►
From YouTube: Config Working Group 3.15.18
Description
Meeting notes https://docs.google.com/document/d/1P3p7zOpX66hPoZBi_CiC36JW7JmoaLWqE2sgHvdq5tY/#heading=h.tko92dmkkj7q
B
I've
restarted
the
effort,
so
it's
eventually
first
thing:
I'm
doing
is
I'm
just
trying
to
come
up
with
a
decent
testing
framework
so
that
we
can
select
with
a
good
set
of
tests
on
the
get-go.
That's
why
I
started
writing
this
containerized
model
for
adding
API
server
into
the
testing
infrastructure.
A
B
B
A
A
Yeah,
that's
what
we
come
after
the
architecture
yeah
so
so
have
this
worked,
and
we
have
immediate
things
to
a
block
in
pallets.
It's
basically
a
lot
pallets
to
accept
the
different
kinds
same
kinds
of
different
API
group.
So
that's
would
be
some
storage
model
chains.
That's
were
coming,
I
hope
I
can
get
in
by
next
week.
A
A
F
B
E
B
B
C
B
H
A
Having
some
discussion
about
how
we
the
right
to
the
format
for
the
global
names
and
for
the
global
conflicts
and
the
namespace
come
fix,
we
have
I
think
the
last
meeting
we
have
talked
about
on
different
formats.
We
can
do,
we
can
either
using
the
different
namespace
and
like
a
mixer,
does
or
we
can
using
the
kubernetes
way
have
cluster
cluster.
A
We
have
basically
different
name
of
the
proto
message
to
represent
different
scope.
So
that's
two
options
we
can
have
or
the
other
one
is
in
Easter.
We
just
give
up
the
we're
just
giving
external
tool
to
author,
the
global
namespace
or
any
kind
of
different
namespace,
and
also
the
global
scope
or
any
kind
of
hierarchical
scoping
system.
We
can
work
with
and
it's
still
config
itself
and
will
be
a
namespace
scoped
as
the
top
scoping
and
and
we
can
having
service
scope
and
it
below.
Is
it
but
a
name
scope
is
the
top.
F
F
Cluster
level
are
back
and
have
that
like
types
or
endeavor
namespace
right
like
that's,
just
not
what
they
do
and
us
trying
to
do
that,
but
we
should
be
aligned
with
what
they're
doing
and
do
things
the
same
way
so
that
it
feels
like
a
consistent
experience
and
that
there's
not
two
different
ways
to
do
things
or
worse
that
we
used
a
I'm.
Sorry,
we
don't
even
do
that
yep.
You
know
right
some
other
thing
to
do
that.
So.
A
How
about
I
kind
of
write
it
up
and
we
can
having
a
involve
the
right
of
number
of
people
amount
of
parties,
and
we
can
talk
about
that
I
think
yeah.
We
should
open
that
discussion
grab,
yeah
I
would
so
any
other
questions
and
sorry
I
wouldn't
know
my
mic
was
the
speaker
was
muted,
so
can
I
hear
you
guys
I.
G
A
B
A
It's
good,
it's
due
schedule,
I
think
it's
a
good
time
for
us
to
revisit
what
what
we
have
done
in
March
and
we're
gonna
looking
forward
for
April.
So
that's
what
we
need
to
come
up
with
the
stage
for
each
of
these
features
in
the
configure
group
and
that's
what
we're
shooting
for.
So
we
probably
go
one
by
one
and
I
think
just
for
myself
and
if
and
for
you
guys,
we
should
also
review
the
the
criteria.
A
A
So,
let's
review
this
first,
so
so
alpha
and
beta
I
think
has
distinction
into
the
code
is
well
tested
at
the
features
safe
of
production
to
use
the
API
said
should
contain
beta,
that's
kind
of
one
distinct
details.
We
need
to
pay
attention
and
and
there's
I
think
there
was
another
one
is
the
end-to-end
test
should
be
covered?
