►
Description
Kubernetes Data Protection WG Bi-Weekly Meeting - 09 March 2022
Meeting Notes/Agenda: -
Find out more about the WG here: https://github.com/kubernetes/community/tree/master/wg-data-protection
Moderator: Xiangqian Yu (Google)
B
Hello,
everyone
today
is
march
9th
wednesday.
B
Today
this
is
the
kubernetes
data
protection
working
group
meeting.
Today
we
have
quite
a
bit
of
agenda
items,
so
we
will
go
through.
Huang
hwang
will
go
through
the
current
status
of
the
cbt
work,
change
block
chair
can
work
and
then
williams
snapshot
is
moving
to
ga.
It
has
no
gga,
but
there
are
some
remaining
items
needs
to
be
done,
including,
but
not
limited,
to
remove
support
of
the
way
one
beta
one
apis.
B
Many
of
many
of
you
might
be
well
aware
of
the
there's,
a
buckle,
incompatible
change
between
we
won
and
we
won
better
one.
So
this
is
going
to
be
a
relatively
long
process
in
terms
of
retiring.
We
want
beta1
apis
and
then
going
moving
forward.
We
also
have
this
xiao
and
how
it
has
some
put.
Some
hurry.
Put
some
cozy
updates
over
there
as
well.
So,
let's
start
with
the
quan
juan
feel
free
to
take
over.
D
Sure
sure,
after
can
you
see
my
screen
now?
Yes
yeah,
so
basically
I
just
want
to
give
an
update
on
what
we
have
discussed
so
far
on
the
cbt
feature
and
some
of
the
work
that
I
done
a
little
bit
independently
on
this
topic.
So
far.
So
so
far
we
have
an
agreement
on
the
the
api,
and
this
is
the.
D
That
we
have
discussed
and
agree-
maybe
maybe
a
year
back,
I
when
I
do
some
more
research.
I
also
and
discuss
with
some
of
the
engineer.
I
also
adding
this
a
little
bit.
Did
you
feel
that
I
think
might
be
necessary
when
we
go
into
the
design?
I
will
explain
a
little
bit
more
about
that
later,
but
basically
the
to
it
in
order
to
access
some
of
the
resource
we
might
have
to
give
it
some
secret.
D
D
I
haven't
asked
the
whole
team
yet,
but
the
the
rest
of
the
api
is
what
we
have
agreed
like
last
year
and
I
last
year
when
we,
when
we
bring
this
to
desired
phase,
we
bump
into
a
problem
that
we
think
that
it
is
better
to
have
like
a
proxy
that
directing
these
requests
into
a
specific,
the
specifics
provider
of
the
differential
snapshot
service.
D
So
I
create
some
kind
of
a
slide
here
so
that
to
explain
what's
going
on
before
that,
I
just
can
quickly
go
over,
like
a
very
brief,
like
once
uses
case,
to
make
sure
everybody
at
the
same
pace.
So
this
is
like
a
backup
workflow
with
different
snapshots,
and
this
is
like
a
normal
typical
using
of
an
application
or
an
application
running
on
a
pod.
That's
using
some
storage,
as
you
can
see,
pvc
pv
and
it
mapping
to
storage
on
the
back
end.
D
In
this
scenario,
we
only
look
at
the
block
volume
only,
but
I
think
we
agree
that
after
we
got
the
block
volume
going
on,
we
will
start
adding
the
five
system
into
other
the
fight
system
into
it
and
in
order
to
make
a
backup
typically
we
come.
We
start
with
like
a
snapshot
of
the
balloon
and
and
then
depend
on
the
vendor
right.
They
might
mount
it
the
pvc
creating
from
that
snapshot
or
some
vendor.
They
just
backed
up
the
snapshot
directly.
It's,
but
it
is
like
typical
workflow.
D
So
without
different
source
snapshot
service,
then
the
backup
vendor
might
be
forced
to
back
up
the
entire
snapshot.
