►
From YouTube: WG Component Standard Office Hours 20200204
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
B
B
No!
I'm
I'm!
Just
like
the
weekend.
I
was
like
yeah
looking
into
the
whole
controller
manager
thing
a
bit.
I
feel
I'm
actually
feeling
I
get
like
more
proficient
every
time
I
look
into
it,
which
is
quite
nice.
B
Actually,
what
I
wanted
to
do
now
is:
I
wanted
to
just
like
briefly
go
with
you
over
the
kind
of
like
work
in
progress
like
ideal
sketch.
I
think
lucas
was
doing
back
in
the
day.
Okay
and
there's
quite
a
discussion
between
stefan
and
jordan.
Let
me
send
you
the
link.
Okay,
yeah.
Let
me
take
a
look.
C
D
B
Awesome
so
yeah
just
so
this
discussion
started
in
around
may.
I
think-
and
I
think
the
general
idea
is
so
they're,
like
they're,
proposing
a
couple
of
alternatives
here.
I
think
each
alternative
is
kind
of
like
kind
of
implies
that
every
controller
is
versioned
on
their
own
right.
So,
like
alternative
one
carrier
on
gvk,
I
don't
know
exactly
what
this
means.
This
question
didn't
get
answered:
yeah.
A
B
A
I
just
yeah
because
joe
is
just
sergey,
is
working
on
that
and
trying
to
fix.
B
Yeah
yeah,
I
found
it
interesting
that
we
could
use
this
argument
in
a
discussion
moving
forward.
A
Before
it
would
need
to
be
yeah,
it
would
need
to
be
a
different
metadata
instead
of
metadata.
It
would
need
to
be
the
standard.
A
A
B
So
yeah,
so
here
I
guess
we
define
stuff
for
the
controller
manager
itself,
things
like
client
connection,
any
controllers
that
are
disabled.
I
guess
I
don't
know
what
so
they
say
like
controller
common,
it
would
set
defaults
that
could
be
overwritten
by
each
controller.
B
I
don't
know
how
this
relates
to
like
the
cube
cloud,
shared
config
thing,
maybe
something
about
like
client
connections
or
so.
A
A
Really,
I
don't
really
know,
I
think,
there's
different,
maybe
different
approaches
like
one
would
be.
You
just
have
this
shared
config
that
sets
it
for
everything
another
one
would
be.
You
have
a
shared
config
that
sets
defaults
for
everything,
but
still
allow
individual
controllers
to
explicitly
override
the
shared
parameters
and
that's
worth
thinking
about.
B
Yeah
yeah,
maybe
this
we
can
incorporate
for,
like
the
you
know,
we
can.
We
can
hint
at
the
current
shared
fields
and
we
can
now
like.
I
don't
know
if
we,
if
we
not
name
it
common
or
something
else,
but
just
like
some
field
where
we
can,
you
know,
define,
defaults
and
stuff
can
be
overwritten
later
on.
B
A
B
A
I
think
it's,
the
idea
is
that
it's
like
some
arbitrary
extension
so,
like
I
don't
know
if
cluster
api
uses
this,
but
one
example
of
something
that
might
use
raw
extension
is
like
provider
config
in
a
machine
template
where
it
can
be
a
different
schema,
depending
on
which
cloud
provider
it's
deployed
on
so
there's
some
sort
of
late
binding
happening.
I
think,
with
the
actual
schema,
that's
expected
so.
B
A
B
Okay,
but
you
would
need
to
do
some
conversion.
B
I
had
said
here.
The
next
step
is
to
copy
using
conversion
into
the
internal
struct.
So
basically,
you
need
to
define
some
conversion
functions
to
work
with
this
raw
extension.
Yes,.
B
A
Just
you
need
since
go,
doesn't
support
like
some
types.
B
How
like,
if
you
default
anything
here
by
just
you,
know,
omitting
it,
I
guess,
does
it
always
default
like
to
to
which
type
does
it
default
to
right?
Oh.
A
By
that
comment,
where's
the
comment:
the
comment
about
defaulting
to
the
container
kind.
