►
From YouTube: Kubernetes kops office hours 20200327
Description
Recording of the kops office hours meeting held on 20200327
A
Hello,
everybody:
it
is
Friday
March,
27th
2020.
This
is
cups
office
hours.
I,
am
your
moderator
facilitator,
just
in
Santa,
Barbara
I
work
at
Google
a
reminder.
This
meeting
is
being
recorded
and
will
shortly
be
put
on
the
internet.
Please
therefore
be
mindful
of
our
could
have
conduct,
which
boils
down
to
please
be
a
good
person
and
I
or
someone's
already
pasted
a
link
to
our
agenda
in
the
chat.
Please
feel
free
to
add
your
name
to
the
attendees
list,
as
I
am
doing
right
now
and
if
you
do
have
any
items
for
the
agenda.
A
B
However,
we've
seen
many
cases
when
it
can,
the
cubelet
can
get
killed
by
mistake
and
some
some
really
weird
issues
and
by
switching
everything
over
to
system
D
only
it
seems
to
make
things
a
lot
better
and
I
bring
it
up
as
well,
because
cube
a
DM
seems
to
have
ran
into
the
issue
as
well,
and
they,
you
know,
had
platen
now
by
default,
push
everyone
that
direction
and
I
think
mini
cube,
and
so
my
main
prompt
was.
We
don't
have
a
flag
to
enable
this.
B
Yet
I
was
gonna,
add
a
flag
to
enable
it
and
then
build
on
top
of
that
to
add
it
easy
switch.
So
the
users
don't
all,
have
to
set
it
themselves
and
then
we
can
decide
later.
If
we
want
to
change
it
default.
Does
anyone
have
any
ideas
around
this
or
questions
or
thoughts
or
am
I
making
this
up
and
I'm
wrong?.
A
The
big
one
I
was
just
gonna
ask
which
I
think
you've
talked
about
is
like
the
extent
to
which
this
is
a
deviation
from
what
upstream
is
doing.
It
does
sound
like
at
least
some
of
the
people
have
done
it
upstream.
So
I
think
if
you,
why
don't
you
start
by
just
opening
an
issue
or
opening
up
here
or
an
issue
which
just
makes
the
change
like
that
and
links
to
the
upstream
issue?
So
we
can
just
have
a
look,
and
then
we
can
talk
about
how
best
to
introduce
yeah.
B
Yeah
and
that's
what
I
was
gonna
look
good.
We.
We
have
a
number
of
issues
that
have
been
open
for
about
three
years
on
this,
that
I
think,
but
they
don't
mention
the
secret
FS
system,
deep
problem,
but
I
think
that's
what
it
actually
referring
to.
We
have
a
number
of
them
that
would
fall
into
this
bucket,
so
yeah
I'll
try
to
group
all
that
together
and
open
up
a
PR,
but
I
just
want
to
make
sure
no
one
was
like
no,
no,
no.
No.
B
A
C
A
D
A
D
A
Sounds
great
I
think
it's
I
think
everyone
be
happy
to
see
basic
off
fully
disabled
in
119
that'll,
be
that
would
be
great,
I.
Think
I.
Think.
The
challenge
that
we're
likely
to
face
is
that,
if
we're
not
using
basic
auth,
we're
probably
using
client
certificate
authentication
and
that
client
certificate
authentication
doesn't
work
if
you're
behind
a
layer,
seven
like
essentially
yeah
like
an
it,
doesn't
work
through
nginx,
for
example.
A
It's
my
understanding,
any
layer,
seven
relay
I,
guess
you
can
call
it
or
or
something
and
there
seven
proxy
that
that
is
not
just
a
layer
for
pass-through
and
I.
We
are
not
using
that
as
far
as
I
know
in
cops
by
default,
so
we
should
be
okay,
but
I,
don't
know
if
other
people
are
putting
things
in
front
of
it
that
they
will
then
be
unhappy.
People.
E
E
A
I
mean
it's
not
like,
we
definitely
have
to
turn
off
from
119
entirely
if
we
can,
even
if
we
still
generate
them
and
they're
not
going
to
work
like
if
we
can't,
if
we
can't
forget
how
to
not
generate
them,
it's
not
the
it's,
not
the
end
of
the
world.
It's
certainly
not
great
like
if
it's
just
like
there,
but
not
used.
So
we
should
try
to
do
that
right.
