►
From YouTube: WG API Expression Bi-Weekly Meeting for 20210119
Description
WG API Expression Bi-Weekly Meeting for 20210119
A
I
put
something
on
the
agenda
we
are.
We
were
asked
to
provide
a
community
meeting
annual
report.
I've
had
this
on
my
list
for
quite
a
while.
I
did
some.
We
got
some
comments
from
jordan,
it's
something
probably
anton
and
I
have
to
to
look
over
again.
I
updated
our
release
target
for
ga,
but
if
you
have
anything,
you
want
to
put
into
produce
artifacts
or
what's
on
our
roadmap,
feel
free
to
to
add
things.
B
A
B
Yep,
so
I've
been
for
the
past
couple
months
and
then
on
and
off
for
a
little
while
I've
been
working
on
the
idea
of
an
idl
or
interface
definition,
language
for
kubernetes.
B
B
B
So,
like
you
know,
there's
no
native
representation
for
a
lot
of
our
types.
It
allows
you
to
write
a
lot
of
stuff
that
we
don't
usually
allow
in
kubernetes
apis
like
floats
for
instance,
and
so
it
would
be
good
to
have
the
90
out
from
that
perspective,
because
in
an
idl
and
in
a
specific,
like
domain-specific
language
for
writing,
kubernetes
apis,
you
could
instead
represent
these
concepts
natively
and
just
not
have
the
concepts
that
you
don't
need.
B
Similarly,
people
that
are
writing
trying
to
write
extensions
so
like
crds
in
languages
that
are
not
go,
are
running
into
issues
because
they
don't
really
want
to
have
to
install
an
entire
go
tool
chain
just
to
compile
their
types,
but
also
they
don't
really
want
to
write
open
api
by
hand
because
that's
clumsy
and
complicated,
and
it
actually
turns
out,
especially
with
some
of
the
restrictions
we've
put
on
how
you
can
write
open
api
schemas
in
crds.
It's
really
hard
to
write
some
things
by
hand,
so
an
ideal
will
be
useful
from
that
perspective.
B
So
one
of
the
things
that
I
kind
of
wanted
to
get
out
of
this
is
I'm
kind
of
looking
for
initial
feedback
and
if,
if
anything
is
is
missing,
so
I
have
a
I
have
a
repo
with
the
stuff.
I
have
currently
that
I
will
post
in
a
minute,
but
I
figured
I'd
just
kind
of
walk
people
through
stuff
quickly,
so.
B
I
converted
part
of
core
v1.
I
have
a
conversion
of
all
of
core
v1,
but
there's
a
lot
of
core
v1,
so
I
figured
I
just
highlight
the
interesting
stuff
so
in
the
idl
kind
of
like
in
kubernetes,
everything
is
in
a
particular
group
version,
and
within
that
group
version
we
can
declare
a
number
of
kinds.
So
kind
is
a
first
level
concept
should
make
it
easier
for
code
generators
to
like
look
specifically
for
where
to
start,
as
opposed
to
kind
of
using
like.
B
Does
it
embed
these
particular
types
to
try
to
figure
out
what
what
what
types
or
kinds,
and
it
means
that,
like
users,
don't
have
to
you,
know
manually
embed,
object,
meta
and
type
and
stuff
like
that,
and
then
we
can
declare
fields
with
all
kind
of
the
types
you'd
expect.
B
We
have
structs,
we
have
list
maps
and
lists
and
sets,
and
all
that
jazz
we
can
mark.
We
have
like
an
actual
first
class
concept
of
optional.
Instead
of
having
to
rely
on
some
combination
of
like
does
it
have
json
tags?
Does
it
have
the
slash
plus
comment,
markers,
etc
and
also
built-in
support
for
defaults.
B
We
have
support
for
validation,
of
course,
and
then
we
have
a
couple
of
kind
of
new
types
that
are
have
been
previously
implicitly
represented
in
the
kubernetes
type
system,
a
lot
so
like
enums
with
specific
values,
but
now
we,
this
introduces
a
first-class
kind
of
construct
for
them,
which
should
allow
us
to
have
better
api
documentation,
make
it
so
that
you
don't
have
to
manually,
define
constants
and
then
separately
also
define
the
validation
when
you're
writing
crds
and
then
also
kind
of
we
have
a
syntax
for
unions
as
well
of
which
we
have
a
number
in
kubernetes.
B
We
just
don't
mark
them
as
such,
so
things
like
volume
source
are,
I
think,
a
great
example
here.
So
I
have
a.
I
have
a
working
compiler
if
people
want
to
check
this
out
the
compiler
kind
of
works
in
two
phases.
B
The
first
phase
is
that
you
compile
to
an
intermediate
form
which
kind
of
compiles
into
proto.
So
we
can.
I
can
show
you
that
it's
not
much
to
show
it's
assuming.
I
didn't
break
anything,
it
does
what
it
does.
