►
From YouTube: Kubernetes SIG K8s Infra - 20220427
Description
A
Hello,
everybody
welcome
to
the
sick,
guys
infra
meeting
of
27
april
2022.
This
meeting
is
being
recorded
and
we'll
be
streaming
to
youtube
later
and
we
all
abide
through
this
ncf
code
of
conduct.
So
please
be
excellent
to
each
other.
So
let's
jump
into
the
agenda
on
what?
If
you
want
to
share
the
screen
and
do
the
data
studio
report
all
those
things
that
would
be
nice.
B
B
E
A
All
right,
there's
two
items
for
discussion
today
is
the
one:
is
the
board
review
which
I'll
put
a
link
into
there?
We're
gonna
have
a
look
at
the
board
and
then,
oh
sorry,
three
there's
a
lot.
Let's
first
go
before
we
do
the
board
review,
because
I
think
that
should
be
the
bulk
of
our
discussion
of
who
does
what
and
when.
If
you
want
to
go
first
with
your
proposal
cancelled
the
meeting,
I
assume
that's
around
kubecon.
B
A
All
right
who
would
update.
F
G
And
getting
consensus
on
that,
I
think
by
the
end
of
last
year
we
had
a
kept
because
we'd
been
asked
to
go
through
to
the
sig
release
engineering.
So
we
attended
several
of
those
meetings
trying
to
drive
a
shared
cross
sig
document
and
that
kept
seems
to
have
been
the
right
place
to
do
that
and
at
the
beginning
of
this
year,
then
hey
good
timing,
ben
kind
of
pushed
that
from
the
kep
into
the.
C
I
mean
I
I
still.
I
still
don't
think
that
this
is
something
that
should
require
cap,
and
I
thought
we
had
an
agreement
on
that.
That
said,
we
have
other
means
where
we
still
probably
don't
have
to
do
this,
but
it
would
be
a
good
idea
just
to
have
better
synchronization
expectations.
I
Ben
you
have
to
give
a
little
bit
more
context
on
what
you
were
working
on
this
week.
Basically,
the
flow
diagram
of
like
the
incoming
request
and
what
happens
with
it.
What
kind
of
checks
is
done?
You
know
what
are
the
redirects
look
like?
I
think,
if
we
explain
that
a
little
bit
more,
I
think
the
release
engineering
team
will
be
more
comfortable
with
you
know,
some
of
the
things
ben
you.
H
H
Oh
yes,
please
do
one
of
you,
so
yes,
so
so
the
original
idea
that
bannon
and
others
had
was
that
we
would
do
the
layer
block
mirroring
during
the
the
image
promotion
process
right
and
that,
because
of
that,
when
the
final
promotion
step
occurred
and
a
container
image
was
promoted
to
registry.case.io,
then
it
would
be
assured
that
all
of
the
s3
objects
that
comprise
the
image
layers
for
that
promoted
image
would
already
be
replicated
to
all
the
mirrors.
H
Ben
put
together
a
solution
where
we're
essentially
mirroring
those
object
layers
to
the
s3
mirrors
after
the
promoted
image
makes
it
to
the
registry
and
instead
of
the
registry,
being
assured
that
all
of
the
image
layers
are
in
all
of
the
mirrors
ben
added.
A
little
check.
A
little
http
head
call
to
to
determine
whether
or
not
that
particular
image
layer
exists
and,
if
not
just
simply
follow
the
code
path,
as
it
normally
exists
to
gcr.io.
H
So
there's
a
little
bit
less
efficient
in
the
registry.case.io
worker,
but
we
kind
of
get
around
the
ordering
problem
that
we
were
having
with
the
promotion
layer
ben
did.
I
explain
it
properly.
C
Yes
and
I'll
add
that
we're
like
we've,
got
like
a
very
simple
attempt
at
caching
those
checks
under
the
assumption
that
no
one
should
be
deleting
players
please.
C
So
if
we've
observed
a
layer,
it
should
continue
to
be
available
for
that
bucket.
The
only
thing
being
that
you
know
these
are
serverless
auto-scale
processes
and
we're
caching,
this
in
memory,
so
as
processes
get
cycled
in
and
out,
we
we
don't
persist
that
at
the
moment,
depending
on
how
we
feel
about
this,
we
have
the
option
to
move
that
out
or
if
we
come
back
and
have
the
ordering
back
to
the
original
plan,
then
we
can
potentially
drop
the
check.
The
upside
is
also.
This
does
make
the
registry
more
robust.