With
the
differential
snapshot
service,
we
can
request
for
the
specifics
block
that
change
or
specific
file
that
chain
depend
on
the
mode,
but
in
this
case
we
only
focus
on
the
block.
So
the
difference,
your
snapshot
will
return
like
a
block
the
chain
block
list,
and
then
the
backup
mover
can
simply
backup
only
the
block
that
chain
this
result
into
a
more
efficient
backup.
D
E
D
E
D
Like
I
said
it,
just
it's
just
one
scenario
I
mean
many
vendors
might
even
directly
back
up
from
the
snapshot.
D
Like
I
said
it
is.
Is
it
a
vendor
specific
thing,
the
the
one
I
show
you
it
just
it's
just
one
way
to
do
that
and
it's
up
to
the
vendor
to
backing
up
it
to
back
up
the
snapshot,
the
how
how
they
back
up
the
snapshot
in
this
case.
In
this
scenario,
is
we
create
a
snapshot
from
it,
and
then
we
create
a
pvc
from
the
snapshot
with
us
with
the
source
specified
the
steps
go,
loom,
snapshot
name,
and
then
we
mount
it
on
the
mover
part
and
so
on
and
so
forth.
E
D
E
D
E
The
are
you
able
to
get
blocks
directly
through
the
api.
Is
it
a
data
path
or
is
it
would
you
have
to
actually
read
from
the
volume?
E
D
But
I
I
I
also
have
another
scenario
in
which
we're
using
vmware
and
we
actually
using
via
access
to
vmware
api
to
directly
move
blocks
from
the
via
the
vmware
first
class
list
to
backup
yeah
in
both.
E
And
so
I
guess
that
does
that
mean
that
the
api
itself
will
support
both
modes
like
reading
blocks
directly
and
then
just
referencing
blocks
in
a
volume
that
you
have
to
restore
yourself
or
will
it
take
an
opinion.
D
D
D
Really,
if
we
can,
if
we
can
access
the
back
end
snapshot
directly
from
the
move
apart,
then
we
don't
have
to
you
know:
do
that
all
that
conversion
all.
G
D
Yes,
that's
not
one
way
to
do
that
and
I
think
ben
right
you,
you
actually
did
an
experience
on
that.
One
and
thomas
russian
in
my
team,
also
needed
some
experience
on
that
conversion,
but
that
is
I
I
would
leave
that
into
I
mean
I
would
leave
that
to
each
vendor
mecca
vendor
to
figure
that
out.
G
I
I
will
say
that
I
don't
like
the
idea
of
having
this
api
in
the
data
path
for
actually
getting
the
contents
of
the
blocks,
because
there's
so
many
performance
concerns
around
the
data
path.
It
seems
better
to
focus
on
getting
a
list
of
blocks
and
then
leave
the
retrieval
of
the
contents
of.
C
D
Yeah,
we
did
specify
that
it's
only
the
metadata
of
the
data
update.
It's
basically
just
a
lot
so
change
chain
block
list
without
the
data.
G
D
So
that
one
of
the
debate
that
we
have
back
back
then,
is
whether
to
how
huge
one
is:
can
we
affect
our
desired
decision,
whether
we
zoo
as
a
service?
We
can.
The
client
can
simply
read
it
or
we
put
it
on
a
cr
put
it
on
an
api
kubernetes
api.
D
So
that
is
the
decision
that
we
made
back
then
that,
because
this
list
can
potentially
be
very
long,
so
we
do
not
want
to
put
it
on
an
object
on
a
cr
object
on
on
the
api
server.
E
I
wasn't,
I
actually
thought
we
didn't
this,
make
a
decision.
You
know,
because
both
both
the
like
all
the
cloud
apis
that
have
changed
while
tracking,
like
aws,
they
give
you
a
token
to
go
grab
the
actual
individual
blocks
same
with
vmware.
E
C
I
thought
this
one
should
work.
I
thought
we
talked
about
this
one
before.
