supposed
to
be
describing
is
that
you
shouldn't,
have
you
know
into
12?
You
should
have
array
and
12..
That's
that's
the
desired
thing,
so
I
think
I
don't
know
so
go
ahead.
A
D
A
Well,
this
is
an
open
specification,
we're
talking
about
the
specification
itself
right,
so
in
theory
yeah
you
have
you
have
somebody
who
is:
who
is
the
provider
of
the
function
that
is
implementing
the
function,
that's
generating
results,
and
then
you
have
somebody
potentially
from
a
completely
different
context
and
community
and
who
is
processing
the
list
of
results
from
a
diversity
of
different
functional
providers,
amalgamating
them
into
a
single
list
where
hopefully,
result
means
the
same
thing
across
all
of
them,
because
they're
all
following
the
spec.
A
Well,
it's
not
struct
rated
we're
talking
about
functions
as
a
whole,
compliant
functions,
so
apple's
thing
like
they're,
probably
I'm
guessing,
building
lots
of
internal
functions
that
are
that
are
being
used
for
this.
This
product
they're
talking
about,
but
also,
if
they're
compliant
with
the
specification.
A
I
am
showing
the
go
imp,
implementation
in
the
function
framework,
but
the.
Where
is
that
if
I
go
back
far
enough,
I
think
I
already
had
it
open
here.
A
C
Maybe
we
should
rethink
a
little
about
how
often
how
likely
it
would
be
in
reality
for
us
to
have
several
proposed
values,
because
if
it's
just
our
edge
case,
maybe
it's
not
worth
to
do
this
whole
change
of
conventions.
But
if
this
is
quite
popular
in
the
industry
for
other
companies
as
well
for
current
function,
users,
then
maybe
it
would
make
sense-
and
I
do
agree
with
katrina
carlos-
that
the
convention
should
not
introduce
ambiguity.
C
A
I
I
thought
I'd
have
new
thoughts
to
the
point
about
how
difficult
this
would
be
to
maintain
in
the
functions
framework
itself,
and
I
think
it
could
actually
be
quite
easy
because
we
don't
have
the
struct
type
to
a
particular
thing
and
we
could
use
unmarshalling
as
a
as
a
like
defaulting
hook
and
like
when
we're
on
marshalling
it.
A
We
could
make
sure
that
the
fields
marshalling
and
unrushing
to
make
sure
that
the
fields
look
the
same
based
on
the
api
version
that
we're
given
so
it'd,
be
a
matter
of
tracking
like
did
the
is
the
object
coming
in
v1
or
v2,
and
interpreting
the
field
appropriately
and
making
sure
both
struct
fields
are
pop
or
populated.
Out
of
that,
or
not.
A
I
don't
know
how
popular
it
is
because,
as
far
as
I
know,
you're
the
only
ones
who
are
actually
using
this
feature.
So
that
means
it's
not
very
popular,
but
it
also
means
that
your
experience
is
the
most
valuable
experience
we
have
as
to
what
it
is.
What
is
useful
here-
and
we
do
want
to
make
it
more
useful,
it's
easier
to
change
early
on
when,
when
you're,
the
only
user
right.
C
That
makes
sense,
I
didn't
know
we're
the
only
user,
I
I
don't
know
for
sure.
Either
who's
come
forward.
E
C
Wonder
yeah,
that's
the
it's
probably
too
early.
I
was
gonna
propose
if
you
have
a
user
base
and
then
we
could
do
like
a
poll
and
then
ask
people
about
because
it's
very,
I
think
this
change
is
not
like
a
right
or
wrong.
Change
is
more
or
less
opinion
based
like
we
think
this
might
be
better
kind
of
change,
so
it
would
be
good
to
talk
to
the
user.
But
if,
if
we
don't
know
many
other
users,
then
we're
back
to
ground
one,
we
don't
know
what
to
do.
C
A
I
I
don't
think
it
would.
I
I
guess
if
we
could
somehow
maybe
with
a
github
search,
that's
obviously
not
going
to
turn
up
anything
private
like
it
wouldn't
have
turned
up
your
project,
but
we
declared
this
v1
for
the
purposes
of
kept
and
customized
collaborating
on
it.
Neither
of
those
tools
actually
uses
this
portion
of
the
spec.
A
But
if
we
want
to
be
pragmatic
and
we
can
determine
with
reasonable
amount
of
certainty
that
it's
not
going
to
cause
anybody
any
pain
and
maintaining
a
v1
and
a
v2
and
going
through
the
formal
process
of
cutting
that
just
on
principle
would
actually
cause
us
maintenance.
Then
then,
maybe
it's
worth
considering
munchie.
I
wonder
if
you
have
a
strong
opinion
about
that.
E
So
I
don't
have
a
strong
opinion
of
that
so
in
cap.
So
if
a
functions
use
this
field,
we
like,
we
will
show
it
up
in
the
eventual
result,
but
for
the
functions
that
we
built,
we
are
not
using
this
field.
A
So
the
captain
will
output
it
through
if
it
shows
up,
but
if
they,
but
you
don't
you
don't
manipulate
it.
In
other
words,.
E
A
A
Because
it
would
be
definitely
definitely
easiest
to
just
update
it
and
make
it
make
the
improvement
for
everybody
in
place.