E
A
Yeah
see
I,
don't
I
would
really
not
I
I
I
think
we
should
key
up
the
kubernetes
version
just
so
that
we
don't
have
people
stuck
on
an
older
cops
version
because
they
want
to
use.
Basically,
the
people
want
to
use
bicycles
with
the
people
that
are
using
it
through
a
like
through
and
nginx,
where
there
seven
that
bouncer,
okay.
A
A
Think
they're
gonna
back
for
it,
because
I
think
it
would
be.
That
would
be
a
very
big
breaking
change.
I
think
the
the
reason
might
be
for
like
more
security
and
give
people
a
heads
up
on
what's
coming.
If
we
did
it
to
1:18,
we
could
continue
to
have
a
feature
flag
or
something
we
could
have.
Then
you
have
some
mechanism.
I,
don't
say
feature
like
some
mechanism
to
turn
it
on,
whereas
in
119
and
then
one
upgrades
to
119,
basically
with
is
off
its
criminals.
119
basic
of
this
off
and
you
have
no
choice.
D
A
Cops
118
and
we
could
turn
it
on.
We
could.
We
could
turn
off
basically
Fault
in
118,
but
continue
to
have
a
the
ability
to
turn
it
back
on
I
know
how
do
people
feel
about
that
is
anyone?
Is
anyone
using
cube,
API
server
behind
a
or
in
a
way
that
they
would
require
a
basic
auth?
They
are
relying
on
basic
laws.
I
probably
am,
but
it's
not
a
big
deal
for
me
to
change
it.
Okay,.
A
A
E
A
You
for
thank
you
for
poking
me.
I
actually
tried
this
morning
and
it
sounds
like
the
process
has
changed.
It
may
have
changed
for
the
better
and
that
there
may
be
a
more
automated
process,
but
something
has
changed
so
I'm
trying
to
update
upload
those
videos
and
they
will
get
uploaded
one
way
or
another,
but
it
might
be
a
little
bit
longer
or
we
figure
out
the
new
process
trying
to
do
some
automated
thing,
and
we
might
have
fancy
info
titles
and
look
spectacular
as
well.
A
E
So
I
I
didn't
have
time
to
write
today.
I
did
digging.
One
of
the
thing
is
the
thing
to
take
into
consideration
is
that
nft
is
slower
than
legacy.
So
at
the
moment.
Yes,
it's
newer,
but
yes,
okay!
The
second
thing
is
the
auto
detect
in
kubernetes.
It's
an
auto,
detect
based
on
something
else,
actually
writing
something
before
it.
E
E
Reliably
I
mean
otherwise
depending
one
was
the
default
in
the
OS.
It
will
chose
that
thing
and
that's
it
you
once
something
has
written
their
other
components
will
write
also
and
once
the
CNI
write
something
that's
it
so
I
did
some
testing
in
there
and
really
the
only
way
to
do
it
is
to
do
it
in
additional
user
data.
Otherwise,
docker
may
start
first
and
if
it
starts
first
and
dad's
rules
to
the
IP
tables
table,
then
there
is
no
reason
to
set
up
alternatives
anymore.
E
A
E
E
E
A
A
A
Yes,
I
feel
like
there's
a
bunch
of
things
have
to
do
here.
It
sounds
like
yes,
your
issue
about
turning
off
docker,
auto
start
is
absolutely
right:
yeah
I
guess
it
will
hurt
anyone
that
is
using
the
image
for
other
things,
but
hopefully
there
aren't
too
many
people
doing
that
or
that
they
can
adapt.
A
E
A
It
was,
it
was
confusing
because
we
weren't
clear
what's
gonna
happen
upstream,
but
it
sounds
like
upstream
has
made
their
move
and
now
isn't
a
total
fix
right.
Oh
it's
complicated,
yeah,
so
yeah.
If
you
are
able
to
open
an
issue
that
would
be
wonderful,
okay,
even
but
this
would
be
a
tracking
issue
right.
There
should
be
a
like,
hopefully
like
collapse,
other
issues
into
this
and
link
out
to
other
issues
that
are
described
like
the
work
that
needs
to
be
done.
I,
hope,
okay,.
D
I
just
need
to
figure
out
how
we
want
to
be
enabling
it
I
linked
to
a
comment
on
the
very
long
issue
about
whether
we
use
a
target
or
a
feature
flag
and
I,
think
which
one
we
would
do
kind
of
depends
on
whether
we
would
want
to
continue
to
support
the
old
syntax
indefinitely
or
if
you
want
to
eventually
migrate,
to
only
support
the
new
syntax.