C
The
sandbox
instance
with
this
enabled,
but
the
moment
we
rolled
that
code
out,
the
ci
didn't
work
because
there
was
layers
that
hadn't
been
promoted
yet
that
cops
clusters
were
trying
to
use.
By
doing
this,
we
are
able
to
continue
developing
an
e-testing
without
blocking
on
having
that
particular
ordering.
C
C
You
know
99th
percentile
latencies
on
the
proxy
that
is
actually
close
to
the
bucket.
Today,
we've
seen
those
hit
about
90
milliseconds,
which
is
probably
fine,
but
it
feels
bad
in
comparison
to
one
or
two.
H
Yeah
and
it's
not
it's
not
prohibitive
at
this
point
right,
so
I
I
would
and
that
that
is
not
to
say
that
I'm
not
actively
working
on
the
code
to
mirror
those
object,
layer
blobs
during
the
promo
process
as
well.
It's
just
because
of
the
ordering
thing
I
think
mostly
related
to
the
signing
process
or
step
within
the
promotion
process.
H
I
Right
right,
the
other
thing
that
I
want
to
add
here
is
when
we
know
that
the
traffic
is
not
coming
from
aws.
It's
a
plane
redirect
at
that
point
to
existing
gcr
stuff
right.
So
right,
the
like
graceful
degradation
of
you
know
worse,
we
have
okay,
the
the
image
is,
image
is
not
in
the
s3
cache
and
we
have
to
do
a
head
call
and
we
have
to
you
know,
check
that
and
redirect
them
back
to
gcr.
That's
like
the
worst
case
scenario.
Well,
that's.
I
H
That's
the
only
code
path
right
now
for
image
layers.
We
always
do
the
head
check.
Well,
may
we
may
be
able
to
get
to
a
situation
where
we
assume
the
layer
is
present,
and
I
don't
know
we
do.
We
do
some
sort
of
head
check
after
the
fact.
C
Great,
I
think
what
tim
was
getting
at
is
that
as
much
as
possible
throughout
the
code
anytime,
one
of
our
assumptions
is
violated.
We
just
go
back
to
a
plane,
redirect
to
gcr
completely
unassuming
the
only
time
we
do
anything
other
than
a
plane
redirected
gcr
is
when
we're
able
to
validate.
This
is
a
request
for
a
blob.
It's
a
request
for
a
blog
with
a
sha-256
hash
and
the
format.
C
We
expect
it's
a
request
for
a
blob
with
a
sha-256
hash
that
we
are
able
to
identify
the
client
ip
for
and
the
client
ip
is
in
a
known
aws
region
and
that
region.
The
bucket
that
we
have
selected
for
that
region
has
the
how's
that
then
we
will
redirect
s3
and
anything
else
happens.
We
just
go
back
to
just
a
plane,
just
go
to
gcr
with
exactly
the
path
you
gave
us
unaltered.
C
Right,
we
have
a
further
enhanced
that,
like
this
thing,
just
gracefully
degrades
back
to,
I
don't
know,
what's
going
on,
just
go
to
dcr
exactly
exactly,
and
we
and
we
are.
I
am
observing
that
we,
even
with
just
a
small
scale,
load
that
we
have
now.
We
do
see
that
the
generally
we're
able
to
keep
an
instance
running
and
that
instance
recalls
that
it
has
already
seen
these
layers
and
most
of
the
requests
are
getting
hits
right.
I
And
and
the
best
thing
about
this,
the
way
we
we
ended
up
was
that
we
are
unblocking
the
release
engineering
folks
from
being
on
the
critical
part
of
the
things
right.
So
yeah
go
ahead.
C
Not
only
can
we
continue
to
iterate
on
the
details
of
the
proxy
be
pretty
reasonable
to
continue
moving
towards
rolling
out
this
implementation.
I
think
we
should
you
know,
do
some
additional
performance
checks
and
things
since
we've
changed
that
a
bit,
but
it's
looking
good,
given
that,
then
the
ensuring
the
layers
are
available
becomes
an
asynchronous
task.
That's
not
blocking
that
doesn't
need
to
be
part
of
the
image
promoter
at
all.
The
current
implementation
we
have
is
ping
our
know
and
ask
him
to
run
an
rclone.
I
And
the
twist
in
the
end
is
when
we
do
end
up
figuring
out
the
cloud
flare
or
something
else.
It
becomes
even
more
simpler
at
that
point,
but
let's
not
worry
about
that
right
now.
