►
From YouTube: GitOps Working Group Meeting 20210913
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
All
right
welcome
to
the
get
ops
working
group
co-chairs
meeting
on
september
13th
guys
here's
the
quick
link
for
convenience.
B
A
Because
we
we
still,
we
had
changed
this
sentence,
but
then
never
really
didn't
just
change
this
bit
and
somehow,
when
I
was
like
oh
we'll
just
use
the
same
word
like
syncing
automatically
that
sounded
a
lot
that
sounded
like
it
would
be
very
easily
confused
with
this.
A
C
I
like
the
idea
of
pull
because
they're
kind
of
getting
flies,
though,
because
we
have
to
talk
about.
Is
it
always
going
to
be
poor
versus
pulling
and
I
mean
pull
versus
push,
and
then
we
want
pull
right,
but
we
talked
about
that
being
very
implementation,
specific
that
not
necessarily
it
has
to
be
pulled
yeah
but
like
if
we
say
pulled
automatically.
C
A
I
think
the
I
mean,
I
think
the
principle
is
well.
I
think
it's
I
think
it's
really
just
about.
I
mean
honestly.
The
principles
is
mostly
about
the
fact
that
this
process
is
automatic,
that
the
process
of
the
the
process
of
getting
those
configurations
into
into
the
runtime
environment,
for
your
agents
is
automatic.
It's
always
happening
or
sorry,
not
not
that
it's
always
happening,
but
then
it's
sorry
that
it's
that
it
that
happens
on
its
own
without
you
having
to
trigger
it.
D
A
A
Is
it
really
necessary,
even
but
I
feel
like
I
don't
really
have
an
answer
to
that,
except
that
it
does
seem
that
it
was
really
important
for
us
and
for
the
group
and
in
the
principles
to
begin
with,
to
say
that
that
this
process
of
reconciliation
is
automatic
and
continuous,
you
know
and
the
the
we
had
previously
smashed
them
together
to
say:
well,
maybe
we
don't
need
it
broken
out
to
a
separate
step
to
say
that
the
configurations
need
to
be
automatically.
A
You
know
brought
into
the
run
time
for
the
agents.
Maybe
we
can
just
simply
say
that
the
agents
automatically
do
it,
and
then
that
caused
a
bunch
of
contention,
and
so
I
think
that's
why
I
just
we
recommended
going
back
to
to
keeping
them
separate.
I
don't
know.
A
Yeah,
I
think
in
this
case
it's
it's
or
synced
automatically
yeah
it
is,
it
is
ultimately
synced
automatically
or
ultimately
primarily
pull
automatically
but
yeah.
We
did
say
that
there
could
be
cases
where,
where,
where
that
message
might
be
sent
like,
for
example,
in
the
case
of
cluster
api,
you
know
I.
C
I,
like
sync
automatically,
I
I
don't
and
matter
of
fact
this.
I
do
not
necessarily
agree
that
sync
automatically
produces
confusion
against
continuously
reconciled,
because
one
is
about
automation.
The
other
one
is
about
continuous
reconciliation.
So
I
I
don't
necessarily
agree
that
stink
automatically
overlaps
with
continuously
reconcile.
A
Yeah,
do
you
think
we
should
use
it
then,
in
the
same,
do
you
think
we
should
put
it
in
the
same
order
as
continuously
reconciled
and
may
make
it
automatically
synced
or
reconciled
continuously.
C
A
A
E
E
And
polling
is
is
confusing
anyway,
because
because
you're
not
necessarily
always
pulling
it's
just
like
that's
how
it
how
it
works
best
most
of
the
time.
But
there
are
scenarios
where
you
might
have
like
what
does
it
mean
to
pull
versus
push
when
you're
managing
a
resource
in
aws
from
reading
aw
from
a
from
a
service
in
aws?
Is
that
pushing
or
is
that
pulling
like
they're,
both
on
aws
but
they're?
Not
it's
not
like.
It's
not
on
your
cluster,
deploying
to
your
cluster.
A
But,
to
be
clear,
I
think
like
this
is
this
is
the
other
thing
about
this.
Is
that
you
know
like
principle
three
as
we're
talking
about
we're
only
talking
about
the
declarations,
we're
not
talking
about
we're,
not
talking
about
the
resources
that
they
describe,
just
making
sure
that
those
declarations
are
are
fetched.
