►
From YouTube: WG Data Protection Bi-Weekly Meeting 20200729
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
A
So
today
we
are
going
to
talk
about
validation
hook
for
snapshots,
so
sean
chen
andy
will
be
talking
about
that
and
then
we
will
go
over
the
data
protection
workflow.
There
are
a
few
people
reach
out
to
us
saying
that
they
are
interested
in
helping
out
with
that
white
paper.
So
I
will
just
to
go
over
that
all
right.
Let's
get
started
shenzhen
think
I
have
given
you
control.
B
Actually,
if
you
don't
mind,
can
we
start
with
that
production?
Workflows,
andy
was
stuck
in
another
meeting.
He
will
be
late
for
a
couple
minutes.
I'm
sorry!
Oh.
A
D
A
D
B
You
give
me
yourself,
but
where
would
brief
introduction
first.
D
D
Oh
great,
so
I'll
start
with
the
motivation.
D
D
D
Okay,
yeah
yeah,
so
the
biggest
culprit
is
actually
the
one
of
semantics.
These
are
already
enforced,
partly
by
the
controller,
but
it
doesn't
block
the
actual
creation
and
update.
D
D
A
D
Sure
sure
that
that
is
a
good
idea,
let's
see
so
in
the
volume
snapshot
object,
we
have
the
spec
dot
source
field
and
the
spec
dot
source
field
has
a
persistent
volume
claim
name
field
as
well
as
a
volume
snapshot
content
name
now.
The
behavior
of
the
snapshot
is
that
when
you
set
the
persistent
volume
claim
name,
you
are
asking
the
snapshot
to
take
a
dynamic
snapshot
to
create
and
take
a
snapshot
for
you.
E
D
D
If
the
default
snapshot
class
is
not
set,
then
or
if
there's
more
than
one
default
class,
then
the
hook
will
allow
the
creation,
but
the
snapshot
will
fail
because
there's
nothing
the
hook
can
do
but
it.
So,
if
that's
the
case,
if
it's
unset
or
if
it's
set,
it
has
to
be
not
empty
so
that
it
has
to
be
a
non-entity
non-empty
string.
D
But
I
think
this
is
a
less
important
validation.
D
One
other
validation
I
want
to
talk
about
is
the
spec
dot
following
snapshot
ref,
so
this
behavior
can
be
slightly
confusing
because
in
the
volume
snapshot
content
in
the
dynamic
case,
all
of
these
fields
will
be
set
by
the
controller
when
is
created
pointing
so
the
the
fields
that
are
that
are
listed
under
volume
snapshot.
Ref
are
the
name,
the
namespace
and
also
a
uid
field.
D
When
it's
created
dynamically,
all
three
of
these
will
be
set
automatically,
but
when
you
are
in
the
pre-provision
case,
the
cluster
admin
has
to
create
the
name
and
the
name
space
field,
but
they
have
no
way
of
knowing
what
the
uid
of
the
volume
snapshot
that
is
going
to
be
bound
is
going
to
be
so
they
cannot
set
it.
So
it
has
to
be
set
by
the
controller
after
it's
bound
and
we
want
to
add
an
immutable.
D
D
D
A
Yeah,
so
the
one
snapshot
beta
design
is
different.
Now
yeah
it
will
prevent
rebinding.
F
F
A
I
think
we
did.
We
talked
about
something
similar
earlier
on
this.
Basically,
so
you
have
a
content
right
that,
if
you
let's
say
you,
you
have
a
one
snapshot
and
content
they're
bound,
and
then
you
delete
the
snapshot
content
still
there,
so
in
this
case
you're
actually
supposed
to
create
another
content.
If
you
want
to
create
another
snapshot,
use
it.
So,
oh.
F
H
A
You
but
you
cross
another,
you
can
well,
you
can
create
another
well
as
long
as
you
have
the
physical
snapshot
id
there,
then
you
can
create
a
new
content
right,
so
I
mean.