C
I
So
let's
go
with
this,
so
the
main
question
I
have
here
would
be
for
a
caleb
and
hippie.
Like
did
you
understand
the
sequences?
Do
you
have
any
questions
on?
You
know
what
what
we're
talking
here.
I
All
right,
so
we
should
write
down
the
sequence
that
we
were
talking
about
and
reflect
that
in
the
cap
update
the
cap.
So
we
are
all
in
the
same
page
and
we'll
tweak
it
when
we
want
to
update
the
code
with
other
things
or
if
the
sequence
changes
right.
Yep.
G
C
I
think
we're
moving
this
now
to
this
is
still
an
like
an
ideal
approach.
That
would
be
a
good
thing
to
do
long
term,
but
it
is
I
don't
it.
It
seems
like
given
real
word
performance
data
of
the
alternate
approach
that
we
should
not
block
on
this.
G
C
I
I
I'll
give
you
another
view
right.
Like
last
time
when
porco
was
talking
about
hey,
I
need
to
be
able
to
validate
using
the
the
chain
guard
validator
stuff.
What
was
the
name
porko.
I
So
what
was
the
validator
that
you
were?
What
was
the
command
line
for
the
validator
for
the
images
cosine
yeah
cosine?
Sorry,
so
yeah.
I
Sorry,
yeah
so
cosine,
so
when
cosine
wants
to
when
somebody
wants
to
validate
an
image
signature
with
cosine
first,
they
will
get.
I
mean
either
way.
If
the
image
the
image
for
the
signing
itself
is
not
in
the
s3
bucket,
then
it
would
redirect
to
gcr
at
that
point,
so
the
validation
would
still
work.
I
So
this
was
the
point
that
we
were
trying
to
make
hippy.
So
even
if
earlier
the
problem
we
had
was,
we
wanted
both
the
signature
of
the
image
and
the
image
itself
in
the
s3
bucket.
Only
then
the
validation
would
happen
now.
With
this
head
check,
we
could
the
we
will
end
up
giving
the
image
information
back
to
cosine.
F
Yeah
that
it
may
work,
but
then
you
will
end
up
with
uneven
copies
of
the
registry,
so
I.
F
C
C
I
So
purchase
so
essentially,
what
we
are
saying
is
like
you
can
do
it
in
your
own
time.
You
know
you
don't
have
to
do
that.
First
before
we
can
roll
this
out.
That's
basically
what
we
are
saying.
What.
F
I'm
saying
is
creating
a
process
that
only
copies
the
actual
container
layers
that
are
not
signatures
going
to
involve
more
work.
You
know.
I
C
Right,
it
will
still
serve
them
from
gcr
in
that
case,
instead
of
from
s3
and
also
the
verification
we
run
probably
won't
see
it
because
it'll
be
running
in
gcp,
so
we'll
all
be
from
gcr
in
that
case.
But
if
a
client
runs
it
and
they're
trying
to
check
this
and
they're
inside
aus
and
we
detect
them
as
insight
advs,
they
will
receive
the
layers
either
from
s3
or
gcr,
depending
on
availability,
but
they
will
be
identical
either
way
and
they
will
always
reserve.
C
They
will
always
receive
all
of
the
layers
just
depending
on
availability.
We
may
or
may
not
be
able
to
ship
the
traffic
to
amazon
and
we're
checking
as
we
go.
A
C
Cash
hits
we
log
cash,
hits
and
misses,
but
also-
and
we
log
redirecting
to
aws
but
again
it
that
we'll
be
able
to
see
the
cost
from
doing
that.
And
when
we
run
the
sync,
the
sync
just
syncs
everything
and
it's
already
responsible
for
ensuring
that
everything
exists.
I
Yeah,
we
are
not
getting
rid
of
the
sync
process.
We
are
still
going
to
do
that.
We
are
still
going
to
do
the
image
promoter
changes.
We
are
still
going
to
do
all
the
things
that
jay
is
doing
to
copy
things
from
one
to
the
other.
Everything
is
going
to
be
so
what
we
are
saying
here
is
oca
proxy
will
be
handled
able
to
handle
all
the
cases,
and
you
know
till
we
have
everything
up
and
running
right
and
once
we
have
everything
up
and
running,
then.
C
A
And
then
the
way
I
understand
because
we're
going
to
continuously
think
lighter
if
something
is
missing,
it's
just
a
question
of
it's
masking
now
and
next
thing:
it
will
be
there
exactly
right,
okay,
so
it's
not
the
long
term.