D
D
C
D
Oh
yeah,
not
yet
not
yet,
okay
yeah,
so
I
I
did
some.
I
did
some
work
here
to
you
know:
try
to
figure
out
how
to
fit
these
components
together,
and
I
look
into
this
csi
drive
and
how
to
enhance
the
csi
drive
to
support
this
type
of
feature
and
from
what
I
understand,
the
csi
driver
is
plugged
in
into
the
kubernetes
framework,
and
currently
it
have
like
a
few
external
things
like
external
staff
separator
or
external
provisioner.
D
These
are
the
sidecar
of
the
csi
driver
parts
that
running
and
it
communicate
with
the
csi
framework
using
grpc
and
I'm
thinking.
I
think
that
we
can
add
our
different
source
snapshots
as
like
another
sidecar
into
this
package
right,
the
vendor.
Whoever
wants
to
support
this
api
it
can.
They
can
enhance
their
csi
driver
package.
If
you
look
into
the
the
like
a
api.
D
Currently,
we
currently
have
the
csi
driver
dock
right
here
and
and
the
vendor
can
specify
their
capaci
capacity
by
you
know
adding
these
this
capacitive
cover,
basically
in
and
currently
you
don't
have
it,
but
we
can
add
something
like
a
different
substance
service
like
as
one
of
the
capabilities
here,
and
if
the
csi
vendor
has
that
capability,
then
it
has
to
implement
this
guest
lock
and
guest
lock,
get
chain
block
and
get
the
ship
block
response
into
there.
C
So
this
will
be
like
an
independent
another
service,
not
the
controller
service.
D
Is
that,
oh,
the
the
way,
I
think,
is
that
that
the
driver
will
we
wouldn't
have
like
a
proxy
part
that
is
like
a
running
running
in
the
kubernetes
cluster
and
this
proxy
part
will
redirecting
or
forwarding
the
request
from
data
mover
to
the
snapshot
differential
snapshotter
from
the
any
vendor.
That's
supporting
this
feature,
that
is,
that
is
like,
like
a
general
workflow
that
I
can
come
up
with.
D
C
C
Is
basically
not
like
not
another
capability,
adding
on
top
of
a
controller
service?
But
you
are
was
your
new
services?
I
guess
that's
right,
yeah,
okay,
yeah
the
new
service,
because
you-
because
the
reason
is
because
we
don't
want
to
add
too
much
burden
on
the
existing
controller
services,
because
this
is
a
separate.
C
Yeah
yeah,
that
part,
is
fine.
I
think
the
reason
that
you
proposed
a
separate
service
is
because,
because
is
that,
because
we
said
we
don't
want
to
add
this
like
otherwise
we're
part
of
the
controller
right,
we
don't
want
to
add
this
as
part
of
the
controller.
D
What,
as
a
matter
of
fact,
I
do
not
know
that
that
the
the
the
proxy
that
the
you
can
see
the
difference
of
snapshot
proxy
on
the
picture
right
now
on
that
one
is.
I
do
not.
E
Know
curtin
I
haven't
test.
I
haven't.
D
Developed
a
prototype
or
anything
yet
I
just
you
know,
come
up
with
the
idea
that
it
would
work
as
like
a
forwarder.
You
know
it's
just
forward
the
request
to
the
corresponding.
I
do
not
know
how
much
impact
performance
impact
that
it
has
on
the
system
on
on
the
on
the
controller,
because
this
this
proxy
part
had
to
run
somewhere.
You
know
in
like
a
central
place,
so
we
said
so
any
any
any
data
mover
can
talk
to
it
to
request
service.
D
Of
course
it
had
to
have
some
authentication
and
an
authorization
to
to
do
that,
but
some
permission
to
do
that.
But
you
know
it
might
be
a
central
point
of
failure
here
and
we
might
do
some
do
something
to
get
around
that.
But
currently
that
is
like
a
proxy
way
like
a
proxy
to
to
to
so
that
is
allowed.