You
know
automatically
without
you
having
to
trigger
it.
Somehow.
A
C
A
Ultimately,
yeah
I
mean
we.
We
said
that
there
we
were
looking
at
some
exceptions
like,
for
example,
cluster
api,
but
I
actually
don't
think
that's
really
when
I,
when
I
think
about
it,
I
don't
think
it's
really
an
exception,
because
it
just
means
that
every
cluster
has
to
be
bootstrapped.
You
know.
C
C
But
that
is
not,
but
there
we're
not
talking
about
the
desired
state
becoming
ready
for
reconciliation,
we're
talking
about
the
actual
reconciliation
process,
which
I
think
it's
the
same
scenario
that
ban
points
out
right
when,
when
you
make
an
api
call
to
aws,
that
is
the
reconciliation
process
that
is
happening
not
not
the
the
pooling
of
the
state.
So
that's
right,
considering
that
I
would
actually
agree
that
pulled
automatically.
C
A
That's
all
I
meant
earlier,
that's
what
I
was
thinking.
I
mean
I
it's
not
my
hill
to
die
on
it.
Just
it
just
seemed
like
it
would
clarify
that.
That's
all.
E
A
Oh,
I
think
I
see
what
you're
talking
about
dan
you're
talking
about
like,
for
example,
if
it's
syncing
like
directly
into
kubernetes
api,
then
what
is
the
yeah?
I
see
what
you're
saying
I
guess
that
could
be
yeah.
That
makes
sense.
I
definitely
I
think
the
it
would
yeah
you're
right.
I
mean,
I
guess
the
agents
need
to
store
it
somewhere.
You
know
they
need
to
do
something
with
it,
but
maybe
not
you
know,
maybe
some
systems
just
hold
it
in
memory
and
just
do
it
constantly.
I
don't
know.
E
A
D
E
A
No,
that's
true.
That's
a
very
good
point
yeah
I
that
makes
sense
yeah
because
because,
ultimately,
I
think
the
reason
it
was
written
the
other
way
previously
was
because
the
fourth
principle
was
it
actually.
The
header
for
the
first
principle
were
just
agents-
software
agents,
you
know,
so
I
think
it
was
really
trying
to
emphasize
the
software
agents,
but
I
think
you're
right.
It
isn't.
Both
of
these
steps
in
fact
happen
with
agents.
C
A
A
C
A
A
So
right,
like
our
goal
here,
sorry,
I'm
just
gonna.
A
I
just
want
to
look
real
quick
at
the
at
the
milestones
that
we
have
or
the
yeah.
So
we
have
issues
milestones
for
rc1
our
milestone
goals.
Were
we
wanted
to
simplify
the
principal
titles
we
wanted
to
resolve
the
notes
and
glossary
items
we
haven't.
A
A
Thanks
clarify
language
to
emphasize
main
point
of
each
principle,
maybe
just
like
ensure
consistency
of
language
right.
That
was
one
of
the
things
we
said
but
didn't
write
in
this.
A
Oh
yeah,
we
had
kind
of
a
version
of
this
earlier,
I
believe
like
comparing
against
and
I
what
I
remember
is
that
it
seemed
like
we
all
agreed
that
that
was
implied
by
by
reconcile
with
the
desired
state
as
the
source
of
truth.
A
E
E
I
was
just
wondering
if
I
could
get
the
actual
system
state
kind
of
closer,
and
it
kind
of
feels
like
attacking
source
of
truth
on
on
at
the
end,
is
like
it
kind
of
does
I
mean
like
yeah,
it's
like
tacked
on.
It
feels
like
some
other.
E
But
when
I
do
it,
this
second
way,
which
I
don't
love
totally
either
for
reasons
I
can't
articulate
yet
I
feel,
like
I
understand
that
I'm
doing
something
to
my
actual
state,
I'm
operating
my
actual
state
against
the
like.
I
don't.
I
don't
feel
like
there's
any
ambiguity
in
that.
That
would
make
me
think
of
what
the
source
of
truth
is.
I
know
what
it
is.
E
I
don't
really
like
the
phrase
bring
it
in
line,
sounds
weird,
but
but
as
it's
it's
telling
me
I'm
looking
at
actual
and
desired
and
I'm
updating
the
actual.
