►
From YouTube: CNCF App Delivery 2020-01-15
Description
CNCF App Delivery 2020-01-15
B
Everybody,
this
is
a
michael
from
from
the
harbor
team
ally.
We
had
the
quick
exchange.
I
ended
up,
not
adding
harbor
to
the
agenda
since
we're
gonna
go
through
seek
runtime
first
for
our
graduation
review
and
if
there
is
a
need
for
sick
up
delivery
to
look
at
harbor
as
well,
I'll
just
come
in
a
future
meeting
and
discuss
it.
That's.
A
A
A
D
D
A
D
F
G
B
D
G
A
D
G
G
D
H
A
G
Ahead,
I
was
just
gonna
suggest,
we've
got
all
of
these
and
I
suggest
two
things
one.
Can
we
get
it
with
the
CNCs
logo
in
it
instead
of
the
kubernetes
one,
because
I
don't
know
how
it's
gonna
change,
look
wise
a
little
bit
like
what
that
will
mean
and
then
can
we
do
some
kind
of
voting
counter
corridor.
Otherwise
it's
easy
to
set
up.
D
We've
run
this
process
with
the
other
groups
is
like.
We
have
now
just
done
our
first
round
of
design
meeting
and
we've
gotten
feedback.
I
will
then
go
back
to
our
designer.
Let
her
know
like
the
here's,
the
things
we
should
change
and
then
we'll
probably
break
them
into
groups
that
people
can
like
vote
more
directly
on
it
and
then
we'll
move
from
there.
Yeah.
D
A
A
D
D
G
D
A
A
D
A
A
A
A
J
I
G
This
this
does
raise
a
couple
of
questions
that
were
brought
up
by
others.
For
example,
the
operator
hub
is
tightly
coupled
to
the
operator
framework
and
OLM.
If
I
remember
right,
and
so
then
there's
other
operator
based
projects
that
are
going
through
this
SIG's,
such
as
Kudo,
and
that
doesn't
use
its
different
thing
from
the
operator
framework.
So
how
does
it
fit
into
the
context
of
operator
hub
if
it's
not
using
this
like?
How
does
that
all
mesh
together
in
a
nice,
cohesive
way
for
the
CNC
F?
That's.
J
J
J
G
If
they
all
come
in
under
one
project
and
one
project
governance
in
one
thing,
what
does
that
end
up
meaning
going
forward?
How
does
it
all
work,
because,
right
now,
yeah
Oh
LM
+
operator
hub
are
tightly
coupled?
So
what
does
this
mean
for
a
different
kind
of
operator,
even
if
it's
from
another
CNC
a
project
like
if
Kudo
comes
in
as
a
sandbox
project?
What
is
something
like
this
mean
and
how
does
that
go
forward?.
J
G
J
That
should
actually
be
fine
and
there's
a
couple
of
things
that
are
that
are
going
on.
That
will
actually
make
this
better,
so
operator,
SDK
is
looking
to
replace
their
Co
builder,
with
SKU
builder,
just
in
general
and
they're
gonna
keep
their
helm
and
they're
gonna
keep
their
ansible
there.
It's
bull
operator
builder.
It
still
leaves
lots
of
room
for
products
like
meta,
C
or
Kudo
to
come
in
and
build
operators.
J
E
Might
be
a
good
time
for
me
to
comment
in
on
from
the
kudo
side.
We
have
some
comments
in
the
operator,
favorite
draft
governance
and
yeah.
We
we
did.
We
did
some
talking
it
at
Q
Khan,
because
we're
also
working
on
introducing
like
a
another
interface
as
well,
for
operators
being
able
to
depend
on
each
other
called
COI.
But
what
we
added
to
the
pr
one
in
the
operator,
framework,
governance
or
draft
was
actually.
E
Getting
getting
some
more
projects
in
in
with
the
operator
framework
umbrella,
so
like
our
test
tool,
kudo
itself,
so
you
almost
have
like
that
medicine
c
approach
and
making
sure
that
all
these
projects
interoperate
well
together.
So
so
you
know
the
if
you
go
get
some
time
to
go.
Look
at
that
PR!
You
can
see
like
where
we're
almost
like
trying
to
like
pull
out
multiple
projects
under
a
large
umbrella
and
like
cover
operators
as
a
whole.