The
data
mover
don't
have
to
know
the
exact
which
csi
driver
will
serving
this
service.
D
D
So
that
is
what
I
got
so
far,
and
I
still
have
so
many
questions
that
I
need
to
answer,
and
I,
of
course
this
is
a
community
effort.
I
so
I
hope
everyone
can
jump
in
you
know
you
know,
have
your
opinion
can
give
me
some
input
on
what's
going
on
and
what
to
do?
One
of
the
things
that
I
think
zing,
zhang
and
and
helps
send
me
email
a
few
days
ago,
is
that
right
now
we
we
in
our
api,
we
have
he's
using
base,
snapshot
and
snapshot
target.
D
These
are
currently
we
think
of
this
as
a
back-end
snapshot
handler,
but
should
it
be
the
volume
snapshot
instead?
Because
if
you
have
the
volume
snapshot,
we
can
track
to
the
volume
snapshot
content
and
in
that
volume
snapshot
content
there
will
be
a
snapshot
handler
in
there
and
then
this
kind
of
question
then
can.
C
D
But
whether
that
I
like,
like
I
said
we
kind
of
still
discussing
about
and
oh.
D
Yeah,
if,
if
we,
if
we,
if
we
don't,
have
the
information
about
which
csi
drivers
this
volume
supposed
to
go
to,
we
have
to
dig
out
somewhere
right
and
one
of
the
ways
to
do
that
is
we
have
to
including
another.
You
know
parameter
here
somewhere
in
this
request
that
that
you
know
I'm
using
this
csi
driver
right.
D
C
D
I
do
not
want
to
take
too
much
time
off
the
team,
but
here
are
the
few
that
that
kind
of
the
state
of
what
we
have
right
now
and
still
have
many
questions
left.
So
what.
D
D
And
I
could
add
the
driver
name
in
one
of
the
in
one
of
the
the
field
here.
Maybe
I
put
it
in
like
a
parameter
here,
or
maybe
I
add
another
field
called
volume,
class
volume
snapshot,
class
name.
D
D
Field
at
least
yeah,
even
if
that
is
the
case
and
then
yeah
I
can,
I
can
add
another
field
inside
this
drug
and
that
helped
me
directing
directing
this
from
the
proxy
from
the
different
snapshot
proxy.
I
can
help
directing
that
request
to
that
csi
driver.
That's
supporting
that
volume.
B
D
Oh,
I
see
what
you
mean
so
this
the
the
api
that
I
I
we
were
posing
here.
He
actually
the
difference
or
snapshots
are
supporting
that
api,
so
by
the
time
that
the
the
chain
get
blocked
request.
Yet
to
this
api
it's
already
gone,
so
we
have
to
have
like
another
another
service
on
top
of
it,
so
that
yeah.
C
C
C
C
D
So
it's
the
controller
node,
well,
the
the
server
it's
just
a.
I
can
just
say
I
just
maybe
I'm
using
the
wrong
word
here,
but
it
is
like
openness,
service
and
the
part
it
actually.
C
Is
are
you
saying
I?
I
guess
we
just
want
to
know
like
it's
a
is
this
like
the
snapshot,
controller
type
of
a
service
like
running
on
a
controller
node?
Basically,
you
want
to
know
like
well,
where
does
it
go
so
this
is
a
this
is
running
independently.
Basically,
it's
not
really
deployed
together
with
a
driver
right,
so.
D
C
The
question
is:
how
do
we
go
from
there
to
this
next
csr
driver
sidecar,
because
we
are
saying
we
don't
do
we
have
a,
I
thought
we
have
some
kubernetes
crds
defined
or
are
we
saying
we
don't
want
those?
I
guess
that's
the
part
that
I'm
saying.
I
C
I
I
C
So
confused
so
I
know
we
have
some
questions
about
the
maybe
the
the
the
confusion
comes
because
there
was
some
discussion
about
the
kubernetes
crd
and
then
some
there
are
concerns
that
are
at
too
much
load
and
then
that's.