A
E
Don't
think
you
can
make
comments
in
markdown.
A
A
Okay,
the
desired
state
of
get
ups
managed
system
must
be
source
of
truth
right.
I
mean
that's
kind
of
what
we
were.
That's
what
we
were
saying
before,
if
you
like,
if
we,
if
we
want
to
make
it
extraordinarily
clear
that
that
is
always
the
case,.
E
And
what
I
guess,
the
point
of
it
would
be
desire
to
get
off
spanish
system
must
be
the
source
of
truth.
A
Yeah,
I
mean
maybe
just
like-
maybe
something
like
that.
I
mean
I
feel
like
this-
could
use
some
more
feedback
and
iterations,
but
ultimately
that's
something
that
we
want
to
say
either
before
we
said
that
we,
the
the
correct
place,
would
be
put
to
put
it
in
in
the
continuously
reconciled
section,
because
that's
how
it's
reconciled,
but
I
think
you're
right
it's.
I
think
it
would
be
equally
good
or
maybe
it'd
even
mark
it
bring
it
home
more
to
have
it
separate.
C
A
A
Suffrage
continuously
observe
actual
state
and
attempt
to
apply
the
desired
state.
Yeah,
simple
short,
it
is
I'm
just
trying
to
it.
Is
I
actually
kind
of
like
that
better?
In
fact,
that's
kind
of
what
what
it's
closer
to?
What
was
there
originally,
but
the
only
difference
is
one
of
the
things
that
was
super
important
for
for
folks
in
the
group
was
actually
no.
This
sounds
good
attempt
to
apply
it's
the
exact
same
thing
as
initiate
actions
to
reconcile
attempt
to
apply
could
be
initiate
actions
to
apply.
E
A
A
That
one
just
say,
observe
actual
state.
A
A
E
A
All
right
so
check
out.
Okay.
First
of
all,
I
agree
with
you.
Second
of
all,
second
of
all,
it
really
is
kind
of
proving
how
tight
some,
maybe
not
exactly
the
wording,
but
some
of
the
original
principles
were
because
just
check
it
out
all
right.
What
it
actually
said
was
was
exactly
what
you're
saying
it
maintains
a
system
state.
It's
just
a
different
way
of
saying
it
right.
It
applies
the
resources
described
in
the
clear
wait,
a
second,
oh!
A
A
Okay,
I
guess
this
version
does
not
say
alert
on
divergence.
There
was
an
earlier
version
that
did
all
right,
never
mind
what
I
was
just
saying.
Remember
like
I
just
I
can
only
point
to
this,
not
because
I'm
working
here,
but
because
this
is
where
it
first
started.
E
A
E
Much
oh
yeah
and
sure
correctness
could
mean
a
lot
of
things
yeah.
I
think
sure
correctness
could
mean
that
you've
made
a
declaration
that
uses
too
many
resources
and
that's
bad
for
the
environment
and
that's
not
correct.
A
A
A
Then
you
guys
feel
like
34
gets
it.
I
think
so,
but
my
question
is:
should
we,
if
we're
going
to
instead
of
saying
like
initiate
actions
which
is
like
a
little
bit
more
open?
If
we're,
if
we
want
to
say
attempt
to
apply,
which
I
think
is
actually
true,
do
we
also
want
to
just
say
and
alert
on
and
alert
on
divergence?
You
know
like
isn't
that
part
of
the
the
idea
is
that
we're
we're
letting
humans
know
when
something
went
wrong?
Well,
this
gets
to
point
five.
E
A
A
Yeah
actually
you're
right,
you're
right
now,
it's
now
at
least
with
this
proposal,
it's
0.6,
but
because
we
we
suggested
putting
a
source
of
separating
source
of
truth
from
this
and
making
it
sound
principal,
which
I
also,
which
I
agree
with
by
the
way,
and
I
think,
if
that's
the
case.
E
I
don't
think
we
need
a
separate
principle
with
the
rewritten
version
of
continuously
reconciled,
because
it
very
clearly
states
that
it
is
the
desire.
It
is
the
source
of
truth,
we're
applying
the
desired
state.
That's
what
we're
doing
so,
then
I
feel
like
we
don't
need
it.