E
A
I
I
think
every
approach
has
pros
and
cons
right.
I
don't
think
there's
anything
that
is
perfect,
so
I
mean
I
think
this
is.
I
think
this
is
just
to
like
avoid
a
ambiguity,
so
this
you,
you
know
you
just
always
have
this
one-to-one
relationship
between
a
snapshot
and
content
right.
So
I
think
that
the
case
that
you're
talking
about
shoes
to
your
work-
maybe
just
like
one
extra
step-
you
do-
have
to
create
a
new
content
in
this.
A
For
the
for
your
for
your
new
name,
spaced,
one
snapshot
right.
You
can't.
A
H
I'm
a
bit
uneasy
with
with
changing
the
semantics
relative
to
pb's
pvcs.
A
I
think
the
reason
that
we
started
to
do
this
is
there
is
a
bug
in
ppc
controller.
Let's
see
if
I
can
remember
exactly
the
case,
I
think
I
think
some
there
are
some
users
running
to
the
problem
when
they
copy
the
original
yaml
file
and
trying
to
deploy
it
in
a
new
environment.
They
want
to
keep
the
origin
the
volume
right
on
the
that
is
now
their
storage
system,
but
because
the
volume,
because
if
the
pvc
does
not
exist,
the
pvp
control
actually
will
automatically
delete
the
physical
volume.
A
A
So
so
we
have
some
extra.
You
know
checks
in
our
controller
to
make
sure
you
know
if
it's
just
because
when
in
snapshot
does
not
exist,
we
don't
go
automatically
delete
the
underlying
bonus
snapshot.
So
that's
that's
a
very
different
from
the
pvpc
approach.
A
A
B
D
B
Change
deeper:
what
happened
for
the
part
right,
so
what
happened?
Was
a
user
has
a
a
problematic
cluster,
so
what
the
cluster
has
been
want
to
do
is
to
do
a
coupe
cutter,
get
for
all
the
pvs
and
dump
this
information
into
a
big
yaml
file
and
remove
the
status
and
just
keep
cuddle
apply
in
the
target
cluster.
In.
B
That
that
they
want
to
use
the
original
pvs,
so
the
problem
with
that
is
that
those
pvs
has
a
deletion
policy
set
to
delete
a
return
policy
set
to
delete.
So
the
moment
you
create
those
pv
objects
in
the
target.
Cluster
is
the
moment
where
the
controller
starts
to
loop
and
find
oh
this.
This
pv
is
actually
lost
funding
right,
so
they
go
ahead
and
delete
the
underlying
disk
of
it.
B
The
snapshot
will
experience
exactly
the
same
behavior
if,
when
the
modern
natural
content
is
not
bounded
to
anything,
I
usually
follow
the
pvc
pattern,
so
we
explicitly
block
the
use
case
of
upgrading
to
updating
the
ui
d2
empty,
to
allow
rebinding
for
this
natural
case
and
the
other
consideration.
I
think
then
remember
the
rom
is
the
semantic
for
allowing
multiple
one
snapshot.
Content
going
into
pointing
to
the
same
underlying
one
snapshot
object
without
understanding
it
by
simply
just
copy.
The
content
object
over
so
this
this.
B
Basically
there
are
a
couple
of
decisions
we
I
think
back.
Then
we
had
a
couple
of
rounds
of
discussion
with
the
storage,
whether
or
not
we
should
follow
exactly
the
same
pattern
as
dvdc,
but
that
was
considered.
Not
a
safety
is
the
way
and
use
the
new
approach
where
allow
at
least
the
law
of
rebinding
of
the
worms,
natural
content
seems
like
a
safer
approach,
so
the
end
of
the
user
can
still
achieve
doing
that.
H
Okay
answer
your
question,
john
yeah,
to
a
certain
extent,
I'm
I
guess
I
I
would
argue
that
the
at
least
the
pvpc
thing
isn't
a
bug,
it's
a
user
misunderstanding,
but
that's
that's
not
really!