So
I'm
curious.
If
anyone
has
thoughts
on
that
or
anything
else,
I
mentioned
in
that
comment,
I.
A
Want
my
to
think
about
this
might
be,
it
does
feel
to
me
that
terraform
has
successfully
transitioned
to
zero
twelve.
It
doesn't
feel
like
they're
going
back
I,
don't
know
whether
any
people
that
use
terraform
out
there
are
are
like
resisting
zero
as
well
or
sort
of
hoping
that
there
is
some
alternate
path
or
whether
effectively
we
are
everyone
is
has
it
has
made
the
move
to
zero.
Twelve
I
know
that
a
migration
tool,
and
things
like
that
so.
D
A
A
The
fact
that
people
aren't
sticking
to
zero
eleven
would
imply
to
me
that
we
in
this
particular
case,
would
that
would
point
push
me
towards
a
feature
flag
that
would
actually
activate
the
old
configuration
so
that
we
can
deprecated
the
old
configuration
entirely.
That's
my
take
on
it.
I
don't
know
whether
this
is
like
an
ongoing
stream
of
these
changes.
I
would
hope
not
like
I
like.
Is
this
a
one-off
or
is
this
like
we're?
Gonna
have
the
same
conversation
in
two
months
about
0:13
I,
don't
know
yeah.
D
I
feel
like
this
will
be
a
little
brown
for
a
while,
but
you
know
I
put
a
vote
on
the
comment
and
there
are
more
votes
for
the
target
rather
than
feature
flag,
although
that
the
way
I
worded
it
suggested
a
feature
flag
to
enable
right,
0:12,
perhaps
a
feature
flag
to
prisoner
of
the
existing
behavior,
would
be
treated
differently
or
essentially
flipping
the
default
of
that
feature.
Flag,
yeah.
B
Second,
that
we
moved
everything
to
one.
You
know
to
12
and
I.
Think
personally,
I'm
not
worried
about
13
I,
don't
know
when
that,
if
they're
slated
to
come
out
with
that,
like
I
took
them
once
they
pre-announced
12,
it
took
like
a
year
for
it
to
come
out
like
they
said,
would
be
out
in
a
summer,
and
it
was
out
like
the
following
summer.
Basically
so
I
I
think
I
think
we
roll
forwards.
Personally,
it
would
be
my
opinion.
G
A
A
D
A
Passing
target
terraform
should
default
to
12
I.
Believe
you,
okay,
so
in
other
words,
it
is
a
breaking
change
to
target
era
form.
So
we
can
only
do
it
on
a
new
cops
release
and
it
would
be
a
headline
like
release
note
and
then
the
presume
it's
a
breaking
change.
But
yes,
it
would
certainly
be
a
big
release.
Note
and
then
the
feature
flag
is
an
accommodation
for
people
that
are
not
yet
ready
to
make
the
change
and.
G
A
D
A
A
A
A
H
So
what
I
wanted
to
propose
was
adding
either
V
CPUs
memory
filters,
possibly
more
and
then
using
the
new
describe
instance,
types
API
from
AWS
to
kind
of
figure
out
that
the
correct
instance,
types
to
use
I
think
this
would
be
useful
for
just
like
Brazilian
auto-scaling
groups
in
general,
but
it's
also
imperative
for
spot
instances
so
that
if
one
capacity
pool
is
interrupted,
it'll
be
able
to
move
on
to
a
different
spot
capacity
pool.
So
I
just
wanted
to
get
some
thoughts
on.
H
If
the
CLI
was
the
right
place
to
add
this,
if
it
should
be,
on
instance,
group
creation
and
cluster
creation,
or
if
it
should
just
go
in
cluster
creation,
since
it
seems
like
all,
the
arguments
are
only
on
cluster
creation
right
now,
yeah
I
just
want
to
get
some
thoughts
on
that
and
see.
If
there's
any
red
flags
before
proceeding
with
implementation,.
A
A
Older
yeah,
I
guess
for
like
the
expanded,
the
expanded
form
like
so
I
mean
one
way
to
do.
This
is
when
you
do
cups.great
cluster.
We
evaluate
your
request
against
the
describe
instance.
Groups
describe
instance,
types
at
that
particular
moment
in
time
and
we
say
you're
a
best
match
for
an
m3
medium,
so
you
are
m3
medium
forever.
A
The
other
way
would
be.