I.
C
A
Yeah,
those
are
from
my
pointing
it's
totally
fair
for
one
project
to
run
their
own
infrastructure
for
their
own
delivery
mechanisms,
also
for
operator
how
still
the
commanders
and,
as
I
mentioned
it
going
forward
that
we
obviously
want
to
have
tied
the
collaboration
between
projects.
Obviously,
but
just
because
something
is
specific
to
one
project
for
me
does
not
necessarily
mean
we
can't
always
kind
of
leave.
We
should
push
back
on
the
project
who
say
the
in
seeds
yeah,
actually
that
that's
what
then
part
of
the
collaboration
is
in
CNCs
should
be
about.
A
G
G
G
I
The
situation
with
that
is
that
for
operator
have
IO
to
work.
We
need
it
kind
of
a
common
definition
of
metadata,
that's
rather
than
for
the
catalog
display,
so
simple
things
like
a
logo,
a
short
name,
a
display
name,
a
description,
links
to
follow-up
stuff,
and
none
of
that
existed
at
its
time
when
we
conceived
operator
IO
and
it's
part
of
the
CSUSB-
expect
that
all
M
also
uses
four
lexicon
concerns,
and
we
already
started
looking
into
ways
to
separate
this
out
from
the
runtime
aspect
before
them.
I
G
That
also
kind
of
dovetails
with
the
application
specification,
that's
being
developed
under
kubernetes
and
even
the
helm
metadata
that
you
can
get
in
a
chart.
Yeah
memo
file
there's
a
lot
of
the
metadata
out
there,
but
my
point
comes
down
to
is,
if
we
look
at
this
as
a
whole
right
I'm,
not
trying
to
like
cast
blame
and
get
into
architecture,
but
as
the
sig
app
delivery,
I
think
we're
supposed
to
do
like
due
diligence
and
look
at
this
from
a
CNCs
perspective.
And
so
what
happens?
G
If
we
have
multiple
say
operator,
things
that
come
together
and
and
now
we
want
to
list
them
in
an
operator
hub
from
different
projects.
If
the
hub
is
tightly
coupled
to
one
project
with
one
way
of
doing
things
and
other
projects
want
to
come
in
or
dual
and
how
do
they
get
their
things
listed
if
they
don't
work
the
same
because
it's
tightly
coupled
to
one
groups
way
of
doing
things,
but.
H
G
J
G
J
Oh
em,
it's
not
a
packaging
format.
Olm
is
a
mechanism
for
getting
operators
installing
figuring
out.
What's
you
installing
the
cluster
I
mean
at
that
point?
It's
just
it's
just
kubernetes
objects,
but
all
em
actually
figures
out
where
it
comes
from
and
then
figures
out
what
namespace
it
needs
to
be
in
and
and
puts
it
there.
So.
G
That,
but
but
I'm
coming
back
to
implementation
here,
if
somebody
packages
a
chart
and
say
rust
order,
they
write
a
chart
and
rust,
alright
I'm,
sorry
an
operator
and
rust,
and
they
packages
it
as
a
helmet
are
to
be
installed.
Can
it
be
listed
in
the
operator
hub
and
work
with
everything
else
and
the
answer
I've
been
told
as
it
has
to
use
the
implementation
of
OLM?
So
unless
you
also
have
OLM
installed
in
your
cluster,
it
doesn't
work
as
you
need
this
like.
How
does
this
work
like?
G
Totally
understand
that
today
there
are
a
lot
of
operators
packaged
up
as
helm
charts,
and
there
are
people
who
prefer
to
do
that
and
if
operator
hub
is
the
destination
to
find
all
of
your
operators
and
now
we're
saying
that
a
set
of
operators
that
people
exist
in
use
and
work
with
today
cannot
be
listed
in
there
because
they're
not
using
a
certain
thing.
It's
not
really
a
full
list
of
all
the
operators
out
there.
It's
a
set
of
those
and
so.
G
For
you,
okay,
but
you
still
have
to
use
operator
SDK.
You
still
have
to
install
OLM
in
your
cluster
and
it
has
to
work
with
OLM
to
be
listed
on
the
operator
hub
and
so
today,
with
the
many
of
the
charts
that
are
out
there
that
do
this.