So
you
see
how
like
garbage
here,
yeah
so
see
how,
like
garbage
collection,
is
defaulting
to
garbage
collection,
controller
configuration-
maybe
I
didn't
say,
defaulted
there,
but
I
I
could
see
that
like.
It
takes
the
name
and
then
defaults
the
kind
to
something
based
on
the
name
in
the
map
right.
The
map
key.
B
B
B
So
for
this
one,
the
immersion
pen
map
key-
I
don't
really
like
see
how
this
is
different.
For
me,
it
looks
like
it's
just
some
sort
of
shorthand
so
included
here
and
the
kind
of
few
it
implies.
You
know
we
get
this
version,
it
doesn't.
A
Does
that
matter
right
like
so?
Yes,
it
is
syntactic
sugar
for
defaulting
the
version,
which
means
you
don't
have
to
type
the
version,
which
is
one
of
the
like
one
of
the
arguments
about
having
gbk's
ever
was
like.
Oh,
my
god,
people
have
to
type
all
this
stuff
right.
So
that's
maybe
like
an
idea
around
that.
I
actually
don't
think
it's
that
big
a
deal
for
people
to
have
to
type
it
out.
A
The
other
thing
that's
sort
of
interesting
about
this
solution.
That
makes
it
a
little
more
of
a
complex
implementation.
Is
that
I
could
say
garbage
collection,
v1
and
garbage
collection,
slash,
v1,
alpha
1
in
a
map,
and
I
wouldn't
get
a
parsing
error
because
they're
different
keys,
but
somehow
we
have
to
resolve
either
which
one
to
use
or
how
to
merge
them
or
like
what
do
we
do?
Do
we
throw
an
air
you
know?
So
that's
has
some
more
complex
edge
cases.
B
D
B
And
this
one
I
don't
really
get.
We
start
with
partial
control,
conflicts
and
the
owners
of
controllers
move
their
flags
into
conflict.
I
don't
really
see,
I
guess
it.
It
means
like
not
having
kind
of
this
map
we
have
here,
but
embedding
them
directly.
I
don't
know
this
one.
Is
it's
a
bit
cryptic
to
me.
To
be
honest,
I
wonder.
A
A
Is
the
the
schemas
for
these
sub
controllers
are
still
rather
complex
and
it
makes
it
difficult
to
generate
the
documentation
for
no
schemas,
because
they're,
just
unregistered
schemas
sort
of.
B
B
Like
this
whole
ink
cluster
config
thing
I
would
probably
defer
for
now,
because
I
think
it's
quite
I
don't
know
it
sounds
to
me
this.
This
alone
would
would
you
know,
require
a
cap
to
think
about.
A
It
it's
quite
complex
yeah
that
would
be
like
the
equivalent
of
dynamic
cubelet
config.
For
this
I
would
prefer
to
avoid
that,
because
that
was
a
lot
of
extra
work.
I'm
not
sure
it's
proven
useful.
A
B
So
yeah
I
mean,
like
my
my
kind
of
action
plan,
is,
I
start
a
cab
with
like
the
status
quo.
What
has
been
done
before
or
like
what
has
been
done
up
to
this
point
maybe
explain
a
little
how
the
actual
controller
manager
handles
you
know,
building
a
client
what
kind
of
options
we
need
for
that
then
like
what?
What
would
a
like
a
component
config,
look
like
with
the
current
state.
B
So
if
we
only
have
the
container
kind
version,
how
would
it
look
like
what
are
the
pros
and
cons,
and
then
I
guess
the
second
one
would
be
to
think
about
what
would
happen
if
we
version
each
controller?
How
could
this
look
like?
What
are
the
pros
and
cons
and
yeah?
Maybe
talk
about
some
implementation
options,
so
we
could
realize
it
what
what
has
to
be
done,
how
we
can
make
things
better
and
just
write
it
all
together.
Basically,.
A
Yeah
I
do
kind
of
like
the
idea
of
being
able
to
set
a
like
defaulting
object.
A
I
need
to
think
about
a
little
more,
but
I
it's
kind
of
nice.