We
push
that
down
into
an
AWS,
auto
scaling
group
that,
like
is
able
to
match,
say
today,
you're
an
m3
medium
and
then
in
a
year
whatever
it
is,
when
I'm
seven
mediums
the
launch
of
a
better
match
for
m7
medium,
and
so
we
have
much
more
instances
would
like
switch
you
to
that,
because
we're
able
to
like
keep
your
preferences,
I
guess
the
question
is:
is
it
more
the
former
or
the
latter
yeah.
H
More,
the
former
because
it
would
be
like
with
mixed
instance
groups.
You
can
do
multiple
instance
types,
but
again,
that's
at
you
know
whatever
instance
types
are
available
when
you
run
the
command,
there's
no
way
currently
to
add
instance,
types
as
they're
released
as
far
as
like
control
planes
on
auto
scaling
groups.
A
C
C
H
The
create
instance
group,
because
that
seems
like
something
that
you
might
actually
want
to
be
static.
You
wouldn't
actually
want
to
reevaluate
instance
types
as
they're
released
in
case.
You
know,
there's
some
parameter.
You
didn't
think
of
that
causes,
like
detrimental
performance
costs,
your
application,
but
if
it
was
on
instant
group
instance
group
create
create
you
could
you
know,
select,
tenants
and
stipes
that
match
your
configuration
and
then
you
that
static
forever.
C
A
Is
this
something
which
we
could
if
we
persisted
the
intent?
You
know
your
preferences,
your
I
think
you
used
memory
and
V
CPU
as
your
as
your
filtering
criteria.
If
we
persisted
that
and
then
also
persisted
the
lock
like
you
know,
I'm
sort
of
like
imagining
like
that,
whatever
language
you
want
to
use
that
has
a
version
specifier
in
a
locked
file
like
you
sort
of
keep
the
version,
specifier
honoré
or
instance,
type
specifiers,
and
then
you
have
a
lock
and
it
will
tell
you
an
update
is
available.
Do
you
want
to
do
that?
A
H
Yeah
I
think
that
beyster
I
think
that
would
be
useful.
I,
don't
know
what
others
thoughts
are
if
it
would
be
useful,
just
cuz
I,
don't
see
it
like
changing
so
I
think
versioning
kind
of
does
make
sense
for
the
instance
type
selection
and
then
locking
it
down
to
that
version
like
just
so
that
you
don't
have
dynamic.
Weird
changes,
like
others,
are
saying
on
like
qualifying
instance,
types
and
stuff
you've
got
to
re-evaluate
every
once
in
a
while
anyways,
so
maybe
just
like
rerun.
C
A
Command,
if
you
want
it
to
protect
this,
as
it
were,
a
toolbox
command.
So
what
is
our
as
a
playground,
and
you
can
certainly
build
a
tool
box
command
to
recommend
instance
types
or
something
like
that,
and
that
would
be
a
great
way
to
discover
to
what
extent
this
is
useful
and
like
to
what
extent
do
I
have
to
specify
a
super
long
string.
That
is
basically
like
the
equivalent
of
specifying
me
like.
Will
this
anyway,
because
I.
C
A
Like
I
know
that,
like
I
was
imagining
I'm
gonna
specify
this,
but
I'm
also
going
to
say
the
architecture
right,
because
otherwise
I'll
get
an
ARM
chip
that
just
like
drops
in
there
and
we
don't
farm
yet
and
yeah,
certainly
so,
but
if
you
wanted
to
do
this
is
a
toolbox.
Demand
I,
think
that
would
be
a
wonderful
way
to
to
discover
what
the
semantics
should
be.
Okay,.
A
B
Like
this
idea,
by
the
way,
so
I
think
this
is
a
really
good
direction
to
head
to
in
the
toolbox
command.
We
would
totally
use
that
because
I
don't
really
care
what
the
specifics
of
my
instances
are.
It's
really
about
memory
and
CPU.
So
there
was
just
a
command.
I
could
just
say
every
once
in
a
while
run
this
and
then
we
well,
we
don't.
We
also
have
tooling
around
cops,
but
we
can
then
check
that
in
and
then
you
know
now
we
have
our
new
static
version.
C
Yes,
so
so
we
want
to
take
privileges
away
from
core
DNS
and
to
do
that,
we
have
to
change
the
service
in
a
way
that
cops
fought.