Why
he's
going
this
route
is
trying
to.
I
C
So
I
think
you
have
a
you.
I
see
that
in
your
in
the
other.
Talk
that
you
have
you
have
a
question
saying:
do
we
need
a
crd?
C
I
I
guess
like
normally,
this
would
be
a
like
virtual
api
right
or
in
like
in
core
kubernetes
apis.
This
would
be
a
virtual
api
right,
so
we
wouldn't
save
anything,
but
you
would
ask
for
some
information.
You
would
get
it
back
from
as
like
a
as
like
a
resource,
but
we
don't
have
the
ability
to
do
that
today
because
of
crds
right
and
we're
not
in
the
core
cube.
Is
that
does
that
make
is
that
is.
I
C
I
Yeah,
it's
a
it's
a
thing
that
exists
so
like
subject
access
reviews
are
virtual
resources.
You
don't
save
a
subject.
Access
review.
It
literally
is
a
request.
That's
made
in
the
api
server.
The
api
server
gives
you
a
response
back,
but
that
resource
isn't
saved
in
ncd,
but
you
can
only.
I
Api
server
can
do
this
okay,
another
option,
but
I
don't
think
we
want
to
do
either
of
those
and
so
yeah.
I
think
a
web
hook
is
like
something,
but
still
that
means
that
we're
like
not
saving
something
in
like.
C
I
I
It
would
still
be
protected
by
our
back.
It
would
have
all
of
the
nice
things
that
come
with
kubernetes
resources,
but
it
wouldn't
be
saved
in
fcd
and
we
wouldn't
have
any
of
the
size
problems
or
scale
problems
or
anything
like
that,
but
I
don't
know
how
to
do
that
with
crds.
So
I
I
don't
have
good
answers.
I
I'm
just
trying
to
make
an
analogy
to
like
what
I
I
think
exists
today
for
certain
resources
in
the
core
kubernetes
api
server,
and
if
that
is
something
that
we
could
talk
to
folks
about
and
see,
if
there's
a
way
forward,
using
that,
like
analogy
of
like
what
we're
trying
to
trying
to
get,
does
that?
Does
that
make
sense
to
anybody
else?
Besides
me,
I
apologize
yeah.
C
I
D
Yeah,
I
actually
do
not
know
like
how
to
right,
because
the
typical
way
to
do
anything
in
in
the
csi
is
to
create
a
cr,
and
then
you
know
controller
reacting
on
it,
but
this
one
is
we
don't
want
any
of
that.
We
just
want
a
like
a
rest
api
service.
We
want
to
send
information,
I
mean
get
information
out
of
that.
That's
it
that's
all
we
have
you.
I
B
Yeah,
I
agree
on
api
extension.
This
is
this
is
fairly
complex,
but
just
to
to
make
sure
I
understand,
even
though
there's
a
proxy
which
is
just
a
service,
the
interface
you
expose
to
whoever
calling
you
does
not
necessarily
need
to
be
the
same
as
the
interface
between
you
and
the
csi
driver.
No,
oh,
I
see.
Oh.
D
B
E
B
D
I
actually
that's
a
good
point.
I
I
did
not
think
about
that.
But
yes,
so
the
the
guest
block
chain,
request
coming
from
the
motif
from
data
mover
part,
can
contain
additional
information
to
guide
the
proxy
to
forward
the
request
to
the
csi
corresponding
csi
driver
right.
That's
right!
That's
how
that's
actually
sound
good
yeah.
C
C
So,
like
the
site,
what
is
the
interface
between
this,
your
proxy
part
and
the
differential
snapshotter
like
right
now?
What
we
have
is
sidecars
waiting
for
a
volume
snapshot,
content
api
object.
But
what
is
the
interface
you
are
going
to
have
because
you
are
not
deployed
together?
You
are
not
going
to
make
a
direct
grpc
api
call.
You
can't
do
that
because
you
are
here,
your
your
proxy
party
is
at
a
central
location,
you're
not
deployed
together
right.