A
Yeah
I
mean,
I
think,
the
reason
that
we
we
wanted
to
emphasize
that
before,
or
that
was
brought
up,
is
because
the
principles
didn't
really
quite
say
that,
but
that's
like
the
the
the
driving
force
of
get
ops
that
I
think
has
emerged
continued
to
to
to
prove
itself
over
time.
You
know
that
that
we're
not
just
talking
about
some
multi-directional
convergence.
Generally,
you
know
that
was
even
a
confusion
between
there
was
even
a
confusion
with
you
know,
at
least
one
person
in
the
in
the
get
ops
working
group.
A
That
to
reconcile
the
actual
state
with
the
desired
state.
Yeah
I
mean.
Maybe
that's
enough.
I
don't
know
I
mean
I
think
as
long
as
we
like
make
the
point
very,
very,
very
clear.
Somehow
I
don't
even
I
almost
don't
even
care
how
at
this
point
as
long
as
it's
as
long
as
it's
just
super
clear
that
that
the
desired
state
always
wins,
you
know
it's
not
something
better.
Yeah.
E
States
the
direction
of
the
reconciliation,
the
weird
thing
about
source
of
truth
was,
as
a
phrase
it's
a
very
strong
phrase,
and
we
actually
even
have
it
bolded
in
there.
So
it
was
like
this
is
such
a
strong
phrase
and
it's
bolded
in
here
in
a
way
that
makes
it
feel
like
it
should
be
its
own
thing.
Yeah.
D
E
Now,
if
we,
if
we
take
this
this
line
34
instead,
it
just
explains
how
reconciliation
occurs
and
the
explanation
says
where
the
source's
truth
is
from,
but
it
doesn't
say
it
in
such
bold
language
that
it
feels
like
a
separate
thing.
I
I
don't
think
it's
an
I.
I
don't
think
it's
an
issue
of
the
ideas
being
too
separate.
I
think
it's
an
it's
an
issue
of
the
wording
making
them
feel
too
separate,
and
I
don't
think
that's
there
anymore.
A
A
I
definitely
don't
want
to
like
bike
shed
this
part,
but
I
think
I
think
you
have
a
point
at
least
that
when
we
were
describing
actions
to
reconcile
that
does
seem
to
leave
some
ambiguity
if
it's
not
spelled
out
perfectly,
but
if
you're
talking
about
attempting
to
apply
the
desired
state
to
the
act,
you
know,
then
that's
pretty
damn
clear.
A
A
What
we're
seeing
here
just
we're
putting
the
word
attempt
in
there,
which
was
which
resolves.
I
I
think
anyway,
some
of
the
cons,
the
concerns
that
some
other
members
of
the
working
group
had
initially
that
it
may
not
succeed
right.
A
So
I
think
that
that
result
that
resolves
that
that
we're
just
saying
it's
attempting
to
instead
of
just
applying
them,
we're
kind
of
also
taking
it
out
of
the
language
of
kubernetes
kubernetes
does
apply,
but
but
I
think
that
the
implication
of,
if
you're,
if
you're,
applying
something
in
kubernetes
we're
talking
about
some
something
like
control
apply
right.
That
doesn't
necessarily
mean
it
succeeds,
but
I
think
we
need
this
to
be
independent
of
talking
about
kubernetes.
So
I
think
yeah.
D
A
A
You
know
yeah
yeah
or
something
like
that.
You
know,
because
you
know
okay,
this
is
all
recorded,
so
it's
not.
You
know,
but
one
interesting
option
might
be
that
what
if
this
were
what
if
this
were
just
a
weird
idea,
just
in
terms
of
order,
what
if
this
were
put
before
reconciliation?
A
A
It
almost
sounds
like
what
we
were
trying
to
get
possibly
part
of
what
we
were
trying
to
get
at
with.
You
know
the
the
re.
We
really
mean
it
principle
at
the
end
that
we
had
to
ditch.
C
Does
it
yeah
to
me
because
I'm
trying
to
kind
of
think
about
all
the
scenarios
where
principle
the
principle
of
continuous
reconciliation
doesn't
already
imply,
given
its
current
warning
that
the
desired
state
always
wins.
So
I'm
this
agent
is
always
observing
state.
That's
gonna
find
this
state.
If
there's
any
other
input
of
configuration,
whether
human,
whether
any
other
file
that
was
applied,
anything
else
is
going
to
modify
the
actual
system
state
that
the
agent
is
observing
and
the
agent
is
going
to
apply
the
desired
state
which
is
declared
and
therefore
single
source
of
truth.