Our
fuel
apply,
make
the
hash
of
so
this
is
a
proposal
to
have
a
field
in
the
channel
which
says,
if
you're
upgrading
from
versions
before
this
do
a
little
replace
instead
of
a
dupe,
it'll,
apply
so
I
think
it's
okay,
I,
don't
want
to
lgt
a
mint
because
it's
from
a
co-worker,
so
I
just
want
to
get
a
sense
of
whether
that's
preceding
or
not
I.
A
This
is
this
is
sort
of
a
proposal
that
I
think
is
happening
elsewhere
in
like
the
apply
logic,
so
like
that's
sort
of
the
other
way
that
people
are
think
about
this,
so
you
would
essentially
annotate
the
new
service
with
a
I
wanna,
say,
apply,
method
replace,
but
you
know
like
some
of
the
got
and
then
it
would
know
it
whatever
the
the
it
was
that
was
applying.
This
would
know
not
to
apply,
but
in
fact,
to
replace
I,
don't
know
well
well,.
A
A
Will
let
me
I
will
write
down
this
one
on
my
lesson.
What
I
will
try
to
do
is
I
will
try
to
find
the
status
of
the
alternate
proposal,
I
guess
the
anti
annotation
tie
proposal
and
then,
let's
look
at
that
on
the
issue
in
our
on
the
PR
and
discuss
I.
Think
this.
This
is
a
good
proposal.
I
think
the
it
might
put
more
burden
on
us
than
we
want.
A
A
F
C
A
C
A
C
F
C
A
C
A
C
C
G
C
J
J
D
J
Put
up
I
mean
it
also
supports,
claim
a
flag
on
it
or
something,
and
instead
of
just
removing
it's
like
I,
feel
like
this
is
gonna
get
fixed
and
then
we're
gonna
put
it
back
in
right.
There
is
some
work
going
on
on
it.
If
you
follow
through
on
the,
but
if
you
follow
down
some
of
the
tickets,
there
are
people
looking
at
it.
It
may
be
going
slower
than
I'd
like
for
certain,
but.
C
C
G
C
A
C
A
A
A
I
So
basically
I
worked
with
sulla
Marcus
and
he
has
asked
me
to
bring
this
up.
That's
pretty
much.
Why
I'm
here
and
I
just
wanted
to
kind
of
see
how
this
goes.
I've
only
first
seen
this
pull
request
a
couple
days
ago,
haven't
really
looked
at
it
too
much
but
seems
fairly
straightforward,
so
I
guess
maybe
somebody
else
has
some
opinions
or
has
looked
at
it.
I
think
he
wanted
to
just
sort
of
nudge
people
about
it.
A
And
to
be
clear,
this
is
to
use
a
second
instance
of
sed
or
a
third
instance
I
Technic
technically
another
a
separated
instance
of
sed
to
support
psyllium,
and
is
this.
This
is
an
option
right.
This
is
not
required,
so
people
can
find
still
run
psyllium
with
their
with
the
API
server,
backed
approach.
C
A
A
C
In
the
one
six
ish
of
helium's,
the
TRD
only
does
not
scale
beyond
50
notes,
so
you
need
the
sed
to
help
propagate
things
more
efficiently.
It
looks
like
in
of
Helium
one
seven
or
one
eight
they're,
going
to
make
performance
improvement
to
where
they're,
even
considering
removing
supporting
backing
at
CD.
But
you
know
in
the
meantime,
yeah.
A
A
Pr
looking
the
PR,
it
seems
pretty
reasonable,
like
the
the
bulk
of
the
changes
are
in
our
test
output
so
that
that
seems
like
it
is
not
a.
It
is
not
a
incredibly
difficult
one
to
carry
and
and
and
I
think,
maybe
the
multi
cluster
one
is
the
the
one
that
tests
the
edge
as
I
understand
it.
If
you
want
to
do
multi
cluster
services,
multicast
or
something
you
have
to
use
at
CD
with
like
that,
some
sort
of
CD
is
the
backing
store.
A
Okay,
yes,
also
because
it's
yeah
we're
gonna
we're
gonna
lose
canal.
We
want
to
make
sure
we
have
psyllium
happy.
Alright,
we
have
one
last
item
on
the
agenda,
which
is
sort
of
time
permitting
item
from
Peter,
which
is
about
how
to
update
the
kids
dependencies.
I
can
always
make
a
video
or
something
of
that
or
put
a
dot-com
about
that
list.
I
propose
you
go
through
the
release
plan
first,
but
first
before
we
do,
are
there
any
other
agenda
items.
