►
From YouTube: WG API Expression Bi-Weekly Meeting for 20210525
Description
WG API Expression Bi-Weekly Meeting for 20210525
A
All
right
welcome
to
working
group,
api
expression,
it's
the
25th
of
may
2021.
We
have
some
agenda
items.
I
just
wanted
to
bring
up
the
examples.
Cap
again
it's
a
there
is
some
was
some
feedback
on
it
and
he
also
mentioned
it
was
read
through
yesterday.
I
think
in
the
cap
reading
I'm
still
split
between
the
kind
of
three
approaches
we
have
like
the
per
resource
example.
A
We
originally
talked
about
for
cubesat
generate
the
super
field,
one
that
is
in
the
cap
right
now
and
like
a
wouldn't
want
suggested,
the
per
structure
example.
So
we
can
at
least
create
sensible
example,
objects
from
it,
but.
B
A
A
Authors
would
be
interested
to
make
it
easier
to
create
the
resources
in
a
from
from
from
a
cli
or
possibly
ui,
and
that
was
having
the
examples
field
in
the
open
api,
which
is
a
map
like
per
per
resource
and
a
per
field,
and
you
could
have
multiple
of
those
like
one
that
is
templatable
and
one.
That
is
just
plain
simple,
and
that
was
the
original
idea,
and
you
mentioned
the
the
fields
and
that
made
quite
some
sense
when
I-
and
I
didn't
know
that
exists-
and
I
stumbled
upon
it
while
working
on
a
crd.
A
C
And
so
it's
going
to
be
difficult
to
convince
people
that
we
need
to
have
some
specific
templating
semantic
added
to
the
examples,
but
I
do
agree
with
you
that
we
sort
of
need
the
templating
I've
see
I've
seen
something
somebody
worked
at
google
on
something
like
these
and
it's
it's
pretty
it's
pretty
cool,
because
so,
as
you
type
in
so
you
have
the
default,
let's
say
deployment
being
created
and
then,
as
you
type
the
name,
it's
actually
also
modifying
the
application
in
the
labels
in
multiple
places,
in
the
object,
as
you
type
it,
and
so
you
can
see
how
the
fields
are
connected
between
one
another,
so
that
I
think
is
templating.
A
Yeah,
I
the
first,
I
think,
the
the
examples
field
we
looked
at
was
a
map
and
you
can
like
provide
multiple
examples
in
the
open
api,
at
least,
and
what
the
idea
was.
We
could
start
simple,
just
add
the
field
as
supported
and
set
like
smart,
a
simple
example
for
parts
and
then
once
we
decide
on
templates
or
what
how
we
would
want
to
do
templates,
we
can
add
another
example
to
the
resource
we
want
to
template
or
crds
can
use
it.
A
They
can
do
whatever
they
want,
and
but
the
mechanism
would
be
supported
and
I
could
already
get
like
pushback,
because
it's
it
opens
up
a
lot
of
data
being
added
to
the
open
api.
But.
C
A
Yeah,
I
think
the
tricky
part
is
differentiating
between
defaults
and
example
like
if
it's
if
the
field
has
a
default
and
the
default
should
be
like
the
example
in
the
default
field
should
not
be
the
same.
If
the
default
make
sense,
there
should
not
be
an
example,
maybe
like,
like
the
result.
The
result
of
an
example
should
be
something
you
would
actually
send.
A
C
A
Only
if
the
default,
the
example
would
be
different
from
the
default.
So
if
you.
D
A
Know
for
for
a
container
the
the
default
for
end
is
empty,
but
your
example
might
want
to
set
an
environment
variable
to
show
how
it's
done.
E
So
I
dropped
so
we
we
ran
out
of
time
in
the
meeting
yesterday,
but
I
dropped
a
link
in
the
chat
to
a
bunch
of
notes
that
I
scribbled
down
real
quick.
I
think
we're
going
to
go
over
this
cap
next
time
we
meet
with
the
kept
reading
club
again
so
hopefully
we'll
be
able
to
have
like
a
better
notes
and
discussions,
but
some
of
the
things
that
came
up
where
people
were
concerned
about
drift
and
how
do
we,
you
know,
keep
up
to
date
with
examples
and
drift
for
them.
E
Some
people
said
that
they
weren't
quite
sure
what
this
actually
was
basically
clarify
what
this
actually
is
in
the
cup.
I
think
it
might
have
just
been
the
russian
the
misreading
on
those
that
the
size
came
up
a
few
different
ways
about
splitting
it
down
and
being
able.