C
But
but
the
question
is,
are
we
trying
to
communicate
something
more
with
the
source
of
truth?
Are
we
trying
to
communicate
something
in
terms
of
disability
into
a
system
from
a
single
data
repository
anything
that
goes
beyond
just
precedence
of
configuration
of
a
binding
feature?
Is
there
anything
else
that
would
be
incorporated.
A
A
E
C
A
You
know
noted
during
the
reconciliation
reconciliation
piece
similar
to
lead
to
how
it
was
like
way
back
in
this
version
you
know
or
whether
it
should
be
part
of
the
closed
loop,
because
that's
what
happens
when
that's
part
of
a
feedback
system
you
know
and
that
feedback
could
mean
multiple
things.
A
I
mean
the
main
thing
it
means
primarily
is
notifications,
but
there
was
another
case.
You
know
where
you
know,
for
example
like
stop
after
x
number
of
attempts
and
notify,
or
something
like
that,
I
mean
I
guess
it's
all.
It's
all
around
notification
so
far.
A
But
really,
like
that's
part
of
in
my
understanding,
that's
that's
like
what
the
closed
loop
idea
was
about
like
in
just
just
to
go
to
back
to
the
prior
art
thing.
This
literally,
this
was
like
important
enough
to
be
a
principal.
A
For
the
group,
as
the
group
was
getting
started,
this
draft
this
version
of
the
get
ops
principles,
it's
pretty
much
the
same
thing
it's
just.
I
think
this
includes
like
alerting
it
just
doesn't
say
so,
just
as
actions
are
performed.
A
A
E
I
I
almost
want
to
rephrase
it
to
just
say
any
other
interaction
with
the
actual
system.
State
should
be
minimized.
E
You
know
like
because
that's
kind
of
what
we're
really
saying
is
like
nothing
about
this
says
exclusive,
but
maybe,
if
you
just
said
continuously
reconciled.
A
Like
if,
if
I
understand
correctly
like
that's,
not
really,
I
mean
the
point
of
a
closed
loop.
Well,
not
not!
Really!
No,
it's
not
like
air
gap.
You
know,
whereas
the
source
of
truth
thing
down
below.
That's,
that's
almost
what
you're
saying
right
now.
A
You
know
that
there's
there's
only
one
way
to
do
this,
yeah
good
point.
Ultimately,
you
know,
whereas
I
mean
it's
funny,
because
these
these
bleed
together
a
bit
right,
but
like
that's
just
what
I
was
thinking
is
that,
like
the
the
point
of
the
closed
loop,
is
that
it
it's
a
it's
not
an
open
loop.
It's
not
just
a
dumb
application.
It's
it's
an
it's!
It's
one
that
takes
into
account
feedback
from
the.
F
Feedback
from
the
yeah,
but
we
we
just
couldn't
figure
out
what
that
feedback
would
be
right.
Yeah,
here's
here's.
C
A
weird
example
that
I
wonder
what
your
thing
means
is:
there's
a
close
look,
and
I
kind
of
hinted
at
something
similar
to
this,
but
this
is
like
a
real
real
world
thing
with
flux.
If
I
have
two
customizations
k2
and
k1
and
k2
depends
on
k1,
okay,
I
apply
k1,
it's
good.
I
apply
k2
that
depends
on
k1.
C
It's
all
good.
I
remove
k1,
meaning
the
dependency
that
k2
depends
on
is
gone.
How
do
you
remove
it
helm
delete?
No,
actually,
I
I
removed
it
from
the
start.
String.
Okay,
it's
done!
Okay.
So
at
that
point,
k2
is
gonna
enter
into
a
state
that
says
dependency
missing
and
any
changes
that
I
introduced
to
the
repo
that
is
the
source
of
k2
are
not
going
to
be
applied.
C
Wait
any
changes
won't
be
applied.
That's
the
behavior
of
flux
as
it
as
as
it
stands.
This
is
like
real
world
thing.
Okay,
I
have
five
two
I
remove.
K1
k2
is
pointing
to
some
repo
some
path
and
I'm
introducing
more
changes
to
that
repo.
I
hear
what
you're
saying
you're
saying
it
can't
be.