A
A
So
it
looks
like
thank
you
to
every
cop
with
this.
It
looks
like
we
have
a
person
listed
copy
from
last
week
and
it
looks
like
I
have
not
created
the
Buster
ami
I
will
again
try
to
do
that.
I
think
this
is
now
the
third
week
running.
The
shame
is
REO
I
apologize,
the
so
we
we
have
118
0
alpha
another
119
0
alpha.
If
anything
comes
up
in
terms
of
like,
if
there's
a
significant
reason
to
do
that,
I
think,
once
we
merge,
we
went
off
once
we
move
more,
it's
a
removal.
A
If
you
went
off
alone.
That
would
be
a
significant
reason
to
do
so.
So
why
don't
we
change
this
from
a
if
anything,
to
a
merge,
V
1,
alpha,
1
removal
and
then
do
the
next
alpha,
so
we
can
expect
so
we
can
expect
the
next
V
1
the
next
118
0
alpha
shortly.
If
there
are
other
things
you
would
like
to
get
in
there,
then
let's
do
that
quickly.
The.
D
A
B
A
A
This
is
like
the
introduction
of
Buster,
where
we
expect
early
adopters
to
try
it
out,
and
then
we
will
deprecated
based
on
the
idea
that
Buster
is
good
enough
and
then
once
we
deprecated,
we
expect
people
to
switch
to
it
or
start
switching
to
it
in
greater
numbers,
and
so
I
don't
do
that
until
it's.
So
we
feel
like
it's
a
sensible
choice
and
it's
I'm
sure
it's
a
sensible
choice,
but
you
know
we
don't
have
data
on
that.
Yet,
okay,
thanks.
E
By
the
way
can
I
ask
somehow
related
question:
would
it
be
an
option
to
start
having
an
Ubuntu
image
in
the
near
future
or
make
it
a
default?
The
reason
I'm
asking
is
that
buster
is
here
to
stay
for
another
two
years
with
already
a
pretty
old
colonel,
the
only
distro
with
long
term.
Support
that
will
have
a
newer
colonel
would
be
Ubuntu
2004,
which
will
be
released
next
month.
E
A
Is
a
good
idea
in
the
past,
so
in
the
past
the
the
non
LTSs
of
a
bun
to
have
proved
problematic
in
terms
of
their
longevity,
like
they
don't
last
for
as
long
as
the
community's
release,
which
is
just
that
was
really
ordered,
and
then
the
other
problem
in
the
past,
with
the
bun
to
LTSs,
was
that
the
timing
was
sort
of
off
in
terms
of
what
made
the
LTS
is.
However,
yeah
2004,
which
is
next
month.
A
A
So
I
actually
think
we
could
have
both
I,
don't
I,
don't
want
to
sort
of
expand
the
support
surface
too
much,
so
we
should
probably
try
to
figure
out
which
one
it
is
that
we
recommend
I
wouldn't
want
to
recommend
both
but
I,
don't
know
how
other
people
feel
about
recommending
one
or
the
other.
We
don't
have
to
decide
this,
yet
we
can
certainly
produce
more
images
in
the
same
way
that
we
produce
are
going
to
produce
a
Buster
and
stretch.
A
E
E
J
A
A
D
D
A
Which
is
I
will
describe
the
challenge,
which
is
the
tags.
When
you
use
a
tag,
it
is
currently
replaced
by
the
char.
So
that's
why
we
have
the
the
longer
shower.
Jhin's
are
the
ones
which
are
actually
used
and
then
underneath
we
have
the
commented
out
versions
which
are
in
human,
readable
form,
and
if
you
want
to
use,
if
you
want
to
update
their
place
directives,
you
effectively
delete
the
shower
ins
uncomment,
the
human
versions,
change
the
to
the
correct
tags
copy.
It
comment
out
the
block
for
the
next
person
to
come
along.
A
It
might
be
the
case
in
118.
I,
don't
think
it's
gonna
be
the
case
in
117,
I.
Think
there's
some
roots,
there's
an
odd
behavior,
which
is
that
the
replace
in
a
dependency
is
not
carried
through
is
how
I
remember
it
but
yeah
anyway.
That's
that's
why
there
is
this
weird
common
double
form
and
the
commented
form,
and
it's
not
too
bad,
it's
just
it's
quite
maybe
a
screen
a
screenshot
will
be
the
best
way
to
do.
It
make
a
little
video
to
show
this,
because
it
is
surprising
and
not
not
terribly
difficult.