So
that
part
I'm
still
not
getting
like.
What
is
your
api?
How
do
you
do
that.
C
Certain
does
not
implement
sidecar
will
be
calling
jfk's
interface
so
like
it's
not
snapshot
or
sidecut.
What
it
does
is.
It
is
waiting
it's
waiting
for
a
volume,
snatcher
content.
Api
object
is
watching
the
volume
snapshot
content
api
object
when
it
gets
it,
it
will
make
a
call
to
the
css
driver
using
the
grp
interface.
So
it's
the
other
way
around.
It's
not
like
yeah
yeah,.
C
The
mechanism
yeah,
but
then
your
side
car.
How
are
you
side
car
going
to
you
like
watching
for
those
differential
snapshot
proxy
part?
It's
required.
I
guess
that's
the
the
scat
changes
blocks
requires.
How
does
that
call
get
passed
to
differential
snapshotter?
I
guess
that's
the
question.
C
What
what
kind
of
interface
is
that
are
you
waiting?
Is
there
some?
I
mean
I'm
just
telling
you
what
what
the
current
excellent
snapshot
aside
can't
do
right,
because
the
basically
communicates
through
api
server.
That's
what
they
do,
but
here
we
don't
have
a
crd.
So
what?
What
is
your
communication
mechanism?
J
J
C
I
So
if
you,
if
you
imagine
that
the
I'll
take
a
stab
at
it,
I
think
I
understand
what
tom
was
saying.
If
you
imagine
that
the
data
mover
pav
is
making
a
request
the
service
over
over
rest
right.
I
So
we
agree
with
that
and
then
that
that
service
and
that
proxy
has
a
bunch
of
other
rest
services
that
are
registered
with
it,
that
the
differential
snapshotters
have
registered
with
that
service,
and
they
say
I
I'm
the
differential,
snapshotter
location
url
for
this
particular
volume
class
and
then
the
service
then
would
like
proxy
the
request
to
that
particular
differential
snapshotter
over
rest
and
that
differential
snapshotter
would
then
be
running
with
the
csi
driver
and
access
to
call
the
csi
driver
over
over
grpc
does
that.
I
I
D
The
the
difference
on
snapshotter
in
that
sense
is
a
service.
It's
not
like
a
controller.
It's
not
listening
to
the
cr
or
anything.
It's
just
a
service
that
so,
instead
of
listen
to
the
cr
object,
it
will
just
listen
to
the
request
request
coming
from
the
proxy
right.
C
Okay,
so
it's
not
a
direct
call.
It's
not
like,
because
the
the
the
line
you're
having
here
is
the
proxy
part
is
making
a
get.
D
D
B
Right
because
this,
if
we
open
a
service
on
the
different
differential,
snap,
shooter,
basically
factory
cycle
it
just
it
has
a
service-
endpoint
connects
to
it
now.
The
problem
is:
how
do
you
control?
Who
can
call
that
we
don't
have
that
problem
for
external
snapshooter
or
external
provisional,
because
they
act
on
non-namespace,
worm,
snapshot,
content
or
pv
resources?
D
Okay,
I
mean
this
is
just
an
update
of
what
we
have
so
far.
I
don't
want
to
visit
944,
so
I
I
think,
unless
you
guys
want
to
continue
talking
about
this,
maybe
we
can
call
another
meeting
to
discuss
like
I
like.
I
said
there
are
so
many
questions
to
left
and
that
this
is
just
one,
my
one,
something
that
I
think
I
propose
so
hopefully
we
can
continue
working
on
it
and
and
coming
out
with
more
a
better
solution
or
a
community
solution
for
this.
D
B
Thanks,
we
are
17
minutes
away,
so
shin
you
want
to
jump
to
do
we
do
we
want
to
talk
about
next
action
items
on
this.
I
guess
yes,.
D
Should
we
schedule
it
for
another
meeting
or
should
we
continue
talking
here.