B
So
you
can
pass
the
some
number
of
paths
that
you
you
want
to
search
for
imports
from
if
you,
if
you
use
imports
and
you
pass
an
input
file
and
it
splits
out
a
lob
of
proto
and
then
that
on
standard
out
it
puts
out
the
text
form
of
proto
the
proto
for
just
for
debugging
purposes.
And
you
can
see
we
end
up
with
end
up
with
some
proto,
which
is
definitely
not
human
readable.
B
You
can
see
we
have
what
kind
of
what
you'd
expect.
We
have
validation.
You
can
see
where
we
specified
the
enum.
It
generated
enum
variants
for
us
where
we
specified
list
maps
it
generated
the
appropriate.
B
You
know
validation
and
kubernete
with
kubernetes
extensions
and
then,
where
we
specified
the
kind
of
tagged
union
it
gave
us
this
wonky
blob
of
open
api.
That
would
be
annoying
to
write
by
hand,
but
that
actually
enforces
that
we
use
the
union
in
the
right
way.
B
All
right,
that's
that's
what
I
had
to
demo.
I
will
post
the
repo
in
the
meeting
notes
in
a
bit
if
people
want
to
like
comment,
try
it
out
file
issues
stuff
like
that
and
hope
to
have
a
cap
within
the
next
week
week
or
two:
does
anybody
have
any
questions
or
anything
comments,
concerns.
A
A
C
A
Don't
worry:
do
you
think
that
this
would
help
us
or
could
help
us
to
to
also
describe
defaulting
and
validation,
that
we
can't
describe
right
now,
like
more
complex
validation?
It
still
needs
to
be
go
codes
that
we
could
generate
from
that
and
also
document.
B
It's
it's
possible,
I
it's
it's
one
of
the
things
that
I
think
would
be
interesting
to
experiment
around
with
I'd.
Also
love
to
have
like
feature
parody
between
between
that
and
crds.
So
but
yeah.
I
I
think
we
could.
Maybe
we
could
maybe
look
at
that,
especially
as
like
a
a
follow-on.
B
Something
like
that.
Oh
I
I
forgot
to
mention,
but
there's
also
like
a
kind
of
an
annotation
mechanism
in
the
idl.
So
hopefully
the
idea
is.
We
could
use
that
also
for
some
of
our
stuff,
like
deprecated
or
like
having
specific,
like
attaching
specific
metadata
to
fields
for
like
feature
gates
and
stuff
so
that
we
could
go
through
and
do
automated
checks
or
generate
automated
documentation
on
particular
fields
or
whatever.
C
B
Definitely
think
we
could.
We
could
do
something
like
that
with
the
generated
go
defaults
at
some
point.
A
B
Yeah,
that's
that's
correct.
I
have
that.
I
have
that
code.
I
haven't
I
actually
I
did
a
prototype,
but
for
for
various
reasons
the
prototype
was
not
and
go
so
I
have
that
code
not
in
go,
and
I
just
need
to
port
it
to
go
but
yeah
this
would.
This
would
generate
type
seco
as
well.
C
Yeah,
thank
you
sally.
It
was
very
good
to
see
that
I
don't
have
anything
special.
I
don't
want
to
take
much
time,
just
a
quick
update
on
the
requirements.
We
have
updated
the
api
clients
task
because
it
is
a
priority.
It
blocks.
Our
api
client
go
work,
so
jeffrey
is
now
working
on
the
pr
for
that.
I
I'm
sure
he
doesn't
have
any
update
yet
because
he
just
started
last
friday,
but
we
will.
C
C
It
is
just
for
90
percent
of
the
time,
and
we
need
that
fix
and
jenny
will
focus
on
that
and
jeffrey
is
working
on
the
pr
for
client
go
and
we
will
share
that
pr
soon
with
people
who
want
to
try
it
from
the
cube
builder
community
and
to
get
the
results
from
which
solution.
We
should
go
forward
with
the
structure,
the
builders
and
that's
all
yeah.
Thank
you.
C
And
we
don't
have
the
decision,
because
we
are
waiting
for
the
pr.
So
every
anyone
who
could
try
one
of
the
forks
for
the
builders
or
for
the
struts
could
know
what
to
do
so
that
be
that
pr
is
not
ready.
Yet
as
soon
as.
B
C
The
priority
so
whenever
we
we
got
one
volunteer
like
asking
for
more
questions
from
cubeler
meeting
that
we
had
last
week,
joe
presented
this
without
a
pr
but
just
presented.
What
we
need
from
those
who
are
volunteering
to
give
us
some
idea
of
what
is
better
to
decide.
And
I
am
keeping
track
of
like
the
names
and
we
will
add
more
people
and
we
share
the
pr
with
them
to
have
the
decisions.
A
Okay,
yeah,
I
I
thought
we
are
joe
already
had
like
working
prototypes,
so
I
I
didn't
know.
C
It
is
a
prototype,
but
there
are
some
yeah.
The
solution
is
ready.