That's
not
really
part
of
the
discussion
here.
I
guess
my
my
concern
is
also
goes
along
with
with
what
ben
raised
earlier
with
regards
to
things.
H
You
know
where
rebinding
is
handy
and
doing
things
like
name
space
transfer,
and
I
guess,
are
we
sure
that
we're
not
painting
ourselves
into
a
corner
where
we're
gonna
have
trouble
implementing
that
in
a
in
a
way
that
we
can
survive
failures
or
crashes
of
the
controller?
That's
doing
that
rebinding
such
that
it
can.
You
know
pick
up
where
it
left
off
relatively
easily.
H
Right
because,
with
the
rebinding
there's
kind
of
a
nice
set
of
of
steps
that
you
can
apply
to
do
that,
transfer
and
you
can
kind
of
look
at
the
state
of
the
objects
and
figure
out
where
you
were
in
the
process
and
the
and
it
kind
of
appears
to
happen
atomically.
H
But
I'm
a
little
concerned
if
we're
having
to
create
new
content
objects,
we're
going
to
run
into
a
situation
where
we
either
have
two
content.
Objects
pointed
at
the
same
snapshot
on
the
storage
system
or
we're
going
to
have
a
time
period
in
there.
Where
there's
no
content.
Object,
pointed
at
the
snapshot
on
the
storage
system
and
both
of
those,
I
think,
can
get
us
into
situations
where
we
lose
track
of
of
data.
I
B
Definitely
something
we
can.
We
can
discuss
more
john
in
a
separate
conversation
and
the
pi
is
there
out
for
review.
So
if
you
do
have
this
kind
of
comments,
please
feel
free
to
add
your
comments.
That's
one
and
the
other
one.
I
want
to
emphasize
a
little
bit
over
here.
It's
allow
refining
of
the
content.
I
shouldn't
mention
this
before,
without
allowing
bounding
the
one
central
content
brings
confusion
right,
so
one
snapshot
may
be
very
clear
between
statically
provisioned,
one
snapshot
and
one
sample
content
and
dynamically
provisioned.
B
F
Two
content
objects
pointed
to
the
same
same
snapshot,
is
not
something
to
worry
about,
because
there
are
plenty
of
other
use
cases
where
you
will
have
lots
of
content
objects
pointing
to
the
same
snapshot.
For
example,
if
you
want
multiple
users
in
multiple
namespaces
to
have
a
reference
to
that
snapshot,
you
unavoidably
will
have
duplicate
content
objects.
F
B
You're
right,
I
was
not
talking
about
multiple
reference
to
the
underlying
worms,
natural
one
of
them
snapshot
in
this
storage
system.
I
was
more
talking
about
one
snapshot:
content
if
we
allow
it
by
rebinding
can
potentially
point
to
a
mismatched
volume
snapshot,
for
example
a
dynamic,
and
that
makes
the
semantic
around
the
finding
not
clear.
D
Yeah
have,
please
can
happen
in
a
pr
in
the
comment
that
way
we
can
flesh
out
all
the
details.
B
That's
right,
the
only
the
reason.
One
of
the
major
reasons
why
we
want
to
do
this
in
this
working
group
is
that
we
know
a
lot
of
parties
will
be
interested
in
this
feature.
So
if
you
do
have
comments,
please
feel
free
to
comment
on
that
pr,
and
we
can
address
those
comments
in
you
know
a
more
detailed
way.
G
I
have
a
quick
question
well
so
just
to
give
folks
time
to
review
the
pr.
What
is
the
timeline
that
you
are
looking
at?
Considering
andy
is
an
intern
he's
going
to
be
here
probably
for
the
summer,
so
we
probably
have
a
tighter
timeline
than
people
expect.
I
think
so.
It
would
be
clear.
It
would
be
awesome
to
know
what
the
expected
timeline
is.