C
We
can
talk
about
this
next
meeting.
Probably
just
think
about
this.
I
I'm
still
thinking
we
should
go
back
just
to
ucrds,
it's
just
the
same
model
rather
than
you
know,
because
we
we
we
go
this
way.
We
have
other
set
of
problems,
just
changing
how
about.
D
C
Maybe
you
should
also
write
down
the
like
the
crd
approach
like
pros
and
cons
that,
because,
let's
say
even
if
we
propose
this
fine,
your
cap
you'll
get
questions
hey.
Why
don't
you
know?
Why
didn't
you
think
about
the
other
approach
right,
so
we
wanted
to
say:
hey.
We
discussed
several
like
three
alternatives.
C
D
J
B
C
D
Okay,
that's.
C
All
right
all
right,
so
this
is
the
basically.
So
what
is
snapshot.
C
Right
yeah,
I
will
start
this
one
so
because,
because
we
actually
talked
about
this
about
a
year
ago,
I
just
checked
the
agenda,
and
then
we
found
out
that
we
actually
talked
about
this
one
april
last
year
right.
So
this
is
the
kind
of
front
that
notes.
We
talked
about
phase
one
phase
two,
so
we
are.
We
are
done
with
phase
one
phase
two
so
right
now
the
api
is
so
basically
in
in
may
2021.
We
when
we
do
a
release
of
external
snapshotter.
C
C
Are
folks
ready
for
that.
So
the
reason
I
was
thinking
about
one
reasons
we
already
talked
about
this
one
a
year
ago.
The
other
reason
is
because
you
guys
know
we
are
working
on
the
supporting
mode
conversion
thing,
we're
adding
a
new
field
in
volume,
special
content
so
and
then,
after
that,
that
will
be
added
to
the
v1
api
only
right
so
that
just
to
make
it
looks
very
different
from
the
event
better
one,
a
piano
there's
another
extra
feel
now.
C
So
I
think
it's
a
good
time
to
think
about
the
removal.
Essentially,
you
want
to
add
more
to
this.
B
Yeah,
just
so
the
group
to
be
aware
so
for
any
public
any
any
apis
made
stable
like
we
want
better
one
apis.
We
need
to
follow
a
pretty
much
standard
procedure
for
retirement.
B
Of
course,
the
more
mature
the
api
is,
the
the
more
procedures
we
need
to
take
for.
We
want
better
one,
I
think
the
then
the
requirement
is
to
making
to
make
make
sure
the
api
is
still
available
in
that
one
year.
Time
frame
we're
approaching
to
that
date
because
we
made
an
announcement
last
year
in
may
right
right.
B
So
we
are
approaching
to
them
point
and
also
the
longer
we
keep
a
non-backup
incompatible
api
in
the
community,
the
more
maintenance
we
need
to
do
so
yeah.
I
I
fully
agree
with
the
shin
this
is.
This
is
a
probably
a
good
time
to
consider
remove
support
of
the
we
want
better
one
apis
completely
from
the
community
and
just
to
because
this
group
here
is
are
the
ones
who
are.
B
You
know
heavy
you
either
using
this
feature
or
depending
on
this
feature,
so
we
want
to
make
sure
your
business
is
not
interrupted
by
by
this
removal
and
if
there's
anything,
we
need
to
talk
about
to
make
the
process
smoother.
We
will
do
it.
There's
no
like
objections
from
this
group.
Maybe
we
should
take
next
steps
to
shin.
Do
we
need
a
not
another
announcement
or
something
like
that
in
in
the
pre?
You
know
in
1.24
to
we.
C
C
So
we
need
to
tell
them
by
the
end
of
march,
I
believe
march
30th.
That's
the
deadline
to
getting
this
included
in
the
blog,
but
there's
also.
C
B
Yeah,
like
likely
we
will,
by
next
meeting
of
this
group,
we
should
have
an
answer
right
either
we
move
forward,
then
we
will
go
ahead
and
reach
out
to
the
deprecation
team
to
come
up
with
a
blog
or
there
are
other
type
of
other
difficulties
we
need
to
address.