A
Okay,
unless
anyone
has
anything
to
add,
I
think
I
think
google
is
the
the
other
place
that
would
have
the
most
potential
for
finding
other
users
if,
if
there
are
indeed
any
that
this
would
affect
so
I
think
getting
that
info
from
munchie
is
the
next
step.
A
A
All
right,
so
I
just
have
two
cups
here
that
are
open
one
is,
is
one
I
brought
up
last
meeting
where
it
was
my
action
name
for
last
meeting
to
make
this
pr
essentially-
and
there
was
one
point
of
discussion
assuming
it
didn't-
make
progress
this
morning
again,
there's
one
point
of
discussion
that
we
didn't
talk
about
at
last
meeting,
which
is
the
change
of
the
name
from
care
and
function
to
care
and
function.
A
Definition
in
the
type
and
the
reason
I
wanted
to
do
that,
as
I
explained
here,
was
to
strengthen
the
analogy
to
crd,
because
what
we're
defining
here
like,
if
you
think
about
it
in
terms
of
customer
resources,
you're
not
defining
a
custom
resource
you're,
defining
sdrd,
which
is
why
making
all
these
changes
to
the
fields
themselves
makes
sense.
A
So
I
thought
it
would
just
help
with
clarity
if
we,
if
we
made
that
analogy
really
explicit
by
reflecting
it
in
the
name
as
well,
and
since
nothing
depends
on
this
yet
now
seemed
like
the
time.
So
that's
why
I
was
proposing
that
in
here
as
well.
I
thought
it
would
be
worth
discussing.
So
we
can
move
this
pr
forward.
If
that's
the
only
point
of
contention.
A
And
this
is,
it
shows
up,
it
doesn't
even
really
show
up
in
catalog,
because
catalog
just
has
a
list
of
these
things.
It
doesn't
actually
have
the
the
type.
So
this
really
shows
up
in
the
function
registry,
I
think,
is
probably
it's
probably
the
main
place.
It's
a
function
registry.
When
you
define
your
function,
you
define
the
full
type.
A
So
it's
not
even
a
huge
deal,
I
think,
but
I
don't
know,
I
thought
it'd
be
clearer.
This
way.
E
Okay,
that
I
think
that
sounds
fine
to
me
so
and
then
I
think
we
can
move
forward
with
this
pr
great.
A
The
other
one
is
super
old,
so
I
just
wanted
to
put
it
back
on
your
radar
and
jeremy's
journey.
It's
not
able
to
make
it
today,
but
oops
wrong,
one.
I
just
clicked
on
the
same
one
I
had
open
here
we
go
this
one.
This
is
mainly
changes
from
a
discussion
we
had
in
november.
I
want
to
say
super
long
time
ago,
with
a
little
bit
of
an
addition,
so
like
this
runtime
provider
stuff,
that's
what
we
discussed
a
super
long
time
ago.
A
A
A
Somebody
asked
us
about
this
in
our
slack
group
about
two
weeks
ago
and
I
think
would
be
great
if
us
group,
we
all,
went
through
this
and
editor
added
our
thoughts,
either
in
slack
or
on
the
issue,
because
it
does
seem
extremely
related
to
our
sub
project.
A
The
use
case
that
they
have
is
very
analogous
to
not
only
care
and
functions,
but
also
pipelines
of
cm
functions
so
how
to
chain
them
together.
They
have
a
server-side
use
case,
but
it's
still
fundamentally
that
same
like
how
do
you
combine
functions?
How
do
you
invoke
them?
A
So
those
parallels
to
kept
this
comparison,
customize
and
there's
potentially
even
a
tie
into
the
alternative
execution
engines
issue
that
I
remember
was
floating
around
somewhere
on
the
kept
repo,
I
believe
a
few
months
ago,
so
yeah,
nothing
particular
to
say,
because
this
is
very
long.
Please
read
the
details
and
get
back
to
this
person
with
your
opinions
either
here
or
in
slack.
A
All
right,
so
with
that,
I
guess
we
move
on
to
stand.
Ups
functions
registry.
F
Yeah,
so
I
added
this
item
under
standup
just
to
make
it
known
that
we
have
a
gcp
project,
sig
cli
owned
with
me,
katrina,
jeremy
and
meg
chi
as
the
owners
and
I'm
planning
to
make
it
build
our
docker
images
and
stage
our
and
stage
them
before.
We
publish
them
officially
yeah,
so
that
exists
great.
A
I
saw
you
also
had
a
lot
of
progress
made
on
the
functions
repo
as
well
lately,
so
that
was.
F
Yeah,
so
the
we
donated
the
helm
chart
function,
there's
ci
that
runs
the
end
to
end
tests.
I
made
another
pr
that
will
run
unit
tests
and
another
one
that
I
saw.
You
commented
on
that
fills
out
all
the
readmes,
so
I'm
working
on
those.
F
A
What
else
do
we
typically
have
in
stand-ups
here?
It's
nothing
for
catalog
or
composition
today,
how
about
function?
Sdk.
E
Oh
thanks
for
a
review,
so
I
haven't
got
a
chance
after
that
to
address
the
comment,
so
I
will
try
to
address
the
comments
in
the
next
few
days.
A
Okay
sounds
good:
if
that's
it,
then
then
I
think
we
can
finish
up
early.
A
Thank
you,
everyone
for
coming
today
and
it's
great
to
see
you
all
and
hope
to
see
you
again
in
a
couple
weeks.