►
From YouTube: WG Component Standard 20191015
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
Good
morning,
all
right
welcome
to
the
tuesday
october
15th
working
group
component
standard
meeting,
I'm
just
going
to
get
started
with
a
quick
announcement,
just
a
reminder
for
everyone.
Who's
joined
as
a
new
contributor
for
mentorship.
We
have
a
doodle
poll
out
at
this
link
for
selecting
a
time
to
set
up
an
office
hours.
So
please
put
your
available
times
in
there.
A
B
Yeah
morning,
folks,
so
like
yesterday,
I
reviewed
objects
pr
about
adding
strict
serializer
to
cubelet
and
as
part
of
the
work
which
I'm
preparing
for
cube
adm.
B
I'm
thinking
about
adding
customized
support
for
component
configs
and
right
now.
Customize
actually
requires
object,
meta
or
at
least
the
metadata
dot
name
field
to
be
present
inside
of
a
piece
of
yaml.
So
it
can
be
actually
customized
and
matching
it
up
between
the
layers.
B
B
Yeah,
it's
actually
a
thing
which
boom
we
in
cuba,
the
embalmed
a
long
time
ago.
So
we
have
a
strict
serializer
enabled
in
warning
only
mode
and
as
kind
is
actually
able
to
basically
customize
cube
adm
config,
it
basically
slams
a
metadata
name,
which
is
just
a
bogus
metadata
name,
and
we
always
warn
like
kubernetes
is
going
to
always
sworn
on.
Like.
B
But
it's
at
least
a
warning,
so
a
linear
path
should
do
and
probably
fixing
this
in
the
future
is
probably
going
to
be
the
way
to
go
forward
and
we
should
probably
think
about
studying,
object,
method,
or
at
least
some
stripped
version
of
it
to
component
graphics,
just
to
enable
customize
support
and
also
another
way
to
tackle.
This
is
to
basically
go
to
the
customize
fox
and
think
of
something
like
singleton
fields
or
something
like
this.
B
C
Would
actually
be
in
favor
of
like
having
a
selector
funk
and
that
you
could
at
least
supply
at
a
library
level
right,
so
that
in
something
like
kubernetes,
where
we
expect
there
to
only
be
one
kind
right
like
you
can
just
match
unkind
for
your
customized
patches.
B
B
Well,
basically,
a
jvk
that's
going
to
have
a
single
instance.
C
B
C
It
would
be
weird
to
have
multiple
cluster
configurations
for
a
single
kubernetes
run,
but
a
tool
that
was
wrapping
kubernetes
cluster
api.
I
guess,
might
have
multiple
in
the
same
place.
Yes,.
B
That's
exactly
the
case
so
yeah
and
for
component
configs,
it's
even
more
obvious.
So,
for
example,
you
can
have
several
different
sets
of
ways
of
deploying
a
certain
component,
for
example,
on
different
nodes
or
different
types
of
deployments,
and
you
may
actually
have
the
same
component
config
existing
in
several
different
config
maps
in
the
cluster,
with.
A
Probably
easier
to
just
include
object.
Meta
in
it
does
is
that
the
only
thing
that
makes
customized
work,
it
doesn't
have
any
other
conventions
that
you
can
rely
on
around
like
how
you
structure
the
the
directory
hierarchy
or
anything
like
that
to
get
names
like,
for
example,
right
customize
works
with
config
maps.
A
B
Yeah,
I
think
I
tried
in
the
past
to
basically
patch
up
a
yamal
config,
which
is
component
config
inside
of
a
data
field
in
the
config
map
in
a
yaml
file,
and
I
used,
I
think,
json
touches
and
it
didn't
work.
So
it's
not
parsing
the
data
fields
inside
a
config
map.
You
can
only
patch
the
the
configmap
type.
A
Got
it
all
right,
it's
probably
easiest,
just
to
add
object,
meta
to
component
configs.
We
should
be
clear
about
it.
Not
the
presence
of
object.
Meta
does
not
mean
it's
an
api.
You
can
post
directly
to
an
api
server.
B
Yeah,
which
is
like
currently
cluster
api,
uses
a
stripped
down
version
of
object
method,
so
they
have
a
local
copy,
and
this
is
basically
object
methods,
most
user
facing
fields
and
the
fields
that
are
basically
filled
in
by
the
api
server
enter
otherwise
read
only
to
end
users
or
probably,
should
be
opaque
to
end.
Users
are
not
included
in
there.
So
we
should
probably
talk
to
api
machinery
folks
to
try
and
invent
a
stripped-down
version
of
object
meta.
B
A
Yeah,
I
agree
with
that.
Well,
maybe
not
crds,
because
crds
would
just
have
the
traditional
object,
meta
right,
yeah
for
for
our
component
config
files
yeah,
we
can
use
our
a
stripped-down
version,
at
least
for
now,
unless
at
some
point
we
decide
that
we
want
the
full
one
and
as
long
as
we
stay
schema
compatible
with
the
full
one,
it
should
be
fine.
B
Yeah,
so
that's
one
of
my
things
and
the
other
thing
is
the
psa
about
cap,
which
I
prepared
for
a
standard,
combat
line
interface.
D
Before
you
go
on
just
a
quick
question,
because
the
prs
for
the
q,
proxy
and
cube
scheduler
have
been
merged
already,
so
should
we
just
basically
create
like
a
retrofit,
adding
the
linear
path
to
those
ones.
B
So
my
next
topic
is
the
standard
cli
for
component
config
violation.
So
this
is
a
cap
which
is
like
it
proposes
some
like
sub
command
based
solution,
like
I
think
it's
only
stefan
who
actually
reviewed
it
and
probably
team,
but
I'll
encourage
you
to
just
going
through
the
details.
I'm
not
like
tied
in
at
all
with
this
implement
like
this
proposal
for
an
interface.
So
we
can
actually
discuss
this.
B
We
have
basically
three
main
paths
to
go,
and
this
is
just
like
showing
up
one
of
them,
but
clearly
the
need
for
allowing
components
to
just
export
some
common
line
interface
to
verify
their
configs
and
do
nothing
more
is
basically
pretty
much
present
and
the
cluster
lifecycle
tools
are
going
to
and
probably
and
the
users
are
going
to
need
this
as
soon
as
component
configs
become
more
more
popular,
which
is
going
to
happen
when
more
of
them
graduate
to
beta.
A
D
Basically,
there's
just
that
one.
Well,
I
mean
we
had
to
talk
about.
You
know,
what's
the
like
standard
way
forward
for
those
fixed
serializers
and
we
kind
of
agreed
on
okay,
let's
do
linear
path
and
one
first
and
then
next
version.
Basically,
when
it's
dedicated,
we
drop
that
regardless
of
api
version
and
then
there's
also
the
one
for
cubelet.
D
So
I
wasn't
sure
what
happens
with
the
remote
config,
so
I
think
it's
using
like
when
we
take
a
remote
config
in
cubelet
and
we're
basically
running
it
for
the
same
codec
factory
right.
So
we
will
have
to
add
the
linear
path
there
as
well.
A
Yeah-
and
I
think
what
happens
with
that-
is
it's
decoded
in
memory
to
validate
it
before
committing
it
after
it
down.
Although
I
I
want
you
to
double
check
that
that
other
function
you
found
is
still
being
used,
okay,
because
just
downloading
it
to
disk
first
and
then
loading
it
through
the
other
path.
A
All
right,
great
everyone
have
a
good
day.
We
again,
we
are
working
on
the
office
hours
hitting
that
setup.
So
please
put
your
times
in
the
doodle
thanks
so
much
everyone.
Thank
you.
Bye-Bye.