First,
then,
we'll
see
how
it
goes.
Does
that
make
sense
to
you
by
next
meeting
we
should
be
resolving
all
this.
C
Yeah
anyone
has
everyone
been
using
v1
already:
that's.
It
was
in
1.20
release,
so
it's
being
a
year.
I
C
Yeah,
I
just
want
to
make
sure,
because
I
remember
when
the
inventor
22
there
are
several
being
better
than
it
has
got
removed
then
caused
a
huge
problem
in
applications.
So
just
we
just
want
to
make
sure
we're
not
causing
that
problem
here.
G
G
C
B
That's
the
whole
rhythm
here
right,
so
I
suspect
you
know
everybody
should
have
already
moved
to
the
latest
version.
But
it's
to
shin's
point
right.
We
don't
want
to
cause
any
interruption
in
anyone's
business
here
because
given
this
feature,
so
if
no
objections
by
next
working
group
meeting
that
then
I
think
we
will
be
just
move
forward
and
remove
that.
C
Okay,
so
we
can
also
send
an
email
in
our
mailing
list,
so
that's
yeah.
C
Yeah
because
I
think
the
most
backup
vendors
are
already
in
this
ruling
group
right,
so
it's
probably
good
enough.
Okay,
we
also
bring
this
up
in
tomorrow's
surgery
meeting
just
to.
K
B
B
Update
you
just
you
just.
C
B
I
Yeah,
I
just
wanted
a
quick
update,
because
I
saw
that
there
was
like
the
cozy
meetings
and
agenda
aren't
available
anymore,
and
so
I
was
wondering
if
where
I
could
track
that,
because
I
saw
that
there
was
like
a
this
and
then
there's
changes
being
made
to
the
repository
as
well
in
regards
to
that
api
review.
So
I
was
just
wondering
if
there
was
like
a
meeting
or
if
it
was
part
of
this
meeting
and
then
oh.
M
C
But
I
can
give
whatever
like
I
know,
basically
right
now
we're
still
waiting
for
the
api
review
to
finish
right
yeah.
So
I
know
michelle
was.
I
know
she
was
actually
actually
trying
to
review
it
because
she
was
having
some.
You
know
some
guitar
issues.
It
looks
like
he
was
having
some
problem
opening
up
that
he
treated
me
that
that
empty
file
for
some
reason
yeah
so
so
definitely
api
review.
Is
I
mean
the
cab
review
is
ongoing
yeah.
C
So
that's
one
thing
that
is
remaining
and
then
cozy
tin
have
been
making
some
updates
already
on
those
ripples
right
based
on
you
know,
whatever
it
is
already
in
the
in
a
cab.
I
believe
there
are
only
some
small
small
things
that
we
need
to
get
resolved.
C
C
Is
there
is
a
channel-
oh
my
god,
here,
sixth
story
cozy.
Oh
there
you
you
joined,
looks
like
that.
You.
C
C
Double
check,
maybe
you
are
trying
to
access
the
old
one
see
this
one.
This
is
pretty
recent
well,
but
this
one's
tough.
M
I
C
I
C
C
I
C
All
right,
so
that's
this
one
and
then
also
the
content
notifier
so
and
now
we
still
need
to
go,
pin
the
sig
node
folks.
I
think
there
are
still
those
unresolved
concerns,
review
comments
from
the
rsig
node
and
also
there
are
a
few
from
some
api
review.
Alright,
so
I
think
that's
still
the
status
we
need
to
follow
up
with
them.
Those
are
those
need
to
take
some
time
to
figure
out
how
to
talk
to
them
and
getting
those
resolved.
Those
are
not
not
easy.
That's
why
we're
stuck
so
sean
chen.
C
Do
you
have
any
update
on
this?
Do
you
have
you
pinned
any
of
those
folks.
C
B
Cool
any
open
issues.