B
That
that's
a
great
question
thanks
for
bringing
this
up,
which
is
andy,
will
continue
his
industry
for
another
two
weeks.
So
it
is
gonna
be
great.
If
we
can
have
this
cap
reviewed
by
the
community
and
emerged
within
the
two-week
timeframe,
while
he
is
going
to
start
implementing
the
webcook
service
piece
and.
D
B
What's
that
you,
your
internship,
will
end
in
in
a
couple
next
couple
of
weeks,
so
it's
going
to
be
great.
We're
going
to
have
this
pr
review
and
emergency
in
two
weeks
right.
A
Right
so
I
think
there's
another
thing
that
I
want
to
bring
up
is.
I
think
I
think
I
know
that
andy.
You
will
be
talking
about
that
a
little
bit
later,
so
this
is
also
going
to
impact
our
ga
timeline
right.
So
I
think
maybe
you
can
just
you
know,
go
ahead
and
talk
about
this,
and
then
I
that's.
Why
I'd
like
this
group
to
be
aware
of
this
because
of
the
introduction
of
this
one
may
affect
our
ga
timeline.
D
D
And
so
how
do
we
solve
that?
Well,
we
will
need
two
phases,
so
the
first
phase
begins
with
a
what's
called
a
ratcheting
validation
using
the
web
hook.
Only
and
the
specifics
of
the
ratcheting
web
validation
is
that
it's
strict
on
create,
so
we
prevent
new
invalid
objects
and
it's
strict
on
updates,
where
the
existing
object
passes,
strict
validation,
so
we
prevent
existing
objects
from
becoming
invalid
and
we
are
relaxed
on
updates
where
the
existing
object
fails,
strict
validation.
D
And
we
combine
this
with
some
process
to
ensure
that
the
invalid
data
in
the
cluster
is
completely
eliminated,
fixed
or
removed,
and
this
is
the
part
that
I
want
to
discuss.
What
exactly?
What
method
do
we
want
to
use?
Do
we
want
to
so
I
see
three
options
and
some
may
be
better
than
others,
so
we
can
auto
delete
invalid
objects.
D
This
might
pose
data
loss
issues.
We
can
try
to
automatically
fix
the
invalid
objects,
but
it's
impossible
to
know
which
field
should
be
added
or
removed.
If
we
have
both
source
field
set,
we
don't
know
which
one
should
be
set
or
what
should
be
added
or
removed,
or
should
we
somehow
highlight
every
invalid
object
for
the
user
and
then
make
it
their
responsibility
to
fix
it
before
they
upgrade
to
phase
two?
B
The
the
problem
is
that
if
we
want
to
tighten
the
restriction
in
the
workbook
or
in
the
schema,
the
existing
invite
one
snapshot,
content
and
one
snapshot
objects
in
the
api
server.
We
will,
we
are
not.
We
will
not
be
able
to
delete
them.
The
reason
why
is
the
workbook
is
going
to
reject
the
delete
request
or
update
request
that
sends
to
the
api
server
the
api
server.
B
D
Me
yes,
but
the
the
process
that
I
want
to
talk
about
is
what
exactly
is
the
part
where
we
fix
the
clusters
and
how?
What
is
the
best
way
to
go
ahead
and
start
reducing
the
number
of
invalid
objects
to
zero.
B
So
andy-
and
I
we
had
a
lot
of
discussions
around
this
so
to
me-
automatically
deleting
objects
for
users
regardless
whether
they
are
embedded
or
valid
or
modifying
existing
crs
in
the
api
server
sounds
like
a
no-go
to
me,
because
that
may
end
up
risking
their
loss,
especially
for
the
deletion
case.
B
Even
though
the
controllers
forbids
the
invalid
objects
to
be
found.
But
if
we
issue
a
delete
command
to
the
for
the
invited
objects,
the
user
maybe
get
confused
because
wherever
the
object
over
there
is
lost.