If
you
have
you
know,
say
every
controller
basically
has
like
some
client
configuration
a
few
other
shared
fields
rather
than
try
to
like
extract
all
of
that
and
group
it
in
some
other
object.
It
might
be
nice
to
have
a
few
like
sort
of
it's
not
really
duct
typed,
but
I
feel,
like
defaulting
objects
where
it's
like
this
is
the
client
default.
So
this
is
the
other
defaults
and
the
controller
manager
takes
that
and
finds
in
all
the
sub
configs.
A
Or
you
know
we
pre-code
that
or
something
it
would
be.
It
would
still
enable
the
same
kind
of
deduplication
that
aggregating
those
fields
into
a
shared
object
would,
but
it
would
also
enable
the
ability
to
override
for
a
specific
controller
if
the
user
determines
that
that's
necessary,
and
if,
today,
those
options
are
actually
possible
to
override
separately,
then
that's
functionality.
We
should
maintain
so.
B
So,
but
that
that
would
require
possibly
embedding
another
type
into
the
existing
configs
right
now,
right.
A
Maybe
it
depends
on
how
they're
structured
like
if,
if
there's
a
shared
substructure
that
already
exists,
then
no
you
just.
You
could
just
reuse
that
if
they're
sort
of
spread
out
differently
in
different
configs,
then
yeah
it's
going
to
require
some
refactoring.
C
A
I
just
I
just
I
don't
think
we
should
be
focused
on
trying
to
enable
like
additional
functionality
beyond
like
providing
a
config
file.
So
if
it's
not
possible
to
specify
or
like
to
override
on
a
per
controller
basis
for
something
we
shouldn't
add
that
yet
because
that
just
makes
it
harder
for
us.
B
Okay,
yeah
sounds
good,
so
I'll
get
that
started
probably
have
some
questions
along
the
way
and
yeah
I'll
just
get
back
to
you,
while
you
guys
in
general,
awesome
super
excited
you're
working
on
this
alex
me
too
me
too,
all
right
cool,
that's
it
for
me,
I'm
gonna
get
some
dinner,
so
have
a
good
day
guys
you
too
take
care.
Thank
thanks.
Thanks
mike
for
the
help.
D
C
A
No
worries:
how
are
you,
how
can
I
help.
C
I
am
doing
good,
thank
you
and
I
have
a
few
questions
like
I,
so
I
am
going
to
create,
so
I
got
most
approval,
so
I'm
gonna
create
a
work
in
progress.
We
are
against
the
kubernetes
repo.
A
C
Finally,
yes,
I'm
excited
before
that.
I
want
to
make
the
comment.
Add
another
comment
with
all
the
validation
and
stuff.
So
I
was
looking
towards
adding
sorry.
I
don't
have
my
personal
laptop
too
and
I
have
everything
in
there.
I
should
have
bought
gotten
it.
I
was
in
a
hurry
when
I
left
home.
C
Talk
talk
you
through
it,
so
I
have
added
all
the
code
changes
that
I
could
think
of.
So
now
I'm
gonna
run
a
make
or
what?
What
do
I
run
next
to
create
the
generated
files.
A
Oh
make
generated
underscore
files.
C
A
Should
do
it
if
it,
for
whatever
reason,
doesn't
there's
a
folder
in
the
top
level
directory
called
hack?
That
has
a
bunch
of
shell
scripts
in
it
and
you,
depending
on
what
is
lacking,
you
might
find
a
script
in
there
that
can
generate
it
usually
generated
files.
Does
everything
it
does?
I
don't
think
it
does.
The
basil
file
updates,
though,
all
the
build
files,
like
the
all
caps
build
files,
so
kubernetes
has
two
build
systems.
A
There's
like
the
old
build
system
which
is
all
make
and
go
build
and
a
bunch
of
shell
scripts,
and
that
is
actually,
I
think,
still
used
for
official
releases,
but
there's
also
the
basil
based
build
system,
which
is
where
all
the
build
files
come
from,
and
that
is
what
it
gets
used
for.