Jeffrey
can
talk
about
what
he
needs
to
work
on,
but
there
are
some
challenges
between
like
creating
two
different
forks
one
for
structs
one
for
builders
and
the
solution
is
there,
but
for
anyone
to
try
it,
there
are
some
complexities
that
jeffrey
is
working
on
to
resolve
that
and
then
people
can
easily
run
the
code
and
see
which
one
is
their
preference.
E
So
joe
has
a
working
fork
for
built-in
types
for
kk.
I'm
currently
working
on
a
port
of
that
for
crds
to
generate
the
like,
struck
base
and
like
to
generate
the
clients
for
crds,
and
the
pr
is
actually
mostly
ready.
I
think
next
step.
I
just
need
to
work
out
a
few
corner
cases
with
joe
and
we'll
probably
be
ready
to
send
out
a
survey
soon
to
get
our
feedback.
A
Yeah,
I
I
know
I
think,
two
places
that
were
mentioned
inside
openshift
right
that
that
would
want
to
try
apply.
So
I
I
I
sent
forward
the
I
sent
them
the
cap
and
if
we
have
a
working
pr
for
the
clients
I
can,
I
can
ask
them
to
try.
Maybe
suggest
it
would
be
pretty
cool
to
see
it
in
a
controller
in
action.
D
D
Oh
sorry,
I'm
probably
going
to
fix
that,
and
so
that
is
basically
there
is
a
existing
but
unfinished
cap
kept
for
the
immutable
field,
and
I
talked
with
andrew
and
I
picked
it
up
with
my
starter
project.
D
So
I
updated
that
kept
in
a
google
doc
and
it
says
it's
better
for
communication
in
the
comments
and
I'm
currently
is
doing
the
draft.
I
I'm
gonna
stand
out
for
the
review
and
and
probably
have
some
conversation
with
you
and
trying
to
have
more
context
on
that
area,
but
just
give
us
status
report,
I'm
working
on
that
and
when
sent
kept
in
google
doc
is
finalized
and
approved.
I
will
module
back
to
github
work.
Sorry
on.
A
Okay,
I
don't
know
if
we
I
can.
Let
me
let
me
share
the
ga
requirements,
real
quick
and
we
can
go
over
them
as
well
find
them.
Here
we
go
yeah
you
already
update
on
the
api
clients.
We
have
a
meeting
later
about
the
sub
resource.
I
don't
know
if
antoine
will
make
it
but
yeah,
I
know
andrea,
is
working
on
it.
A
Yeah,
if
not,
I
I
I
think
we
need
him,
because
my
my
order,
yeah
hiding
managed
fields
in
api
server
responses.
I
don't
know
if
notice
here
doesn't
look
like
it.
My
last
state
was
we
have
a
cap
that
is
merged
or
in
the
process
of
being
merged.
A
If
github
is
loading.
Okay,
now
the
cap
is
still
open.
I
think
it
has
a
few
outstanding
comments
and
then
this
can
be
merged.
As
implementable
yeah
danielle
has
to
take
another
look
and
yeah.
We
should
definitely
make
that
pretty
soon
like
implementation
is
mostly
done
as
far
as
I
know
so,
we're
good
is
julian.
Here,
no.
A
Yeah
here
about
the
mutating
admission
controllers,
I
have
a
pr
because
I
needed
a
pr
because
locally
the
tests
didn't
work.
For
some
reason.
It
breaks
right
now
and
I
don't
really
understand
why
so
far,
so
I
have
to
give
it
another
look
in
general,
it
shouldn't
be
a
big
problem,
but
we'll
see
I
get
some
weird
errors
and
I
have
to
find
out
which
what
they
mean
but
yeah,
I'm
on
it.
F
E
Oh
sorry,
I
was
muted
yeah,
so
we
have
another
googler
kevin
who's,
I'm
going
to
be
working
on
this
and
I
think
we
sort
of
found
the
problem
we're
still
working
on
a
solution
for
this
but
yeah
this
is
taking.
We
have
some
work
in
this
right.
E
A
All
right
yeah,
instead
of
swiping
jenny,
if
you're
looking
at
it,
feel
free
to
check
me
up
it's
it's.
It's
weird.
C
Yeah
I'll
hit
I'll
message
you
about
it.
After
I
take
a
look
at
the
current
status.
A
E
Oh
just
want
to
add
one
thing
for
the
conformance
tests,
so
I
was
able
to
successfully
report
a
couple
of
our
integration
tests
to
ewe
tests.
It
seems
like
the
port
is
pretty
trivial.
I
created
a
pr
for
it,
but
I
think
at
least
on
the
conformance
side.
It's
not
gonna
be
too
difficult
to
set
up
the
tests,
given
that
our
integration
coverage
is
pretty
good,
ready.
E
A
All
right
looks
good,
then
thanks
everybody.
If
there
is
anything
else,
just
post
it
on
slack
and
thanks
for
coming
by.