B
So
another
more
moderate
kind
of
you
know
approach
is
to
insert
some
announcement
into
the
community
saying
that
hey
we'll
introduce
a
stretching
network
for
now,
and
it's
going
to
prevent
you
from
creating
new
united
one
snapshot
and
one
central
content.
However,
it
cannot
deal
with
existing
united
once
so.
B
We
ask
the
users
explicitly
to
delete
those
objects
before
they
upgrade
to
a
more
restricted
validation
around
those
crs.
So
the
this
I
I
personally
prefer
the
lateral
one
I
want
to
share
with
this
idea
of
the
community
and
just
see
whether
you
guys
have
a
better
idea,
a
better
approach
of
doing
that.
F
How
about
like
spamming,
the
log
with
warnings,
if,
if
there's
invalid
objects,
that
haven't
been
deleted
yet
to
try
to
get
someone's
attention.
F
B
F
I
think
there's
always
going
to
be
some
idiot
user,
that
just
ignores
all
the
messages
and
all
the
emails
and
all
everything,
though,
and
at
some
point
you
have
to
get
their
attention
by
completely
failing
to
do
something
that
they
that
they
want
to
do
like,
like
you
mentioned,
trying
to
upgrade
to
some
future
version.
J
A
B
J
K
J
B
Sure
sure
I
agree,
you
know
there's
no
easy
way
to
prevent
all
the
age
cases.
We
just
try
to
be
more
cautious
over
here
in
terms
of
a
community,
maintain
the
controller
go
ahead
and
do
something
automatically.
J
J
J
A
Yeah
so
now
yeah.
I
guess
that
definitely
is
safer.
A
So
let's
say:
if
the
user
just
ignored
all
the
warnings,
they
still,
you
know
they
upgraded
to
the
later
version.
You
know
they
can
no
longer
delete
those
the
bad
records
objects
so
and
then
what
can
they?
Basically,
they
can
still
create
a
new
content
as
long
as
they
know
the
snapshot
id,
but
they
can
use.
They
can
still
use
that
for
another
for
new
new
content
right
I
mean
as
long
as
they
know,
because
we
are
not
restricting
a
snapshot
id
to
be
used
by
only
one
content
right,
so
they
can.
A
Let's
say
if
they
have
something
there.
You
know
pointing
to
the
snapshot
that
is
on
the
storage
system.
They
can
still
go
to
the
new
system.
They
can
create
a
new
content
point
into
that
thing
and
then
create
a
new
following
snapshot
about
to
bind
oh
yeah.
So
then
they
will
have
a
bad
record
and
a
good
record.
A
F
G
F
That
you
can
arbitrarily
update
them
with
as
long
as
they're
invalid.
To
begin
with,
you
can
update
them
with
new
and
valid
contents
and
we'll
never
force
it
to
become
valid
and,
and
the
goal
is
at
some
point
to
make
sure
that
there's
no
invalid
ones
left.
But
if
we
don't
automatically
do
it,
a
user
is
going
to
have
to
do
it
and
we're
going
to
have
to
make
it
painful
for
the
people
that
don't
do
it.
F
J
Mean
you
act
like
users
are
going
to
make
these
things
by
themselves,
they're
going
to
be
getting
software
from
somebody
else.
That
did
these
things
there's
going
to
be
incorrect.
Implementations
and
you
know
people
are
gonna.
You
know
they're
gonna
have
to
go
back
to
the
vendor,
but
I
mean
we
say
here.
You
know
that
we
have
this
rat
that
we're
able,
in
this
one
release
to
be
relaxed
on
some
updates,
where
the
existing
object
fails
for
validation.
B
Yes,
because
once
the
validation
is
in
is
not
possible,
you
can
even
create
those
new
added
records.
Then,
what's
the
point
of
dealing
with,
therefore,.
J
D
J
J
F
J
J
B
J
B
F
A
Not
sure
I
don't.