I
didn't
understand
this
one,
but
they
wanted
to
leverage
the
fields
for
testing
to
keep
the
size
down,
and
then
there
are
questions
about
read-only
fields,
how
they
are
represented.
E
What
would
this
be
for
example-wise,
and
then
someone
said
this
link
was
good.
So
that's
what
I
have
from
yesterday
but,
like
I
said,
hopefully
we're
to
give
it
another
shot.
Next
time
we
meet,
I
don't
get
the
arguments
about
breathing
new
fields.
E
It
was
just
correct
questions
about
how,
if
you're,
generating
from
fields
via
the
the
spec,
how
do
read-only
specific
fields
show
up
in
there.
A
E
C
Okay,
that's
why
yeah
the
drift?
I
think
it's
the
so
I
think
it's
kind
of
the
same
time
we
have
with
the
documentation
anyway.
The
good
thing
with
the
examples
is
that
we
can
actually
test
them.
I
think.
A
A
And
I
think
they
should
actually
decrease
drift
between
documentation
and
and
code
because
they
are
defined
on
the
code
level
and
if
a
best
practice
on
something
changes.
The
example
should
change
and
even.
A
So,
with
the
what
I
think
we
can
do
with
the
exam
parts
on
with
the
granularity
is
we
don't
necessarily
have
to
set
as
examples
for
every
single
field,
but
we
can
set
examples
for
the
parents
if
that
makes
more
sense
in
some
occasions.
A
C
E
Are
you
saying
document
what
a
pod
spec
itself
looks
like
and
reuse
that
everywhere
a
pod
spec
is
pulled
in
or
deeper.
A
A
Struct
or
on
the
containers
list
item
with
one
container
item,
but
then
we
would
have
to
build
an
example
for
the
entire
list,
and
that
is
not
necessarily
typed
or
that
there
is
a
risk
for
for
drift,
because
we
would
be
defining
the
structure
of
a
container
in
the
documentation,
an
example
part
and
not
generating
it
anymore.
So
probably
adding
one
item
to
a
list
might
be
the
best
and
overwriting
it
by
setting
it
to
empty.
If
we
don't
want.
C
C
Yeah,
I
don't
know,
I
think
you
know
what
I
think
we
need
to.
We
need
to
try
it
out,
because
it's
going
to
be
very,
very
hard
to
know
exactly
how
this
is
going
to
behave
until
we
see
it
run.
C
And
force
tracked
versus
fields,
I
think
it's
going
to
depend
on
the
specific
tracks
on
the
specific
fields,
and
so
we
should
leave
the
door
gun.
F
A
A
C
Yeah,
I
agree
with
that.
It's
it's
too
bad
that
we
can
only
give
one
example
yeah.
I
was
right.
Let
me.
F
C
It's
not
it's
not
part
of
the
schema,
it's
part
of
the
yeah,
that's
that's
where
they
did
that.
A
C
A
Here
I
mean
that
that
was
another
thing.
I
don't
know
if
you
remember
what
we
talked
about
was
if,
if
it's
not
wanted,
what
we
can
also
do
is
create
a
an
example
kind
that
can
be
created
for
linking
to
resources
like
it's.
It's
a
specific
kind
of
contains
an
example
resource
and
can
be
query
just
like
any
other
resource.
A
E
E
I
I
don't
know
something
like
deployment
with
one
pod
deployment
with
two
parts
right,
that's
a
terrible
example,
but
I
can't
think
of
a
better
one,
but
that's
that's
what
I
was
thinking
was.
There
was
a
way
that
we
could.
You
know,
name
space
this
and
allow
developers
to
like
list
or
auto
complete
those,
and
we
could
have
multiple
real
world
starting
points
that
people
would
actually
use.
E
F
C
C
A
A
A
A
A
C
A
Yeah,
that's
that's
a
open
discussion
item
if,
if
and
how
and
if
that
would
be
possible
with
template
examples
and
or
what
kind
of
templating
would
be
possible
and.
A
C
F
A
A
A
E
C
C
Yeah,
I
I
think
that's
that's
an
interesting
time
to
solve,
but
again
I
think
we
can
generate
it
from
the
example
values
in
the
object.
E
I
think
that's
fair
so,
and
this
brings
back
to
the
the
read-only
fields
that
I
brought
up
earlier
is:
if
you
run
this
command
and
you
look
at
the
output,
it
has
something
like
creation,
timestamp
null
status,
you
know
empty
object,
and
so
right
now
this
is
what
people
are
doing
and
then
they
go
through
it
and
manually
release
the
fields
that
they
don't
care
about,
so
anything
that
we
give
them
whether
it's
a
you
know
single
example
or
multiple
is
better
than
doing
this
yeah.