A
A
C
I
would
consider
some
I
consider
beta.
You
should
be
able
to
install
sto
and,
if
possible,
we're
we're
doing
the
right
thing
with
validation
right
now.
If
you
have
the
latest
Corinne
is
clustering.
Slowly,
Co
we
don't
validate.
You
have
to
go
through
some
coffee
steps
to
get
things
set
up.
Installation
processes,
robust.
A
I
A
A
So
my
another
worry
about
the
bringing
into
beta
just
see
next
month
is:
we
are
working
on
Christopher
gonna,
starting
working
on
the
mixer
clients,
kinetic
mixer,
kubernetes,
strd
clients,
and
that's
will
actually
incur
quite
a
big
change
on
the
Cuba
kados,
because
now
cubicle
it
doesn't
use
any
of
the
mixer
CID
clients
and,
and
it's
kind
of
for
hot
:
assume
every
comes
back
in
the
same
API
version
from
mixer
convicts.
If
we
want
to
change
that,
that
will
be
non-trivial.
C
In
here
that
doesn't
making
a
changeable
are
updating
all
of
our
tests
to
use
the
different
groups
and
versions
as
well.
Assuming
a
su
control
is
passing
them
through
then,
it
should
be
covered,
probably
ought
to
a
bitter
end
and
has
to
use
this
do
control
and
in
more
cases
or
using
cue
pedal
for
a
lot
of
the
resource
prediction
now.
So
if
we
could
swap
in
is
typical
for
coverage
and
end.
A
Yeah
but
itself,
so
another
thing
is
I:
don't
expect
that
the
changes
in
easter
color
will
pita
and
maybe
miller
april,
because
there
will
be
non-trivial
announced
to
easy
of
color
to
using
mix
eliminated
clients
instead
of
having
a
dynamic
clients
kind
of
capture
them
all.
So,
I
think,
is
shooting
for
beta.
Is
it
a
bit
risky
and
are
you
thinking
we
should
shooting
for
that?
It's.
C
C
A
A
A
A
C
C
C
C
Was
it
ignored
any
infills?
There
was
a
way
to
parse
these
things
more
relaxed
matter.
Right,
yes,
yeah
I
disappear
for
that
and
if
it
got
it
held
up
because
there's
concerns
that
we
were
going
to
draw
by
warning
and
not
completely
Airy.
Now
we
would
let
we
create
undefined
behavior
the
system
and
that
we
ought
to
catch
that
earlier
invalidation.
So
with
what's
the
optimum
of
error,
state
state
or
you
just
drove
everything
and
for
it
for
pilot,
we
wouldn't
start
likely
or
we're
gonna
gather,
wouldn't
starter.
C
B
C
A
Concept,
so
it
this
is
really
not
like
makes
a
scope.
It's
it's
so
I
would
say
it's
more
like
a
runtime
settings,
a
lot
of
raised
flags-
and
you
know
in
there.
So
actually,
what
not
the
config
in
the
domestic
configured
actually
being
working
on
to
move
into
CID
is
the
authentication
policy,
so
dim
is
basically
now
or
the
MTL
flag
is
defined
in
the
Nash
setting
and
he's
moving
that
into
CR
D.
So
you
can't
have
incur
service
MPs
on
off
or
and
there's
something
additional.
A
B
A
B
A
There's
a
implications
that,
if
we're
moving
to
beta
or
stable,
we
need
to
have
some
backward
politics
guarantee,
and
so
that's
a
4mm
TOS
that
we
actually
really
want
to
remove
that
a
flag
out
of
the
configure
map
was
this
authentication
policy.
Cid
is
a
so
we
want
to
having
the
there's
a
one
in
one
place
to
define
the
FTOs
in
the
series,
and
that
is
the
ID.
We
talks
about
namespace
scope
or
global
scope,
so
we
don't
want
another
kind
of
flag
instance.