Unless
they
make
modifications,
even
though
they
do
operators,
they
can't
be
listed
in
operator
hub,
and
so
is
it
operator
hub
or
is
it
OLM
Saba
right
and
this
gets
into
if
we
look
at
it
systematically
across
the
CNC
F
not
just
like.
G
If
this
were
the
operator
framework
hub
or
something
okay,
it's
very
scoped
to
this
project.
It's
very
scoped
it
that
way,
kind
of
the
way
the
home
hub
is,
but
if
it's
listing
all
operators
everywhere
and
then
we're
artificially
limiting
some
of
that
listing
and
I
can
understand,
there's
lots
of
technical
constraints
in
the
way
it
came
to
be,
and
why
came
to
be
and
I'm
not
arguing
that
I'm
just
saying
as
a
CNC
F.
It
wasn't
me
who
noticed
this
right.
G
The
TOC
brought
this
up
in
the
first
place
and
I
can't
remember
who
it
was.
But
somebody
said
hey
what
about
this
stuff
and
they
actually
talked
about
all
the
hubs
right
because
right
now
there
is
the
there's,
a
security
hub,
there's
a
helm,
hobb
we
introduced
operator
hub,
there's,
actually
effort
working
on
to
see
how
do
you
consolidate
into
one
hub
and
I
know
a
couple
of
the
folks,
Dianne
and
folks
were
at
a
meeting
at
coop
con
on
this
they're.
Looking
at
how
do
you
consolidate
everything
into
a
single
hub
right
like?
G
Why
do
we
need
a
cloud
native
security
hub?
Why
do
we
need
a
helmet
if
customers
have
to
go
all
over
the
place
where
users
have
to
go
all
over
the
place
right
I
can
see
some
of
my
operators
and
operator
hub
I
can
see
some
of
them
that
home
hub.
What's
the
difference,
how
do
I
navigate
this?
What's
going
on?
G
It's
an
entirely
confusing
experience,
but
I
think
it
was
Joe
bata
who
actually
brought
that
up
and
one
of
the
TOC
meetings,
and
so
when
we
look
at
this
right
and
we're
still
fracturing
things
with
this
current
stuff,
what
what
does
that
mean
and
how
does
it
impact
end-users?
Not
us
project
owners,
but
the
people
who
come
to
consume
stuff
and
navigate
it.
I
I.
H
Thought
the
whole
point
is
not
to
be
kingmakers.
If
we
have
one
hub
for
everything,
then
I
think
we're
being
very
decisive
about
how
things
should
be
done
by
offering
an
alternate
way
to
do
it
in
that
alternate
way.
Welcomes,
however,
you
build.
It
will
provide
an
easy
way
that
we
can
get
it
in
here
as
well
so
lot,
so
we
can
show
it.
It
needs
to
have
a
minimal
amount
logo,
the
CSV
to
talk
about
the
versions
and
what
are
the
dependencies
things
like
that,
I
I,
guess:
I'm,
I'm
Matt?
G
So
two
clarifications
a
one:
the
operator
hub,
isn't
saying
anybody
can
build
an
operator
anyway
and
get
it
listed.
It's
saying
if
you
build
and
package
your
operator
a
certain
way
you
can
get
your
operator
listed,
so
it
is
automatically
limiting
it's
not
saying
anybody
can
come,
it
is
doing
king-making
and
then
the
second
thing
is
to
point
out
the
one
idea
of
one
place.
That
was
others
that
was
Joe
on
the
toc
and
it
was
a
point
on
the
TOC
Dan,
the
executive
director
of
CN
CF,
he's
been
talking.
G
C
The
central
hub
per
second
I
think,
like
that,
it's
useful
in
Scott.
It's
like
merits
that,
let's
just
table
that
for
a
second
so
flipping
your
argument
around,
say,
operator,
frame
workers
and
the
CNCs
and
helm
wasn't,
and
you
would
say
that
we
would
be
on
this
call
saying
you
can't
package
and
get
listed
on
helm
hub
because
you're
not
package
at
a
helmet
art
you're
using
this
CSV
thing.
Isn't
that
the
same
argument?
Wouldn't
we
will
not
want
that
outcome
as
well.
G
Okay,
so
the
helm
hubs
different,
because
it's
helm
packages,
if
you
create
a
helm
package
and
you
follow
the
structure
for
helm,
then
yeah
right.