A
A
F
A
If
we
forever
stay
in
the
first
phase,
then
that
means
we
can
never
ga
right.
The
is
that
the
the
criteria
we
can't.
A
B
Okay,
at
this
stage,
I
think
I
I
hear
what
they
were
saying,
but
in
general
crd
level,
validation
is
more
preferred
compared
to
web
group
ones.
The
web
book
will
give
you
some
extra
latency
whenever
you
create
a
request
or
delete
or
update.
The
request
goes
to
the
web
hook.
First,
before
the
app
api
still
takes
any
action
to
persist
the
thing,
so
that
is
a
wrong
trick,
and
that
is
actually
a
service
call
one
snapshot
itself.
B
It's
a
very
latency
kind
of
you
know
a
sensitive
feature,
so
we
are
concerned
of
having
the
webhook
in
place
for
every
single
create
for
updates.
You
know
if
we
want
to
enforce
the
immutability
of
certain
fields.
We
cannot
save
that,
but
for
create
and
delete.
If
we
can
save
the
round
trip,
that'd
be
great.
J
Know
what
what's
the
actual
impact,
so
if
so,
there's
existing
users
who've
been
using
these
snapshots
and
may
have
some
bad
crs
floating
around.
I
don't
my
understanding
is
there's
not
that
many
people
actually
using
this
feature
at
this
point,
is
that
everybody
else's
understanding
or
is?
Am
I
missing
something
there.
A
So
we
need
to
get
that
from
so
there
are
storage
vendors
and
there
are
backup
vendors
right.
So
I
think
storage
vendors
there
are,
I
think,
most
drivers.
Now
I
I
use
a
lot
of
drivers
now.
I
have
implemented
to
see
the
snapshots
and
then
but
for
background.
E
A
Know
how
many
well.
A
J
I
L
Yeah,
I'm
just
I'm
thinking,
I
think
so.
L
A
Okay,
so
are
there
any
what
about
other
like
backup
vendors?
Are
you
using
css
snapshots?
What's
the
status.
J
But
but
they're
only
I'm
not
sure.
L
So
we
we
don't
allow
casting,
does
not
allow
upgrades
between
alpha
and
beta
snapshots,
but
we
do
support
both.
But
you
can't
you
can,
for
example,
kind
of
mix
and
match
right.
A
B
A
Production
ready
for
any
vendor,
yet
maybe
there
are
I
just
I
I
don't
have
that
data,
but
I
know
that
at
least
from
the
csi
driver
point
of
view
many
driver
has
implemented
this
feature.
A
So
so,
if
they
you
know,
if
the
user
created
those
snapshots,
I.
A
D
Okay,
I
did
have
one
one
more
thing
I
wanted
to
talk
about,
which
was
the
certificate
management
of
the
webhook
server.
But
if
we
are
running
out
of
time,
we
can
talk
about
that
in
the
pr.
A
Okay,
yeah,
so
probably
we
oh,
I
think
I
I
I
want
to
know
so
chanching.
Can
you
hear
us
now?
I
we
can't
hear
you
he
said
he's
talking
about
it.
You
can't
hear
chanting.
B
Yeah
I
was
talking
about
like.
A
A
A
I
was
I
was,
I
guess
I
was
wondering
I
think
so
I
think
the.
What
is
the
can
you
go
to
the?
What
like
the
graduation
criteria,
the
like
the
timeline.
The
plan
is
what
for
for
the
first
release
and
basically
meaning
that
in
one
dot
that
means
in
when
that?
A
Okay,
so
so
what
we'll
see?
If
we
can
get
this
one
down,
you
went
down
19,
but
otherwise
it's
going
to
be
1.20,
and
then
we
need
to
wait
for
at
least
one
one
release,
or
it
could
be
even
more
before
we
go
ga
have
you
guys
got
any
clarification
on
that
part
like
how
many
releases
should
you
have
to
wait?
Ashanti
and
indy.