I
think
so.
E
Yeah,
I
think
that's
I
like
I,
I
built
a
original
plug-in
for
this
as
a
proof
of
concept.
I
call
it
generate,
but
it
really
doesn't
matter
to
me
what
we
call
it.
But
yeah
I'd
like
to
see
something
like
this
upstream
for
sure.
C
I
do
like
the
idea
of
having
the
the
plugin
for
now
and
you
how?
How
does
this
work
you
are
you
able
to
actually
do
something
with
it?
I'm
assuming
you
do.
E
C
Can
you
demo
it
the
next
during
the
next
meeting,
maybe
eddie
sure
I
can
do
that?
That'd
be
nice
and
if
you
have
the
chance,
maybe
to
look
at
whether
you
want
to
do
it.
Both
perk
doppler
field
of
the
entire
object
and
see
if
you
can
get
some
results
from
that,
maybe
that'd
be
cool.
E
A
All
right,
oh
yeah,
and
the
other
one
is
also
me.
We
discussed
what
we
oh.
We
talked
about
a
few
times
about
this
possible
cube
concentration
about
how
to
convert
an
operator
to
using
ssa
and
letting
people.
Do
it
a
bit
on
hands
or
hands
on
and
yeah,
and
I
sent
an
email
to
the
kubecon
team.
The
deadline
was
yesterday
or
sunday.
I
time
zone
differences
and
they
said
I
should
just
submit
it,
which
I
did.
A
There's
the
the
options
are
limited
right
now
they
only
have
one
speaker
talk
to
speaker
talk
and
panel.
There
is
nothing
in
their
concept,
that's
hands-on
or
or
interactive
or
like
a
breakout
session.
Sadly,
but
I
submitted
it
anyways
and
yeah.
I
left.
I
said
it's
like
a
two-speaker
and
left
one
blank,
so
I
I
I
don't
know
if
they
get
back
to
me.
They
said
they'll,
look
at
it
because,
right
now
they
don't
have
anything
practical
on
their
session,
which
is
on
their
timetable,
which
is
a
bit
sad
so
but
yeah.
A
If
you
are
interested
in
being
at
least
another
speaker
or
helping
with
it.
Please,
because
I
don't
know
if
I
I
want
to
do
that
alone,
if
there
are
more
people
and
we
they
want
to
convert
operators.
I'd
love
to
some
have
somebody
that
is
going
to
help
people
convert
operators.
D
Yeah
I'd
be
willing
to
help
kevin
with
whatever
and
happy
to
help
you
kind
of
figure
out
how
to
proceed
from
there.
If
you
have
things
already
in
mind,
then
you
know
happy
to
listen
and
chat
on
on
ideas
stuff,
like
that.
A
Yeah,
if
you
want
to,
I
can
see
if
I
can
update
the
talk
and
edit
your
info.
If
you
send
me
or
I'll
yeah
I'll
chat
you
up,
I
I
still
don't
know
if
it
like,
if
they
even
want
to
talk
like
this,
because
I
really
want
to-
or
I
at
least
prefer
to
do
something
more
practical,
because
I
just.
E
A
Okay,
okay
yeah,
maybe
maybe
that
works
yeah.
B
For
what
it's
worth,
this
is
alex
from
hachikore.
By
the
way,
I
also
submitted
a
talk
where
I
wanted
to
share
how
we
adopted
server
side
apply
in
in
terraform,
so
that's
gonna,
be
like
sort
of
like
application
like
side
of
things
story
about
how
servers
are
blocking
can
actually
be
useful.
A
A
Looking
forward
to
that,
I
I
yeah.
I
was
very
happy
about
the.
I
saw
the
blog
post
about
terraform
supporting
servers
that
are
playing
like
ooh
yay,
but
I
hadn't
have
no
time
to
look
into
the
actual
bits
and
pieces.
B
It's
it's
quite
early
for
us
as
well,
but
it's
working
wonders
for
for
basically
productivity
and
like
speed
of
of
development.
So
it's
it's!
It's
been.
It's
been
great.
So
far,.
C
B
B
I'm
I'm
also
open
to.
If
I
get
accepted,
I
don't
know
open
to
doing
like
a
shared
talk
with
anyone
of
you
who's
interested
in
sort
of
like
doing
the
the
the
sig
point
of
view.
