►
From YouTube: Kubernetes SIG Service Catalog 20170605
Description
The SIG meets to discuss beta criteria, pmorie/go-open-service-broker-client, and to see a demo of kubectl plugins for the catalog.
B
C
A
A
A
E
E
F
A
G
A
A
H
H
H
H
So,
in
order
to
use
these
plugins,
that's
two
things
you
have
to
set
so
how
the
protein
can
members
is.
The
first
thing
you
ever
do
is
to
export
environment
variable
so
that
the
coop
sealed
GPL
will
find
where
your
plugins
directory
is
so
right.
Now
we
don't
do
that
so
right
now,
I'm
going
to
pause.
H
Let
me
show
you
how
this
work
so
touching
that
the
mo
is
where
the
matter
is.
If
corn
oopsie-doopsie
here
is
going
to
run
into
here
and
look
for
protein
to
mo
and
then
in
here
is
just
basic
basic
detail:
could
figure
out
how
to
invoke
the
commander
you
want
so
in
here.
We
just
want
to
run
this
binary.
So
let
me
show
you
how
this
work.
So
now
we
have
set
the
environment
variable
and
you
know
to
invoke
our
commands
sub
CTL
and
currently
in
order
to
run
any
trapeze.
H
H
Should
be
right
in
a
second
real
fast
and
in
the
shooter
Oh
it
because
I
haven't
I
have
been
compiled
the
primary
sources
first,
one
so
the
creation
of
service
brokers,
that's
not
the
binary
file
created,
so
I
just
don't
have
to
go,
build
it
and
if
you
want
to
try
out
a
PR,
this
is
what
you
don't
have
to
do.
This
I
don't
build
the
binary
for
the
other
two
commands
and
not
this
one.
H
So,
in
our
case,
I
pretty
much
is
going
to
go
through
our
walk
through,
but
instead
of
using
the
call
coop,
CTL
I'm
I'm,
going
to
use
the
command
plug-in.
So
in
the
first
one,
broken
name
will
be
UTS
broker
and
then
the
your
URL
to
access
the
broker
will
be
this
so
run
that
and
it's
going
to
create
a
broker
and
then
just
going
to
show
you
the
broken
name
and
between
specify
namespace.
So
in
return
has
one
and
show
you
what's
created,
what's
being
created
so
the
next
step.
H
So
setting
two
men
will
be
clucking,
gray,
surface
instinct
and
the
usage
will
be
the
first
argument:
surface
craft
name
and
then
the
plain
name,
the
instance
name
in
the
namespace.
You
want
to
want
it
to
be
period
in
June,
okay,
so
a
copy
and
paste
command.
So
the
surface
instance
name,
the
premium
will
be
default
and
then
the
broken
instance
name
is
GDS
instance
and
the
namespace
is
going
to
be
in
test
namespace.
So
I'm
going
to
run
that,
and
you
should
create
a
service
instance
in
that
namespace.
H
H
So
first
to
look
at
the
namespace
is
okay
and
then
you're
going
to
create
it,
and
it's
going
to
return.
The
information
I
was
being
created
to
you.
Okay,
that's
the
second
command
and
the
third
last
command
will
be
to
bind
them
together
by
the
broker
and
the
service
instance
together,
so
plug
in
I.
Think
fine
service
and
the
usage
for
this
is,
in
the
instance
name
and
the
binding
name
and
the
name
space
in
this
to
be
a.
H
H
Ok,
so
this
is
the
final
surface
come
and
run
that
and
now
it's
successful
we
look
at
the
name,
space
and
you'll
create
it,
and
then
you
can
tell
whether
what's
the
name
of
the
binding
and
then
a
secret
name,
currently
a
secret
name
soon,
as
we
fought
to
avoid
for
the
binding
damage
and
there
is
I
think
that's
pretty
much
to
it
any
coming
a
question.
Well,
yeah.
F
F
H
You're,
correct,
like
the
plug-in
architecture,
currently
does
not
support
optional
facts,
so
we
what
we
can
do
is
we
can
do
it
as
an
option
argument,
but
I'm
not
doing
that.
Just
because
I
know
a
POS
being
worked
on
for
optional
facts,
so
I'm
just
going
to
wait
for
that
to
come,
and
when
that
drop
we
can
switch
over
to
use
pass
by.
H
E
So
one
of
the
things
that
I
think
is
interesting
is
and
good
demo
salmon.
So
in
this
particular
case
where
they
showed
it
is
binding,
the
secret
that
gets
created
from
the
binding
into
the
namespace,
which
is
which
is
sort
of
okay.