D
It
looks
like
xiangtian
is
having
some
problems
with
his
web
client.
A
Oh
okay,
did
you
get
any?
I
I
didn't
read
your
cap
lately,
I'm
not
sure
if
there
was
any
update
on
that
like
how
many
releases
are
we
talking
about.
D
A
A
So
maybe
just
right
around
the
time
that
you
are
you're
leaving,
we
can
get
us
down
I'll,
be
that'll,
be
good.
D
A
All
right,
yeah
so
chanting.
Sorry,
we
can't
hear
you,
I
don't
know
if
you
can
still
hear
us,
hey
thanks
andy,
so
we
can
continue.
You
know
the
other,
the
reviews
on
your
pr.
So
even
if,
if
anyone
is
interested
in
this,
please
add
your
comments
there.
A
I
think
yes,
I
think
so,
so
let
me
just
go
back
to
this.
I
think
I
want
to
go
back
to
this.
This
database
more
flow.
I
know
we
just
went
through
this
one
briefly
last
time,
so
so
we
want
to
work
on
this
as
a
group,
but
there
are
multiple
topics
in
here
right,
so
we
we,
there
are
people
who
ping
shan
chandras
offline
and
saying
that
they're
interested
in
one
candidate,
so
we
I
know
I
have
this.
This
one
is
actually
a
page.
A
I
just
put
down
some
names
here
that
you
know
the
sections
that
you
may
be
interested
in
helping
out.
So
so
the
outline
you
know
the
initial
basically
just
like.
Why
do
we
want?
Why
do
we
want
to
do
their
progression
and
use
case?
So
we
have
so
we
put
a
few
names
here
just
to
to
you
know,
write
write
up
some
use
cases
for
this.
A
What
is
currently
available
so
that
one
should
be
straightforward.
Then
what
did?
What
are
the
missing
building
blocks?
So
I
think
right
now
there
are
multiple
sections
here
so
so,
like
the
first
one
is
rolling
backups
we
have
so
I'm
sorry.
How
do
you
pronounce
it
film
from
darian
see
wonky
phone
phone?
A
G
A
Okay,
alexey
and
elliot
yeah,
so
they
are
interested
in
helping
out
with
this
backup
section.
So,
basically,
this
this
big
section,
we
want
to
define
the
problems
just
to
you
know,
talk
about
why
we
need
to
do
why
we
need
the
volume
backups.
While
we
are
saying,
okay,
those
are
still
missing.
A
What
is
the
problem?
What
what
we
want
to
do
there
right?
So
that's
what
we
want
to
put
together
for
for
each
of
those
topics
right
so
the
phone
box,
and
next
one
is
a
backup
repository.
A
So
this
could
be
the
this
could
be
the
object
store
or
it
could
be
other
backup,
vendors
backup
device
yeah.
So
I
have
phone's
name
here,
but
I
think
this
one
actually
for
the
for
the
cozy
project.
Actually,
this
would
be
also
appropriate.
So
if
you
notice
anyone
working
on
cozy
want
to
work
on
this
section,
that
would
be
great
too
and
the
other
populator
so
ben.
I
added
your
name.
Yes,
you
are
working
on
it
anyway.
Right
is
that
okay.
A
Tom
here
you
know
what
in
cupro
just
me
and
then
the
yeah,
so
I
think
if
we
can
get
started
and
then
you
know
there
are
so
some
of
the
later
other
sections.
We
can
work
on
a
little
bit
later,
but
you
know
if
we
can
get
started
with
some
of
the
sections
like
missing
blocks.
A
M
A
Yeah
you
can
actually,
so
let
me
share
this
one.
You
can,
if
you
even
add
your
name,
if
you
want
to
let
me.
A
Sure
I
need
to
be
a
editor.
C
A
Yeah
so
perfect.
A
A
Can
you
hear
me,
yes,
go
ahead.