It
just
can't
be.
C
E
A
A
F
Know
anything
to
do
with
yeah,
because.
E
C
A
That
does
something
kind
of
different,
and
you
know
you
as
long
as
the
software
agent
follows
that
dan
just
wrote
follows
the
principles
he
can
make
it
do
whatever
he
wants.
A
A
It
yeah
ultimately
right
exactly
so.
You
know
there
may
be
so,
for
example,
in
a
reconciler
that
I
don't
know,
uses
let's
say
like
service
catalog,
a
service,
catalog
concept.
You
know
it
may
attempt
to.
I
mean
a
particular
control
or
a
software
agent
right
may
attempt
to
say:
hey,
look
something
broke
in
the
network.
A
I'm
gonna
go
back
to
the
api
for
this
cloud
provider
and
say
I
need
that
network
connection.
You
know
it
may
do
those
kinds
of
things.
That's
not
unlikely
at
all.
C
A
That's
an
interesting
observation,
but
I
don't
know
that
we
can
that's
a
good
question.
I
mean
I
don't.
I
don't
know
that
we
that's
that
hasn't
come
up
so
far.
This.
E
E
A
That
that's
a
that's
a
really
good
example.
That's
that's
a
very
good
example,
and
that's
what
flagger
does,
for
example,
and
that's
what
other
tools
that
that
that
work
with
get
ops
that
leverage
deployment
strategies
I
mean
there
aren't
very
many
that
are
you
know,
but,
but
so
I
think
that's
hard
says
that
right,
okay,
yeah,
yeah,
yeah
there
we
go
exactly
right
right.
So
so
that's
a
good
example
of
of
why
we
want
to
keep
that
closed-loop
principle.
You
know,
because
I'm.
E
Is
that
because,
well
operating
in
a
closed
loop
could
mean
if
we
said
something
about
meta
state?
Maybe
if
we
want
to
make
a
point
about
metastate,
that's
something
we
could
do,
but
I
kind
of
think
that's
also
in
a
way
covered
by
how
we've
described
the
state
all
the
way
up
up
and
down
here,
because
if
you
don't
have
this
in
here
in
the
glossary,
you
can
explain
what
desired
state
means
and
desired.
State
can
include
meta
state.
E
You
can
conclude
like
attempt
x
and,
if
not
fall
back
to
y
apply
security
policy
apply
whatever
yeah.
That's
true.
As
long
as
they're
done
in
a
declarative
way
and
they're
not
done
in
an
imperative
way,
then
it's
fine
right.
C
Built
that
just
here
you're
having
your
desired
state,
a
policy-
okay,
you
add
to
your
desired
state,
something
that
breaks
that
policy.
All
right,
you
have
don't
allow
certain
types
of
objectives
in
the
name
space
or
something
right.
So
you
have
a
policy
that
denies
that,
and
you
add
something
to
your
desired
state
that
actually
breaks
that
policy
and
you
push
that
to
your
desired
state.
A
C
A
Loop
means
you
do
whatever:
well,
not
just
whatever
it
it's
it's
it's
it.
This
is
just
defining
a
set
of
practices
that
have
grown
over
the
last.
I
don't
know,
however
long
20
years
or
whatever
you
know.
So
it's
like
what
we
should
do.
We
know
generally
what
we
should
do.
We
should
either
stop
the
works
and
send
out
notifications.
So
human
can
resolve
it.
We
should
automatically
roll
back
to
the
last
working
version.
E
A
Desired
state
yeah
I
mean,
if
you
remember,
there
was
a.
There,
was
a
discussion
that
I
brought
up
like
kind
of
a
long
time
ago
about
saying
that
the
wording
for
our
old
principle,
five
should
say
something
like
that
that
the
rules
for
what
the
rules
for
what
controllers
or
sorry
software
agents
do
should
be
should
be
contained
in
the
desired
state.
A
No
right,
just
initial
really
just
just
initiate
different
actions,
and
we
and-
and
we
can't
really
describe
what
all
those
actions
are,
because
we
only
can
envision
the
ones
that
we
know
about,
but
there
there
may
be
other
actions
that
are
initiated.
I
mean
who
knows
you
know
next
year,
maybe
a
big
craze.
You
know
there's
there
might
be
another
new
deployment
strategy
that
no
one's
ever
thought
about,
but
but
yeah
honestly,
the
so
far
these
closed
loop
things
say
it
they.