A
lot
of
the
testing
in
ci.
C
I
understood
I
didn't
get
the
hassle
build
work
when
I
was
there.
Was
this
workshop
at
the
contributor
somewhere
designed
to
modify
not
keep
like
I'll
keep
confident.
I
guess
so.
I
couldn't
get
the
basil
build
work
then.
So
I
used
to
only
make
in
the
workshop,
but
I'm
gonna
go.
Give
it
another.
Try
to
try
basil.
A
Yeah,
so
there's
a
couple
things
to
keep
in
mind
with
basil
is
one.
The
most
common
problem
is
that
the
build
files
are
not
up
to
date
so
to
update
them.
You
can
run
a
script
called
hack,
slash,
update
dash
basil.
A
So
that's
how
you
update
the
build
files.
The
other
really
common
problem
is
that
you
are
using
a
bazel
version.
That's
too
new,
so
like
on
my
computer
at
work,
I
think
it
installs,
like
version
two
or
something
but
kubernetes,
still
uses
like
version
0.23,
or
something
like
that,
because
there
are
lots
of
backwards
and
compatible
changes
between
that
version
one.
A
C
A
You
will
ultimately
need
to
get
the
basel
build
working
for
the
pre-submit
test
to
pass
like
everything
will
have
to
be
correct
for
that
to
work,
but
yeah
you
can
come
back
to
that
later.
The
other
thing
to
know
is
that
sometimes
the
basil
and
make
build
systems
interfere
with
each
other.
So
it's
usually
a
good
idea
to
like
run.
Make
clean
before
you
basil,
build
or
run
basil
clean
dash,
dash
expunge
before
you
make.
A
And
sometimes
it
works
fine,
so
I
don't
usually
include
that
in
my
like
workflow,
but
before
every
command
I
run,
but
when
I
run
into
a
problem
I
just
run
I
clean
up
both
and
then
try
again
and
often
that
helps
fix
it.
C
Okay
and
after
this
I
had
another
question
that
what
are
all
the
validations
validation
files
that
I
would
have
to
touch.
A
I
think
it's
just
so
you
have
you,
have
okay,
so
you're
migrating
a
flag
to
config.
You
need
to
add
validation
for
it
to.
I
think
it's
just
validation.go.
It
should
be
by
the
external
type.
A
It
should
be
by
the
internal
type.
Sorry,
I
think
that
we
it
converts
before
it
validates,
but
then
you
should
search
around
to
like
references
to
the
variable
that
you're
moving
just
to
see
where
a
different
validation
happens,
especially
for
like
older
flags,
there's
often
validation.
That
is
just
like
super
super
deep
in
the
cubelet
application
code,
like
it'll,
be
like
right
before
the
value
is
actually
used.
C
Okay,
so
I'll
make
sure
that
I
search
wherever
the
differences
are
I'll
cross
45
needs
a
validation
or
not,
and
I'll
squash
this
once
I
get
an
overall.
If
I
could
get
it,
I
will
squash
and
then
make
a
pr
or
I'm
going
to
create
a
work
in
progress.
Pr,
so
that
you
and
lee
can
take
a
look
at
it.
It's
more
easier
to
take
a
look
at
there
and
then
taking
a
look
at
my
working
branch.
C
All
right,
I
will
ccu
if
you
don't
mind,
so
that
I
get
more
inputs
and
see
where
what
I'm
doing
wrong,
and
I
was
thinking
that,
if
I
get
this
one
successful,
then
the
next
time
we
could
just
move
a
bunch
of
files
like
bunch
of
flags
not
just
create
a
single
pr.
C
C
Yeah,
thank
you
mike.
I
I
think
I
have
most
of
things
and
I
understand
things
a
little
better.
I
will
ping
you
if
I
have
any
queries.
C
All
right,
then,
so
I'm
gonna
drop
up
the
call
now
and
I
need
to
go
authentic
dogs
to
give
them
an
update
on
the
release
docs.
So.
A
A
Okay,
thanks
folks,
I'm
gonna
call
the
meeting
here
take
care
and
have
a
great
day.