B
Okay,
do
you
want
to
also
let
everybody
know
what
is
the
expectation
of
putting
the
name
over
there.
A
I
think
I
yeah
so
okay,
so
let
me
just
just
repeat
that
and
then
you
can
add
something
so
yeah.
So
basically,
if
you
put
your
name
there,
I
mean
that
means
that
you
are
owning
that
section
right.
So
people
who,
who
are
names
are
here.
I
expect
you
guys
to
communicate
with
each
other
and
work
work
on
this
together
come
up
with
something
so
for
the
first
phase.
A
This
is
we're
talking
about
a
white
paper
right,
so
talk
basically
define
problems
and
talk
about
what
other
possible
solutions
and
then
I
think,
the
second
phase.
We
will
also
work
on
the
caps,
but
I
think
I
think
the
first
phase-
it's
not
going
to
be
caps,
because
I
don't
want
to
tie
the
white
paper
completely
with
with
caps,
but
I
think
we
we
first
need
to
define
what
are
the
problems
that
we're
trying
to
solve.
A
A
A
B
Yeah,
I'm
sorry.
I
have
no
problem
with
the
web
client
so
having
your
name
over,
there
simply
means
that
you
will
be
putting
effort
trying
to
define
the
problem
in
a
specific
area,
and
what
does
that
mean
in
data
protection
domain?
B
And
what
are
the
proposals
right,
yeah,
the
then
the
nation
exteriors
as
shin
was
describing,
if
there's
a
gap
in
in
the
current
kubernetes
environment
and
what
are
proposed
native
kubernetes
constructs
to
introduce
into
the
community
and
the
end
of
the
day.
This
will
be.
Some
kind
of
you
know,
guidance
how
we're
going
to
plan
our
projects
in
this
working
group.
B
So
it's
I
would
say
it's
really
really
heavy
in
terms
of
work,
so
feel
free
to
to
contribute
in
any
of
those
domains.
If
you
want
it.
A
Yeah,
okay,
so
so
does
everyone
know
that
we
we
do
have
a
slack,
so
we
can
communicate
on
slack
and
also
mailing
list.
A
It's
probably
helpful
for
people
who
has
name
there
to
put
their
slack
slack
id
and
the
email
address
so
that
if
people
want
to
reach
out
to
you,
it's
easier
to
find
you.
Okay,
all
right.
A
Any
any
questions
does
anyone
has
any
questions.
I
think
they're
all.
I
think
that
there's
someone
also
reached
out
to
shanghai
yesterday,
so
we
haven't
got
a
chance
to
to
to
talk
about
that
yet
to
to
respond
yeah,
I
think
there's
he
says
stephen.
A
Case,
okay,
sure
yeah,
that's
what
I
thought
too
so
yeah.
That
sounds
good.
So,
oh
sorry
can
you
is
that
right,
stephen
yeah!
You
got
it
right,
though.
Well
done.
Okay,
all
right!
Okay,.
A
Okay,
all
right,
I
think
I
think
that's
all
we
have
today
anything
else,
if
not
that's
it.
For
today,.
A
So
yeah,
so
you
can.
We
can
communicate
with
each
other
on
slack
and
also
email.
So
that's
why
I'm
asking
everyone.
If
you
see
your
name
here,
please
add
your
email
address
and
also
your
slack
handle.
Does
everyone
know
that?
Does
everyone
know
that
we
have
this?
We
have
the
slack.
I
don't
know
if
everyone's
not.
I
A
A
So
I
will
share
that
email
address,
yeah
email,
so
I
think
everybody
should
have
email
address,
let's
start
with
email
and
then,
if
then,
after
afterwards,
you
know
you
can
decide
to,
you
know,
communicate
with
slack
or
not.
But
if
you
go
to
the
this
this
page,
this
is
our
agenda
page.
This
has
all
the
information
how
to
join
mailing
lists,
how
to
join
the
slack
channel.
So
that's
our
slab
channel,
yeah.