H
A
B
B
Okay,
regardless
I
think
there's
still
some
little
amount
of
work.
That's
left
here
right
so
right,
but
regardless
of
how
we
mark
this,
we
probably
need
to
make
sure
that
we
define
the
behavior.
You
know
the
policy
with
how
to
handle
with
these
like
missing
fee,
or
you
know,
mistyped
field,
this
particular
time,
and
that
may
be
at
a
couple
of
steps
right.
So
that
seems
to
be
should
be
done
or
goodness,
regardless
of
hobby
market.
B
Then
it
looks
like
we
can
either
mark
who's,
beta
and
stable
and
before
switching
to
CRTs,
and
that
affects
the
backward
compatibility
and
I.
Don't
think,
there's
a
forcing
function
for
us
to
mark
this
as
stable
at
this
point.
So
we
should
just
mark
this,
as
beta
you
know
do
that
upfront
work
mark
is
beta,
move
over
to
CID
work,
so.
A
If
we
market
this
as
beta,
given
the
fact
that
we
just
adding
some
tests
so
what's
the
CRT
is
that
it
requires
to
to
also
providing
a
operating
pass,
which
means,
if
people
are
they
using
this
flag
to
turn
on
MPLS.
So
we
need
to
have
a
tool
to
help
them
converting
this
into
CR
DS
and
if
we
I'm
not
sure,
if
that's
I,.
B
A
A
A
B
C
Do
we
have
a
warning
for
that
polygamists
will
miss
mismatched
versions,
no
we're
not
enforcing
any
sort
of
reversion
the
client
so
you're,
not.
We
have
version
endpoints
servers
and
information
in
the
various
components,
but
we're
not
accurate
or
not
aggregating.
All
that
information
and
saying
here's
your
entire
control.
My
version
there's
a
mismatch.
Yeah.
C
B
A
B
So
forgive
an
alpha-beta
to
me.
That's
paper
shuffling
right,
so
there
are
two
important
things
right:
define
the
behavior
so
that
everybody
is
aware
of
it
or
the
big
warning
sight
and
then
together
we
provide
tooling
for
migration
on
that
right.
So
those
are
two
real
like
ground
level,
things
that
name
that
that
has
an
impact
right.
So
if
you
don't
provide
a
migration
tool,
we
need
to
document
like
you
know,
we
need
to
have
some
sort
of
release,
notes
or
something
that
clearly
specifies
look.
This
is
changing.
B
C
C
C
You
can
validate
it
and
you
need
to
make
sure
that
your
evaluators
installed
before
your
to
think
that
your
your
C
or
D
is
right.
Now
we're
not
now
we
just
apply
everything
at
the
same
time,
I
hope,
but
eventually,
if
the
eventually
gets
created,
it
doesn't
work.
You
have
the
interdependencies
between
your
components.
A
Okay
right,
so,
let's
leave
for
alpha
and
I
think
I,
think
of
in
terms
of
communication,
I
think
having
it
to
alpha
stays
and
basically
it's
a
big
information
for
the
user.
So
you
just
if
we
put
anything
in
the
mesh
where
the
config
and
just
be
aware
that
it
can
break
anything
and
provide
much
guarantees,
but
I
think,
as
we
just
discussed
it's
fine
for
now
and
and
there's
no
driving
force
to
push
us
to
the
next
stage.
So
we're
gonna
keep
that
okay.
So
the
meeting
is
also
I.
A
Think
it's
the
30
minutes,
so
it's
official
running
out
of
time
so
for
the
rest
of
the
work,
because
we
are
kind
of
a
little
behind
on
some
of
the
galley
def
and
we
were
rowing
to
offline
for
we're
looking
for
for
the
next
stage
and
because
we
all
did
then
the
trackers
I
think
we
pretty
know
where
we
are
and
what
we
can
looking
for
for
next
week.
My
challenges
that
up
and
is
working
on
the
config
architecture
talk
we
have
been.