A
I
think
they
may
not
clarify
it
in
the
way
that
we're
describing,
but
so,
for
example,
the
first
one
different
actions
are
initiated,
depending
on
inputs
in
the
desired
and
actual
states.
What
that's
really
the
reason
that
I
initially
thought
that
that
would
be
a
good
way
to
put
it
and
it
it
doesn't.
A
E
We
are
at
time
by
the
way
I
have
10
more
minutes
and
then
I'll
be
getting
a
phone
call.
So
as
soon
as
I
get
that
phone
call,
I
got
to
jump.
C
I
think
at
this
point
it
seems
that
there
is
really
still
a
lot
of
ambiguity
as
to
what
the
close
loop
thing
is,
and
I
would
even
argue
that,
if
it's
so
complex
to
make
sense
of
it's
likely,
not
a
principle,
and
I
would
I
would
say,
is
there
anything
else
that
you
want
to
cut
like
other
than
principle
fire,
anything
else
that
we
need
to
look
over
principles,
one
through
four
or
are
we
happy
with
those
admittedly
saying
we
are
not
covering
like
notifications
when
something
doesn't
yeah
can't
be
done?
Are
one.
E
E
If
you
think
that
these
are
strong
enough,
that
they
actually
should
be
included
as
their
own
descriptive
thing,
and
we
can
include
kind
of
those
two
principles
that
are
on
the
cutting
room
floor
with
all
the
notes
on
them
and
see
if,
if
people
come
back
with
strong
foods,
but
I
would
be
fine
with
that.
A
Except
for
except
for
except
for,
like
you
know,
intelligent
response
and
and
notifications
yeah,
I
think
so,
okay,
so
I
was
just
so.
I
tried
to
write
down
what
we
said
at
the
end
like
after
we
got
off
and
up
with
this
little
paragraph
is
about
these.
You
know
these
are
just
the
questions
from
this.
You
know
like
is,
but
I
just
didn't.
I
didn't
write
back
the
notifications.
C
And
you
know
what,
when
I
think
about
notifications,
I
think
that's
also
like
implementation.
Specifically,
I
think
what
we're
really
talking
about
here
is
that
agents
should
have
a
mechanism
to
escalate
to
humans
when
things
are
beyond
its
control,
yeah,
whatever
what
whatever
that
is,
whether
create
a
pull
request,
saying
you
know
what
your
version
is
wrong.
C
I'm
gonna
create
a
pull
request
to
switch
it
back
to
the
version
that
actually
did
work
or
I
need
your
help
or
you
know
what
I'm
saying
like
there's
got
to
be
an
escalation
like
like
I've
pre-meditated.
That's
what
I
want
to
look
for
meditated
escalation
agents
should
have
a
meditative
destination.
A
Yeah,
I'm
just
gonna
put
the
word
notifications
in
there
right.
This
is
a
little
bit
long,
but
I
think
you're
right
dan
right
is
this
what
you
were
writing
or
yeah?
I
think
you're
right.
This
is
exactly
so.
This
sentence
this
paragraph
right
over
here.
This
is
super
accurate,
but
it's
super
dry
and
I
don't
think
it's
gonna
contain
anything.
It's
just
basically
like
explaining
control
theory,
so,
like
I'm
gonna
just
like
delete
all
this,
because
I
think
this
is
really
what
we
mean
exactly
what
you
wrote.
A
So
I'm
happy
with
sending
this
to
well.
Also,
okay,
minus
the
question
about
the
source
of
truth.
I'm
happy
with
sending
all
of
this
to
just
as
it
is
to
the
maintainers
for
feedback
for
rc1.
A
E
D
A
You
know
pre
defined,
you
know
like
if
we
do
that,
then
it's
contained
in
there,
but
otherwise
we
don't
say
anything
about
what
happened
like
that.
It
should
even
do
anything
if.
A
A
A
A
A
There
we
go
yeah
all
right,
awesome,
we've
got
our
minutes,
but
we
achieved
the
goal.
So
that's
great
and
we'll
get
a
round
of
feedback
from
from
the
maintainers
pretty
rapidly.
You
know
for
rc
one,
because
that's
the
the
goal
or
get
ups
comes
coming
up
quickly.