The
thing
is
missing
and
we
continue
to
lose
money
because
it's
not
there
right.
C
No
and
then,
depending
on
how
long
you
know
it
takes
for
all
these
other
things,
we
have
the
option
to
put
in
a
continuous
asynchronous
post-promotion
sync
very
cheaply,
just
running
existing
tools
in
a
loop
just
so
we
have
a
stop
gap
of
making
sure
that
the
layer
should
be
getting
populated
until
we
can
do
the
smarter
thing
and
sync,
it
continuously
online
more
effectively
when
we
know
that
we
need
to
sing
something
yeah,
we
could
just
run
like
a
daily
sync
or
something
yeah.
I
So,
going
back
to
what
hippie
was
talking
about
ben,
we
we
need
at
least
a
small
markdown
file
when
you're,
when
you
and
j
are
making
changes
to.
You
know
the
code
path
of
the
incoming
request
and
where
it
goes
and
how
things
get
served.
We
should
have
at
least
a
small
markdown
where
which
we
are
updating
when
we
are
updating
the
pr.
C
To
have
I
want
to
have
this,
I
want
to
have
a
doc
about
the
testing
regimen,
but
at
this
point
we
just
put
something
out
trial
it
in
the
sandbox.
It's
looking
good,
so
we
need
to
get
this
documented
as
th
this.
This
is
what
we're
doing.
I
Yes,
I
agree
with
you
and
then
what
the
two
things
that
you
mentioned
then
hippie
and
caleb
can
take
that
and
roll
it
in
the
cap
and
we'll
we'll
do
the
you
know
approvals
and
whatever
else
is
required
for
that.
I
C
Yes,
I
see,
oh,
I
got
a
treasure
train
went
by
eddie.
I
couldn't
hear
us
apparently.
B
I
Yeah
we
have
to
roll
it
out
as
quickly
as
possible.
So,
yes,
we
should
try
to
do
alpha
with
what.
So,
essentially
we
are
hydrating
here
right
like
so
when
ben
and
jay
update
the
the
dock
and
the
markdown,
we
roll
it
into
the
cap,
and
we
run
through
the
process
of
you
know
getting
approvals
and
pushing
it
into
alpha
when
we
do
125..
I
I
125
I
we
should
be
able
to
change
the
code
between
feature
freeze
and
code
freeze.
We
should
be
able
to
change
the
code
to
point
to
the
new
registry.
B
Okay;
okay,
without
redirect
to
s3.
I
C
Yeah,
I
think
we
have
it.
We
have
a
different
timeline
on
when
we
feel
comfortable
rolling
those
that
functionality
over
to
the
to
the
right
to
the
production
domain,
as
opposed
to
when
we
feel
comfortable
that,
like
we
can
use
the
domain
at
all.
B
C
C
I
We
should
be
able
to
do
that
once
we
release
124
immediately,
then
we
can.
I
C
Right
but
at
the
moment
we're
not
going
to
get
good
data
because
the
thing
that
we
see
is
we
have
one
bucket
and
one
region,
so
we
have
terrible
latency
from
the
other
regions.
So
I
think
the
remaining
blocker
we
have
for
any
of
the
tracks.
Making
progress
is
finishing,
getting
the
initial
sink
into
the
per
region
buckets,
but.
I
One
other
question
one
second,
so
the
other
question
I
had
for
jay
was
j.
Do
you
have
some
ci
jobs
that
are
running
in
aws
that
we
can
point
them
to
repository
dot
case
dot,
io.
H
Okay,
but.
C
We
have
that
today.
What
I'm
saying
is
to
that:
probably
the
next
step
prior
to
us
feeling
comfortable
about
rolling
this
to
register
a
case
studio,
seeing
performance,
and
for
that
we
need
to
get
it
to
get
good,
realistic
performance.
We
need
to
get
the
regions
in
place,
because
the
main
performance
issue
I
see
right
now
is
when
we
are
distant
from
a
bucket
and.
C
We
do
have
a
k,
ops
job.
I
would
like
to
turn
on
more
more
kops
and
gcp
jobs
just
to
be
just
to
be
sure
before
we,
because
this
is
a
yeah.
This
is
a
decent
amount
of
change
versus
the
the
plane
redirect.
B
C
That
makes
sense
yeah,
I
think
doing
that
plus
tur
plus
pointing
some
more
ci
jobs
at
it
will
give
us
some
additional
confidence
and
performance
data.