One
thing
I'd
like
us
to
discuss
at
some
point:
I,
don't
know
whether
this
is
necessarily
before
after
beta
but
I
do
want
to
discuss
the
option
of
changing
it
so
that
you
can
bind
to
not
just
a
namespace
but
to
a
quote
application
right.
E
So
you
could
say,
for
example,
play
for
the
word
main
space.
There
you
could
say
deployment,
slash
deployment,
name,
replica,
set,
slash
replica
set
name
that
way
we
can
actually
bind
to
a
particular
application
in
there
and
have
it
do
the
pod
preset
stuff
under
the
covers
automatically.
But
we
can
discuss
that
later,
but
yeah
I've.
A
Had
similar
thoughts
and
definitely
good
demo,
a
good
demo
Simon,
so
the
only
thing
that
I'll
call
out
is
that
I
think
it
would
be
great
if
we
could,
if
you
can
also
do
this
demo
for
cig
CLI,
and
we
could
get
some
feedback
about
like
naming
of
be
like
sub
commands
and
like
that
from
the
folks
who
who
spend
a
lot
of
time.
Thinking
about
those
things
as
pertains
to
CLI,
so
that
we
writing
in
a
good
territory
when
it
comes
to
those
kind
of
conventional
things.
A
A
F
H
E
Don't
like
getting
the
psychs
you
like
to
talk
about,
as
you
said,
Paul
the
the
format
of
the
command
look
on
the
command
line,
not
only
just
the
install
process
but
potentially
where
to
download
these
things
from,
because
it'd
be
really
nice.
If
there
were
pre-built
versions
of
these
binaries
already
available
for
people
to
easily
download
yeah.
I
E
E
E
F
I
H
Good
way
to
put
things
to
suggest
to
6000
si,
maybe
right,
since
we
are
having
thoughts
on
how
people
would
discover
this,
how
people
download
this,
maybe
they
can
support
a
plug-in
repository
where
people
can
discover
parking
commands
fly
from
the
coop
Cpl
right,
so
they
just
download
right
some
coops
ETL
without
leaving
the
contest
at
all.
Exactly
no
GFCIs
doing
that
and
there's.
E
E
Business
as
Simon
said
from
the
CS
command
line,
you
can
actually
get
the
list
of
plugins.
You
can
download
a
plug-in,
you
can
upgrade
a
plug-in
and
you
never
have
to
do.
Curl
commands
manually,
it's
just
download
the
right
version,
installs
the
right
directory.
It's
just
magic
under
the
covers
yeah.
H
For
this
is
also
super
good,
so
I
just
type
in
the
command
for
Cal
foundry.
This
is
a
list
of
all
the
plugins
like
this
Kyle
finds
with
Tamara
teeming
things,
so
there
will
be
possibly,
and
you
can
have
multiple-
we
probably
pop
repository,
so
one
for
collage,
one
from
the
community.
You
like
this
is
how
they
do
it.
You
can
just
download
some
here
cool.
A
So
Simon
will
you
have
a
take
away
to
get
this
demo
then
on
to
the
next
CLI
agenda,
what
you
can
show
it
and
once
once
you
do
that,
will
you
perhaps
also
as
a
takeaway,
send
an
email
to
the
the
google
group
and
put
something
into
chat
so
that
folks
know
so
and
we
can
go
and
watch
see
what
they
say.
Oh
great,
alright,.
B
A
J
J
Know
it's
hard
with
your
future.
Travel
detector
bubble
work
at
all
universe,
you're
in
anyway,
so
we
released
on
Wednesday
last
week
and
it
was
Bo
9
and
the
biggest
things
that
were
in
that
one
are
adding
a
support
in
alpha
for
the
service
brokers
schema
support
that
was
added
in
so
that
is
that's
I'm.
J
Also,
there
was
a
lot
of
refactoring
done
between
between
Oh
8
and
oh
nine,
for
the
controllers
of
the
controller
that
go
all
like
grown
quite
unbuilding,
so
the
first
attempt
to
kind
of
clean
that
up
was
done,
the
refactored
into
separate
files
and
lastly,
the
second
refactoring
was
done
to
break
the
unit
tests
and
the
corresponding
files.
So
that
was
that
makes
it
much
much
easier
and
hopefully
will
make
merging
easier
as
well.
J
And
then
there
was
a
lot
of
cleanup
done
to
make
the
protocol
yet
more
conformant,
mainly
issues
like
consistent
controlling
per
spekt
spective,
a
you
should
be
passing
the
query
parameter
in
some
cases,
giver
stopping
that
into
a
body
and
vice
versa,
so
that
that
work
was
done
for
this.
This
way
and.