It's
not
saying
it's
everything
that
can
be
installed.
It's
not
all
kubernetes
applications
right.
If
we
said
it's
kubernetes
applications
and
oh
yeah,
you
have
to
do
it
as
a
helm.
Chart
then
we're
kind
of
defining
the
way
to
install
kubernetes
applications
right.
G
Instead,
we're
saying
this
is
the
home
hub,
where
you
find
things
that
can
be
installed
with
hell,
instead
of
all
kubernetes
applications,
we're
not
saying
it's
all
the
things,
and
so
that's
where
I
would
argue.
It's
different
because
operator
hub
is
all
things
operators
right,
except
no,
it's
only
operators
that
can
be
installed
a
certain
way
and
it's
branded
as
all
things
operators,
and
so
that's
confusing
front
users
because
they'll
come
here.
Let's
say
it's
all
the
things:
operators
not
reeling,
that
many
types
of
operators
can't
be
listed
so.
J
I
mean
I,
don't
want
to
go
too
deep,
but
I
will
actually
say
that.
That's
that
is
not
that's,
not
true
it.
This
will
work
with
anything
that
we
call
an
operator.
What's
an
operator,
it's
a
set
of
CR,
zero,
more
and
a
controller
that
does
some
kind
of
function
in
the
cluster.
This
will
actually
work
with
any
of
those
and
that's
what
and
that's
and
I
think
it's
a
good
thing
and
I
did
have
a
conversation
with
Joe
and
I.
J
Don't
actually
think
he
understood
the
project
actually
laid
it
out
for
him
and
the
way
that
this
is
the
only
change
that
I
would
make
to
what
Red
Hat
is
saying
is
that,
instead
of
coming
in
product
first
I
would
actually
come
in
with
outcome
first,
so
the
operator,
the
operator
framework,
this
whole
thing
is
an
operator
SDK
get
you
started
with
building
operators,
and
then
there
is.
Then
there
is
an
operator
hub
showing
operators
that
are
that
are
available,
and
then
there
is
OLM.
J
You
need
a
way
to
manage
what
operators
are
available
to
what
namespaces
in
your
cluster
operator
and
OLM
is
agnostic
to
you.
It
actually
could
technically
be
operable
agnostic
to
operator
hub,
it's
definitely
agnostics
operator
SDK,
and
then
you
have
maybe
there's
another
piece
of
actually
managing
what
runs
in
a
namespace,
but
they
already
have
this
concept
of
subscriptions
and
the
other
thing
operators.
So
this
is,
it's
I.
E
What
just
to
found
some
type
of
a
brine
dish
for
a
moment,
and
that's
that's
also
one
of
the
really
valuable
things
about
you
know.
Kudo
employ
moving
into
this
like
another
spectra.
Interoperability
here,
so
that
we
can
continue
to
you
know,
try
to
get
more
and
more
inclusive
as
we
go
and
cuz
really
like
operating
frameworks
about
delivering
operators
right
and
within
there
there's
multiple
ways
of
delivering
and
serving
up
operators
to
people.
E
That's
still
focused
around
you
know,
Brian
said
as
far
as
like
it
being
that
core
concept
right
rather
than
just
some
delivery
package
right,
yeah,
helm,
chart
or
something
else,
seen
a
bundle
something
else
or
here
right
so
that
that's
that's
how
I
see
like
like
in?
Why
we're
this
way
as
well,
because
yeah
sure
it
might
you
know
you
can
get
the
impression
that
that
it's
that
way
for
now,
but
all
the
projects
that
we
have
under
that
umbrella
would
be
expected
to
interoperate
well
right
and
be
able
to
be
served
up
well
on.
E
E
A
B
A
Just
at
the
very
beginning
of
this
whole
document
and
the
whole
discussion,
so
I'd
agree
that
there
might
be
changes
that
we
want
to
do
work
on
certain
projects
to
be
more
widely
adopted,
or
that
the
answer
we
have
very
basically
what
have
to
list.
You
think
that
operator
framework
was
operator
hub
is
something
that's
significant
enough
within
the
cloud
native
app
delivery
community
that
it
makes
sense
to
be
powered
of
CN
CF
as
a
project
I
think
they
do.