Yep.
B
Any
ep:
do
you
need
more
information
for
the
cape.
G
Go
back
through
this,
I
didn't
wasn't
able
to
type
during
the
call,
so
I'll
need
to
go
back
and
listen
again.
Okay,
I've
listened,
I
feel
like
we
have
enough
information
there.
It
would
be
useful
if,
when
we're
capturing
these
things,
we
do
add
stuff.
We
we're
doing
a
lot
in
talking
here.
We
do
a
lot
in
chat,
but
we
don't
do
a
lot
of
issues.
We
just
create
prs.
F
G
And
so
it's
hard
to
attract.
I
know
that
there's
a
movement
happening
we
try
to
keep
following,
but
as
far
as
trying
to
be
able
to
help,
because
I
want
to
right
now
from
the
ii
team
trying
to
show
up
and
know
what's
valve
of
value
is
has
been
hard,
and
so
I
do
feel
like
I've
got
enough
to
update
the
cap
and
that
my
primary
ask
from
this
group
is
to
go
through.
Take
this
information
just
talk
about
and
put
it
in
the
cap
and
that's
I
don't
want
to
think
that's
a
blocker.
I
Yeah
we
hear
you,
let's
do
the
first
rev
and
then
we'll
be
more.
You
know
cognizant
about
leaving
comments
on
the
kept.
Thank
you
and
doing
the
markdown
and
the
other
document.
That
ben
was
talking
about.
C
Yeah
sorry,
I
meant
to
be
able
to
come
to
this
meeting
with
these
things
documented,
but
dim
surfaced
the
whole
pile
of
bugs
for
me
this
morning.
So
I'm
sorry.
C
Oh,
no,
those
absolutely
need
some
of
those
are
very
concerning
go
for
it.
B
Okay,
any
other
subject,
I
think,
we're
good.
Do
we
a
question
for
jay
and
caleb?
What's
the
technical
blocker
for
the
production
bucket.
H
I
I
can
go
ahead.
I've
been
yeah
having
some
some
issues
with
with
I
am
and
organizations.
Recently
anyway,
I've
figured
out
a
path
for
caleb
on
the
getting
the
I
am
role
created
within
the
sub
account
appropriately
and
I'll
I'll
coordinate
with
you
tomorrow.
If
that's
cool
with
you,
we
need.
H
E
F
A
From
jay's
side
we
should
have
those
buckets
before
the
end
of
the
week,
okay,
but
is
that
that
about
writer?
A
B
Okay,
so
for
next
meeting
we
have
the
bucket
we
have
s3
bucket
somewhere
with
a
copy
of
the
image
blobs,
and
we
have
oci
proxy
in
sandbox
environment
with
reduxing
to
it
right.
C
Yes
and
there's
still,
there's
still
there's
still
ongoing
technical
work
around
around
doing
sinking
in
the
promoter,
whether
or
not
the
everybody
that
we're
still
that's
still
happening.
B
Okay,
ryan:
you
do
you
want
to
go
over
the
board
or
because
we.
A
At
the
moment,
the
board
is
not
heavily
used
directly
because
there's
not
much
issues
coming
in
and
pro
related,
so
what
I'm
doing
to
keep
us
up
to
date?
If
you
look
at
the
board-
and
you
see
the
the
second
last
column
where,
if
you
open
the
board
quickly
I'll
open
my
site,
so
I
can
speak
to
it.
A
A
So
I
see
ben
is
spitting
out
prs
like
it's
nobody's
business,
so
I
catch
the
pr's
as
they
go
past
my
my
email
and
I
put
them
in
validation
to
do
so.
Anyway.
I
have
looked
in
validation
to
do
column.
That's
the
ps
that
came
in
since
friday
friday.
I
cleared
out
all
the
work
there
was,
I
think,
50
some
odd
tickets
in
the
validation
to
do
columns
so
end
of
every
week.
So
if
anybody
have
any
objections
about
things,
certain
validation
to
do,
let
us
know
an
implementation
to
do.
A
G
H
G
B
I
did
a
review
on
on
off
the
board,
but
I
think
most
of
of
what
we
say
in
this
meeting
capture
accepts
the
the
kept
part
which
is
basically
update
with
any
process
we
implemented
lci
proxy.
Most
of
the
aspects
of
what
we
talked
today
are
already
captured
as
issue.