J
J
One
thing
is:
is
poor
and
I
have
been
chatting
about
this
quite
a
bit
about
the
broker
client
that
really
belongs
to
the
so
this
catalog
or
if
it
repeats
on
weibo
and
Paul,
has
put
out
I.
Think
you
put
a
link
to
it
in
the
chat
poll,
but
yeah
I'm
gonna
put
a
link
to
it
in
there
yeah.
So
there
is
there's
another
bleep.
A
A
Kind
of
getting
my
getting
a
link
to
this
thing
into
the
document
tube.
So
on
the
note
that
deal
I
made
that
he
and
I
and
I've
talked
a
little
bit
about
this
with
Doug
and
and
Aaron,
and
we
we
talked
about
it
as
a
group
quite
a
while
ago,
but
I
think
that
it
would
be
amazing
if
there
was
like
an
official
open,
Service,
Worker
client
for
go
so
that
people
building
integrations
with
this
API
and
platforms
that
are
written.
Go,
do
not
have
to
implement
it
themselves.
A
There's
some
slight
degree
of
self-interest
there,
since
we
all
happen
to
be
working
on
exactly
that
with
kubernetes
and
obviously
kubernetes
is
written
in
go.
So
in
my
copious
spare
time.
You
know
in
between
a
jet-setting
around
and
all
of
the
photo
shoots
for
the
cover
of
Time
magazine
and
so
forth.
That
I
routinely
have
to
do
I
decided
as
a
project
of
mine,
to
build
a
go
line.
Client
for
open
service
broker,
and
part
of
that
was
as
a
exercise
to
get
me
to
to
to
force
myself
to
learn
inside
and
out.
A
The
spec,
which
I
have
found
a
couple
places
where
I
didn't,
have
the
right
knowledge,
which
was
kind
of
frustrating
to
me
as
a
really
pedantic
nerd,
not
to
know
the
right
things.
So
there
was
that
motivation
and
I
also
have
noticed
that
our
client
that
we
use
in
the
catalogue,
incubator
Rico,
is
not
very
well
tested.
So
my
two
goals
were
to
ensure
that
it
implements
the
spec
perfectly
and
and
also
is
really
comprehensively
tested.
As
far
as
what
request
body,
what
request
parameters
does
it
send?
Does
it
deal
correctly
with
malformed
responses?
A
Does
it
deal
correctly
with
certain
conventional
error,
codes
and
stuff
like
that
that
the
broker
can
provide
so
I'd
put
a
link
to
that
in
the
chat?
I
took
some
care
and
also
designing
this
to
be
a
drop-in
replacement
to
a
great
degree
for
the
one
that
we
have
in
the
repo
regarding,
like
the
creation
function
that
we
use
to
to
avoid
well
to
allow
ourselves
to
differentiate
the
actual
implementation
of
the
client
that
you
create.
A
So
I
will
make
an
issue,
and
probably
take
some
time
this
week
to
draft
a
pull
request
that
that
demonstrates
that
it's
not
supremely
intrusive
to
change
the
client
out.
But
what
I
think
would
be
great
is
if
we
could
donate
this
to
open,
serviceworker,
API
and
just
have
that
be
the
official
client
and,
as
part
of
the
process
for
open
serviceworker
API.
Make
sure
that
we
update
that
thing
and
likewise
I
think
the
Cloud
Foundry
stuff
is
implemented
mostly
in
Ruby.
E
A
A
So
I
actually
had
a
conversation
with
with
Shannon
trying
to
on
a
somewhat
related
note
of
trying
to
find
if
they
have
like
a
reference
implementation
of
a
broker
that
they
used
to
test
to
test
cloud
boundary
with,
and
it
turns
out
that
they
do,
and
he
at
least
seemed
receptive
to
the
idea
that
being
split
out
into
a
different
repoed.
So
we'll
see
if
we
get
some
traction
with
them.
If
you
don't
I'm
happy
to
put
it
anywhere,
it
makes
sense,
yeah
cool.
A
So
if
you
live,
that
URL
I
have
coveralls
working
on
that
repo
and
I
think
that
we're
at
like
90
percent
test
coverage
in
it,
which
is
about
what
you
can
get
to
with
that
testing
dumb
things
that
are
like
factory
methods
that
have
almost
no
logic
and
stuff
like
that.
So
I'd
be
really
happy
to
get
any
feedback
that
folks
have
I'm,
probably
going
to
make
a
poor
request
for
this
in
the
next
called
days.
That's
the
schema
support
in
into
this
repo,
but
it
already
supports
the
the
alpha
context.
A
Feature
and
I
have
flags
that
allow
you
to
differentiate
at
when
you
create
the
client,
what
version
of
the
API
it
should
send
and
whether
you
should
enable
alpha
features
or
not.