In
the
same
talk,
I
don't
know.
C
I
personally
don't
feel
like
talking
right
now,
but
okay,
if
anyone
is
interested.
A
Yeah
I
mean
we
can
sync
up
on
how
like
yeah.
This
mission
goes.
I
I
haven't
talked
in
front
of
that
many
people.
I
think,
but
I've
done
one
two
talks,
but
yeah.
D
D
B
I
already
submitted
it
because
I
obviously
realized
the
deadline
is
coming
on
on
sunday
and
I
did
it
solo,
but
I
guess
we
can
amend
it
to
add
someone
else
on
it.
F
A
E
Yeah,
the
maintainer
tracks
that'll
be
sent
out
to
the
the
weeds,
alias
at
some
point
and
it's
a
totally
separate
cfp.
So.
A
A
Good
yeah,
that's,
I
think,
that's
it
for
mind
who's
the
guideline
of
controller
with
ssa,
that's
getting
easier.
F
G
Not
from
me,
actually
I
I
found
a
link,
I
don't
know
if
that's
a
whole
document
or
not,
but
I
think
that
might
be
helpful
for
controllers
using
source
applied
it
that's
the
official
line
as
a
guideline.
I
guess
it
to
me
seems
a
little
bit
too
simple.
G
But
I
I
I
wonder
if
anyone
is
working
on
a
detailed
guideline
about
how
controllers
should
be
using
software
supply
and
the
best
practices
and
such.
C
Oh
yeah,
no,
so
we
totally
have
a
blog
post
ready
for
122
for
how
how
to
do
that,
and
I'm
sure
we
can
share
it
with
you
and
you
can
if
we
haven't
already,
but
we
can
absolutely
share
it
with
you
for
a
preview,
and
you
can
tell
us
what
you
think:
okay
yeah,
that
would.
C
G
C
Now,
it's
probably
a
google
doc
and
it's
not
shared
because
it's
supposed
to
be
published
after
122.
A
At
red
hat,
but
everybody
so
far
was
I,
I
think
people
mostly
holding
back
until
it's
ga.
So
I
only
know
of
exploration,
but
nobody
actually
converting
anything
yet
so
anything
blog
posty
could
still
be
helpful.
C
Yeah,
no,
we
have
a
few
people.
People
have
started
to
go
on
that
wagon
in
google,
so
everyone
wants
to
shoot
to
see
the
document
and
now
how
am
I.
A
A
A
A
We
have
to
work
items
that
I
think
at
least
I
have
still
haven't
touched
since
last
meeting.
Do
you
want
to
clean
up.
A
C
A
It's
still
funny
because,
like
I
read
the
email
about
this,
when
it
was
sent
a
few
weeks
ago
and
a
week
later,
I
caught
in
a
discussion
because
we
actually
have
a
discussion
or
a
field.
That
is
a
reference
and
it's
not
done
properly,
and
I
was
like,
oh
see,
wait.
I
have
a
discussion
here,
how
it
should
be
done
from
upstream,
and
I
could
use
it
as
a
suggestion
to
fix
how
we
do
references.
B
First,
is
one
of
my
users
ran
into
an
issue
where
we
do
a
dry
run
in
case
the
crd
doesn't
have
structural
information
just
to
sort
of
like
check
the
old
that,
like
the
the
overall
saneness
of
the
manifest
we
do
a
dry
run
call
during
the
planning
phase,
and
if
that
crd
has
an
admission
controller
guarding
it,
then
the
admission
controller
may
throw
an
error
saying
that
dry
run
isn't
implemented.
B
C
F
C
Run
the
trail
run
on
a
web
hook.
That
is
not
ready
for
it.
The
problem
is,
we
want
to
avoid
side
effects.
So
if,
if
a
web
hook
is
going
to
change
the
state
of
something
on
a
request,
then
it
needs
to
know
that
the
try
run
the
request
is
dry
run
most
of
the
time.
It's
just
a
question
of
accepting
that
it
doesn't
have
any
side
effect
and
changing
the
value
of
the
in
the
configuration
and
just
say:
yeah,
that's
fine!
I
don't
do
anything
but
yeah
by
default.
B
You
got
it
well,
basically,
it's
it's!
The
web
hook
is
out
of
my
control
because
that's
the
user's
sort
of
domain
from
terraform's
point
of
view,
we're
just
being
martha,
cube,
cuddle
right,
so
we're
we're
just
applying
the
manifest
and
as
part
of
applying
the
manifest
we
wanted
to
do
a
dry
run
to
make
sure
that
it
breaks
in
the
planning
phase.