E
G
One
of
the
things
that
we
were
with
that
might
be
useful
is
when
we're
wanting
an
issue
to
be
created
and
it's
not
created,
and
we
talk
about
in
the
meeting
just
in
the
text
field
at
the
bottom
of
this
board,
you
can
type
in
just
a
string,
and
then
that
becomes
a
draft
issue
you
can
assign
that
to
a
person
so
that
the
creation
of
the
issue,
even
though
it's
not
there
yet
can
be
picked
up
and
edited
by
someone
and
turned
into
an
issue
just
from
a
tracking
perspective.
If
that's
abusive,.
A
Okay,
and
even
if
you
create
things
when
I,
your
issue
is
on
the
board
right
at
the
top
in
implementation
to
do
so,
even
if
you
just
create
the
issue,
I
catch
them
up
with
them.
On
the
board
comment
in
there
to
visualize,
oh
make
the
issue:
five
words:
there
we
go.
I
Yeah
I'm
going
through
my
head
trying
to
figure
out
like
if
we
talked
about
anything
new
new
work
and
I
don't
see
any
new
work
that
we
ended
up
talking
about
today.
A
I
So
yeah
and
the
process
there
it
will
be.
You
know
what
we
talked
about,
which
is
when
ben
ben
has
to
write
the
initial
doc
for
the
the
http
directs,
and
then
we
we
add
a
document
and
a
markdown
file,
but
it's
not
new
work,
but
that's
work.
That
ben
has
was
planning
to
do.
As
part
of
you
know
the
previous
prs
that
he
filed.
I
C
Yeah,
I
want
to
clarify
that
we
do.
We
do
have
some
documentation
about
how
this
works
in
the
just
the
readme
for
the
proxy,
but
I
want
to
break
out
a
more
detailed
one
now
that
it's
a
little
bit
more
complex,
okay,.
I
Anything
else,
if
there
isn't
nothing
else
here,
then
I
was
going
to
ask
you
or
know
about
you
know
the
cloudflare
stuff.
How
far
are
we
and
were
you
able
to
figure
out?
I
think
we
were
talking
about
something
you
were
going
to
ask
tim
hawkin
or
something.
B
I
Know
send
a
reminder
and
cc
me
please
and
if
he
doesn't
respond,
I
will
like
dm
him
on
twitter.
B
B
So
I
have
a
question
for
it's,
not
a
technical
question.
Basically,
it's
a
question
about
privacy.
I
open
an
issue
about
that,
so
I
don't
know
how
to
answer
that
question.
I
talked
to
josh
and
bob
about
this.
They
told
me
it's
fine,
okay,
let
me
refresh
this
so
basically,
jim
sent
an
email
about
testing
registered
case
satire
with
cube
adm.
B
B
I
And
where
you
can
assign
this,
I
think
you
already
assigned
this
to
me.
Let
me
look
at
some
things
and
ask
a
few
people.
Okay,
so
yeah
leave
the
leave
this
to
me
and
I
will
try
to
follow
it
up
or
not.
Okay,.
C
I
I
just
want
to
mention,
I
think,
for
the
type
of
thing
we're
doing
the
the
one
you
mentioned
for
the
goling
package.
Proxy
is
a
really
good
example
of
what
what
would
be
nice
to
have.
But,
of
course,
we
have
to
talk
to
I'm
sure,
cncf
and
folks.
I
Yeah,
I
have
a
couple
of
leads.
I
will
chase
this.
You
know
early
next
week
would
be
some
sometime.
I
I
can
dedicate
to
this.
I
know
I'm
sorry.
I
didn't
really
listen.
You
saying
no
worries,
so
I
will
start
work
on
this
next
week.
Okay,
yeah.
B
A
Something
that's
also
on
the
board
is
about
other
than
securities
prior
privacy
security.
What's
his
name,
bannister
said
he
is
happy
to
help
and
review.
So
once
we
get
to
security
and
and
jay
is
also
looking
at
security
as
they
go
through,
so
it
kind
of
did
reach
out,
but
I
don't
think
then
got
back
to
us.
Yet
you
just
review
where
we
are
at
the
moment
on
security.
I
Yeah,
I
think,
having
a
separate
call
to
walk.
Sf
tim
through
whatever
we
have
right
now
would
be
a
good
idea
to
yeah
yeah.
A
A
A
I
I
G
And
it's
all
good
we're
all.
I
love
that
we
can
innovate.
We
are
the
most
innovative
place
with
community
things,
that's
what
I'm
excited
about
and
as
we
learn
to
do
this,
making
writing
it
down
and
making
examples
for
others
to
follow.