So
it's
written
with
some
of
our
concerns
in
mind
about
like
ex
broker,
doesn't
work
correctly
when
alphas
turn-on
or
ex
broker's
pinned
to
this
version.
I
so
did
a
little
bit
of
thinking
ahead.
Trying
to
design
for
that.
A
So
moving
on
the
next
thing
on
the
agenda
is
status
of
beta
now.
I.
Don't
think
that
we
have
it's
been
a
couple
weeks
since
I
think
we
talked
about
this
in
detail.
So
just
to
recap
for
everybody,
kubernetes
api
s
have
three
different
degrees
of
a
three
different
like
semantic
categories
that
they
break
down
into
regarding
level
of
support.
So
an
alpha
api
is
like
this
could
change
at
any
time
and
it
might
be
deleted
and
no
matter
what
happens.
A
But
for
all
intents
and
purposes
we
have
to
live
with
any
GAAP
I
that
we
release
for
the
foreseeable
future.
So,
as
far
as
that
pertains
to
us
I'm
sure
everybody
dialed
into
this
called
probably
knows
this
already,
but
we'll
repeat
it
just
for
anybody
that
doesn't,
and
anybody
watching
on
the
internet
that
our
API
surface
for
Service
Catalog
is
alpha
right
now
and
we
need
to
decide
what
our
decision
plane
is
from
when
we
think
the
API
is
ready
to
go
to
beta.
A
I
will
note
also
for
the
API
machinery
oriented
people
in
the
room
that
I
realized
today,
then
I,
don't
think
that
there
is
a
list
of
things
that
are
required
to
be
able
to
say
that
a
kubernetes
api
is
beta,
which
we
might
want
to
do
some
thought
about,
because
there
are.
You
know
there
are
a
number
of
different
good
practices
and
expectations
that
users
have
that.
How
kubernetes
api
is
behave.
For
example,
if
you
have
a
namespace
resource,
the
expectation
is
that
those
things
are
cleaned
up
when
the
namespace
is
deleted.
A
So
presumably
that
is
a
requirement
for
an
API
being
GA.
The
question
is
which
of
those
things
are
required
to
say
that
an
API
is
beta
and
I
know
that
we've
recently
took
in
the
communities
community
recently
have
taken
api's.
The
beta,
for
example,
stateful
set,
but
I
do
not
think
that,
as
part
of
the
exercise
of
doing
this
for
other
api's
that
we
actually
sat
down
and
fought
in
a
vacuum
which,
of
these
things
to
take
API
group,
X
or
Y
to
beta
are
really
generic
requirements
about
generally.
A
What
is
an
API
need
for
us
to
consider
it
data
from
a
operational
standpoint
to
good
practice,
the
endpoint
so
we're
going
to
going
to
have.
We
have
a
time
set
up
for
Wednesday.
If
you
talk
to
kind
of
hash
out
what
we
think
the
our
decision
plane
is
we're
going
to
bethe
I.
Do
not
think
that
we're
going
to
have
the
luxury
of
waiting
for
the
larger
community
to
go
to
beta.
A
The
discussion
about
there
should
be
something
in
the
kubernetes
community
Rico
that
like
clearly
articulates
the
requirements
for
things
being
beta
versus
GA,
so
I
will
I
will
put
the
coordinates
for
that
meeting
where
we
talk
amongst
books
in
the
sig
into
chat
and
so
forth.
So
we
can
all
listen
in
and
that's
essentially
what
I
had
to
say
about
that
I'm
we've
got
about
20
minutes
left,
but
we
could.
We
can
share
ideas
for
what
we,
what
we
think
a
good
decision
plane
is
and
perhaps
get
inputs
thinking
in
advance
of
when's
back.
E
Well,
one
of
the
things
I
think
you
took
an
excited
to
do
was
to
try
to
get
somebody
from
here.
What
like
it's
called,
but
somebody
who
has
knowledge
about
how
api's
are
supposed
to
be
designed
right.
What
our
resource
model
looks
right
or
not
in
these
tasks,
have
them
to
give
our
our
stuff
a
look
over
to
did
you
ever
get
a
chance
to
do
that?
I
have.
A
Not
had
a
chance
to
do
that,
I
am
crossing
my
fingers
in
hope
that
I
can
do
that
before
Wednesday
cool
it
actually
wasn't
a
cig.
It
was
like
I
was
going
to
try
to
get
time
with
one
of
the
owners
of
kubernetes
like
someone
in
the
top
level
owners
file
to
to
sit
down,
and
do
that
with
me.
But
I
haven't
had
time
yet
that.