Is
everything
perfect?
A
No,
might
you
have
specific
requirements
on
where
the
project
should
be
moving
in
a
certain
direction?
Yes,
and
actually
governance
could
help
us
here
I.
Think
overall,
it
is.
It
is
adding
value
here
and
discuss
about
nomenclature
me
or
what
we
call
the
operator
have
and
other
things
and
I
think
we
have
flexibility
there.
We
can
move
a
project
in
a
certain
direction.
Overall,
it
actually
does
add
value
and
also
to
to
the
other
point
you
don't
need.
Actually
all
them
to
be
installed
to
use,
operates
on
the
operator.
A
Have
we
it,
for
example,
dynaTrace
use
operator,
have
operators
and
not
use
oil
and
to
install
so
the
question
is:
do
we
consider
the
project
in
what
it's
doing
valuable
enough,
and
is
there
a
value
in
having
it
on
in
more
than
the
CN
CF
I?
Think
if
first
one
for
me?
Yes,
it
is
because
a
lot
of
people
of
Roatan
operators
is
there
value
of
consolidating
it
was
others.
A
Yes,
if
we
figure
out
that
there
is
a
restriction
that
other
projects
have
problems,
ways
to
submit
their
operators,
then
this
is
something
we
should
work
on
with
the
project.
That's
also
why
it
would
be
an
incubation
project
and
not
a
graduated
project,
obviously
so
that
that's
my
my
take
on
it.
I'm
not
saying
that
everything
is
going
to
be
is
perfect
right
now,
but
do
we
really
talk
about
something,
that's
significant
that
we
can't
allow
it
at
all.
J
And
that's
actually
a
really
good
point:
we're
talking
about
we're,
not
talking
about
a
graduated
project
here,
we're
talking
about
the
beginnings
and
and
hopefully
would
stewardship
from
the
CN
CF
that
we
can
actually
turn
this
thing
into
something
that
would
be
a
great
thing
for
the
whole
community.
So
I
mean
with
that.
That's
why
I'm
actually
saying
yeah
we
should
move
forward.
Are
we
where
we
want
to
be
right
now?
J
G
I
asked
a
specific
question
before
and
it's
an
easy
way
to
shut
me
up.
I
said
if
somebody
writes
a
an
operator
in
rust,
the
controllers
written
in
rust
and
it's
packaged
up
as
a
Helmut
art.
How
could
it
be
listed
on
the
operator
hub
and
I've
been
told
as
many
of
the
conversations
that
it
can't
be
today?
If
somebody
can
tell
me
how
to
do
this
right,
somebody
can
tell
me
how
to
do
this.
G
C
Two
ways:
one:
you
can
run
the
helm,
SDK
command
against
it
and
get
everything
you
need
all
built
up
and
packaged
or
if
you
think
in
that
home
chart
it's
really.
There
is
a
container
you
package,
a
container
with
breast
code
in
it
there's
a
deployment
and
our
back.
You
can
provide
that
in
the
OLM
spec
format
of
CSV
and
listed
oh
okay,.
C
G
But
lots
of
the
operators
that
I
deal
with
don't
have
anything
to
do
with
services
running
in
a
cluster
and
many
of
the
things
for
OLM
it.
You
know
they
don't
make
sense
for
a
lot
of
the
uses.
We
have
operators,
and
so
it's
it's
easiest
to
just
install
it
as
a
held
in
chart,
and
so
I
mean
really-
and
there
are
a
lot
of
people
who
do
this
I
need
a
way
to
package
it
up,
get
it
listed
and
run
it
with
or
without
OLM,
and
if
I
could
see
that
how
to
do
that.
G
C
A
G
I
might
suggest
when
this
goes
towards
the
TOC,
and
it's
in
the
plans
that
this
experience
is
probably
called
out,
because
otherwise
it'll
be
discussed
on
the
call,
if
not
by
me,
I,
wouldn't
be
surprised
if
others
do,
but
it
will
come
up,
and
so,
if
ahead
of
time,
you
say
our
plan
is
we're
able
to
handle
these
other
city.
This
is
what
we
do
today.
G
A
C
A
A
I
Will
be
the
case
in
the
future
right
now,
it's
basically
a
static
website,
as
it
reads
the
content
of
the
repository
once
into
a
no
chance,
backend
and
then
search
this
up
with
DUI
you
have
today.