If
there's
syntactic
issues
in
the
manifest
rather
than
plan,
then
then
failing
the
apply
which
basically
messes
the
terraform
state.
If
it
fails
yeah.
C
B
C
Yes
and
maybe
ask
the
person
to
get
the
web
hook
fix
somehow.
B
C
Yeah
exactly
that's
unfortunate,
but
that's
the
only
thing
we
can
do
and
again
for
the
webhook
person.
It
should
be
easy
to
fix
and
if
it's
a
web
hook
that
you
don't
own
but
it's
owned
by
someone
else,
I'm
sure
they
will
be
happy
to
fix
it.
Because
it's
let's
say.
B
A
C
B
C
B
C
B
Yeah,
I
think
just
like
informing
the
user
is
good
enough
for
us
for
now.
That's
that's
one
thing
that
I
recently
ran
into
and
the
other
one
is
in
case
of
some
crds
non-structural
ones
included,
but
I
think
others
as
well.
I
don't
actually
have
enough
information
to
like
classify
them
more
precisely.
B
The
api,
the
open
api,
v2
endpoint,
doesn't
actually
include
type
information
for
them.
So
at
that
point
I,
the
the
background,
is
that
the
telephone
provider
actually
relies
on
the
type.
B
In
the
open
api
document
to
generate
the
type
information
for
the
internal
representation
in
terraform,
so
we
do
a
just
in
time:
conversion
between
open
api
and
reforms,
type
system,
and,
if
that's
not
present,
what
I'm
planning
to
do
is
basically
list
the
crds
off
the
cluster
and
fetch
the
open
api
information
directly
off
of
the
crd
object
itself.
B
So
that's
a
bit
of
work.
I
was
yeah.
C
We
were
working
on
that.
To
be
honest,
so
one
thing
that
jeff
is
not
here
to
mention,
but
jeff
is
working
on
making
sure
we
expose
the
open
api
v3
directly
from
the
cluster
instead
of
v2,
yeah
and-
and
I
don't
think
we'd
be
modifying
the
open
api
as
we
do
so
like
we
do
for
open
api
too.
B
I
see
that's
nice,
that
would
be
nice
and
is
there
any
like
time
timeline
for
that,
because
I'm
just
about
to
work
on
on
that
short
circuit
mechanism.
On
our
side,
it's.
C
It's
gonna
take
multiple
releases.
I
think
at
least
I
mean
it's
not
gonna
happen
in
122,
maybe
123.
If
we're
lucky.
B
Cool,
oh
and
the
other
thing
is,
I'm
not
sure
if
this
is
the
right
place
to
mention
that,
but
the
caching
rest
mapper
is
very
bug
written.
I
had
to
discuss
that.
B
I
think
you
opened
the
burger,
I
couldn't
look
at
it.
Yeah,
that's
that's
one
problem.
I
actually
ran
into
a
couple
of
into
a
different
one
in
the
meantime
for
which
I
didn't
get
a
chance
to
document
it.
It
it
randomly
sort
of
like
doesn't
discover
resource
types
in
in
certain
situations,
so
I
had
to
disable
it,
but
I
took
a
pretty
nasty
penalty
performance
penalty
without
it,
so
it's
actually
useful
in
our
case,
but
it
was.
It
was
too
unstable
to
to
keep
using.
At
this
point
we
are.
B
Actually
I
mean
it's
it's
that
important
in
our
use
case
that
I'm
I'm
actually
willing
to
put
some
effort
into
fixing
those
problems
we
just
have
to
stabilize
the
the
provider
first,
which
is
gonna,
take
a
couple
of
months.
I
guess
so
if
if
no
one
actually
gets
to
the
rest,
mapper
I'll
be
willing
to
have
a
go
at
fixing
those
things.
C
That's
awesome
and
if
you
do
so,
let
me
know-
and
I
can
help
you
and
help
you
navigate
things
if
needed.
B
Yeah,
that's.
That
would
be
my
my
my
ask
if
somebody
could
actually
like
on
board
me
to
it
and
something
yeah.
B
F
B
I
mean
it's
really
useful.
To
be
honest,
it's
I
haven't
run
into
any
issues
with
it,
except
for
the
rest,
mapper
thingy.
Everything
else
was
pretty
easily
handled.
A
So,
for
all
I
heard
from
the
world
out,
there
is
managed
fields
are
so
large.