We
do
have
a
database
format
in
place
that
we
are
looking
to
use.
So
there
are
no
inconsistencies
anymore
between
what
it's
in
github
what's
in
a
database
format
and
that
will
run
in
a
container
but
right
now,
it's
literally
once
the
site
starts
up
it
closed
the
repo
that
is
called
community
operators.
I
C
J
A
A
C
So
the
Olin
and
SDK
I
think
we're
the
ones
that
you
called
out
and
here
are
kind
of
like
we
talked
about,
can
be
used
together,
but
don't
have
to
be
used
together.
So
there's
no
reason
for
their
releases
to
be
synced
up
as
well
as
the
website
once
again
is
just
reading
that
community
operators,
you
know,
bundle
of
metadata.
C
F
C
C
Can
use
those
and
you
can
stop
there
if
you
wanted
to
or
you
want
some
the
lifecycle
benefits
of
the
lifecycle
manager,
the
things
that
Brian
was
talking
about
earlier
subscriptions
and
kind
of
more
of
a
multi-tenant
environment,
and
that
type
of
thing,
then
you
can
start
using
the
you
know,
CSV
format
of
the
lifecycle
manager
and
do
all
that
that
content
is
the
one
that
is
more
tied
to
operator
Cubs
listing
like
we
talked
about,
so
thank
you.
They
don't
have
to
be
all
connected.
Okay,.
F
I
see
this,
as
my
own
says
my
first
question.
The
second
question
is:
it
seems
that
the
Opera
SDK
om
in
Opera
hub
actually
have
different
maturity
levels
from
my
perspective,
because
they
are
actually
solving
very
different
problems.
I,
don't
know
if
it
meets
the
goal
for
all
of
the
incubation
criteria
and
I
think
we
may
raise
a
discussion
around
there.
C
F
So
from
my
perspective,
I
can
read
that
over
the
SDK
is
pretty
pretty
mature,
but
for
the
or
am
and
overdub
III
see.
Is
it
still
an
early
stage?
I
think
that
a
lot
of
operators
has
beliefs
in
the
Opera
hub,
but
for
or
am
it
seem,
there's
adoption
is
not
very
I
mean
it's
not
comparable
with
the
Opera
SDK.
Am
I
right
or
I
just
miss
something
I
think.
C
That's
one,
those
ones
because
it
gets
installed
on
folks
clusters,
it's
harder
to
gauge
than
the
SDK
which
you
can
go
see
how
many
people
have
cloned
it,
and
you
know
that
type
of
thing
I
think
that's
where
it's
a
little
bit
harder,
but
there's
definitely
thousands
of
openshift
clusters
that
are
installed.
That's.
J
C
L
G
And
I'd
also
like
to
point
out
well
there's
a
lot
of
open
shifters
out
there.
It's
important
that
the
CNC
have
to
not
play
kingmakers
here
and
we
have
got
a
lot
of
member
CNC
F
certified
distributions,
and
so
we
shouldn't
go
off
of
open
shift,
but
everything
should
run
off
of
what
is
it
compliant
kubernetes
clusters.
So.
E
E
K
J
J
K
J
A
Yeah
DB
only
I
think
a
bit
of
an
open
question,
for
me
still
is
when
I
look
at
the
repose
and,
of
course,
github
is
not
the
best
place
to
necessarily
raise
out
the
whole
creation
materials
like
heavily
having
some
smallest
contributions
from
at
least
two
more
organization,
which
is
very
vague.
I
would
like
to
see
a
supporting
statement
that
you
have
like
really
substantial
contributions
from
from
two
organizations.
I
think
most
people
right
now
are
from
redheads.
Well,
this
is
not
a
formal
requirement.
I
think
it
would.
A
J
A
A
A
Okay,
I
think
then
we
are
actually
on
top
of
the
are
today
well,
alas,
the
hour,
but
at
the
end
of
the
meeting
I,
don't
know
what
it
makes
sense
to
start
anything
else
on
the
agenda
today.
I
don't
think
so.
Unless
somebody
wants
to
bring
something
up,
that
can
be
quick
and
boost
the
other
topics
to
to
the
next
meeting.
Okay,.