►
From YouTube: IETF108-LPWAN-20200730-1300
Description
LPWAN meeting session at IETF108
2020/07/30 1300
https://datatracker.ietf.org/meeting/108/proceedings/
A
Okay,
so
time
is
up,
we
are
the
lp1
working
group.
We
have
15
minutes,
so
we
will
ask
you
to
present
within
your
your
five
or
10
allocated.
Minutes
won't
have
time
to
make
a
huge
discussion
as
well
as
we
normally
do,
but
at
least
we'll
be
able
to
do
a
status
of
the
current
work.
A
So
please
keep
in
mind
that
this
is
an
atf
meeting.
So
all
the
best
practices
for
the
atf
apply
in
particular
be
courteous.
Try
to
speak,
slow
enough
for
people
to
understand
you,
and
obviously
there
is
the
ipr
thing.
So
if
what
you're
introducing
as
ipr
on
it
and
you're
aware
of
that
ipr,
please
let
us
know
or
tell
the
chairs
after
the
meeting
or
tell
the
mailing
list.
A
A
So
that's
where
you're
asked
to
log
in
with
your
data
tracker-
and
you
can
only
log
once
on
each
video
session
and
the
meeting
will
be
recorded
and
probably
played
on
youtube,
so
be
careful
to
us.
Even
more
so
we
thank
dominique
for
taking
the
the
minutes
and
doubling
told
us.
There
was
already
seven
people
helping
in
the
notes.
So
that's
really
cool
blah
blah
blah,
so
the
material
is
locked
as
usual
on
the
atf
side,
and
with
this
let's
go
straight
to
the
agenda.
A
We
don't
have
much
time
so
we'll
present
a
brief
status
of
the
draft
and
then
it's
going
to
be
10
minutes
for
one
calories
on
the
sick,
fox
draft
10
minutes
for
olivier
on
the
laura
one
draft.
Then
anna
will
tell
us
about
shakeover
and
biot.
A
And
it's
a
little
bit
small,
I
don't
know
if
you
can
see
it,
but
basically
what
this
tells
us
is.
There
is
only
one
discuss
left
on
the
static
context
for
co-op
second
cooperation
for
co-app
and
that
we
have
named
a
shepherd
for
the
lorawan
document.
A
We
we
don't
have
much
news
about
the
oem
draft
so
dominique.
If
you
want
to
just
tell
us
a
word
about
that
and
then
the
chic
of
a
ppp
draft
was
was
transferred
to
interior.
So
that's
why
it
doesn't
show
anymore.
So
let
me
do
you
have
news
about
the
the
orient
draft.
B
It's
still
there,
but
no
progress.
Sorry.
A
Okay,
but
it's
still
there
are
you
planning
to
you
plan
to
move
it
forward,
so
I
will
be
moving
the
slides
for
people,
but
if
you
want
to
present
from
your
screen,
you
can
also
do
that.
Just
let
me
know
if
you
prefer
to
present.
Otherwise
I
will
just
be
moving
the
slides,
and
with
this
we
start
with
juan
carlos.
C
C
All
right,
okay,
so
perfect,
so
hello,
everyone,
juan
carlos
suniga
and
I'm
going
to
present
a
quick
update
on
the
chic
over
sick
fox,
I'm
putting
my
video
just
in
case
it's
up
to
the
chairs.
If
you
want
to
upload
it
or
approve
or
not
the
so
the
updates
quickly
on
the
hackathon,
we
had
some
plans
to
to
do
some
interoperability
but
of
course,
we're
still
having
issues
with
the
with
the
covey
the
confinement.
C
So
what
we've
been
doing
is
some
offline
coding
between
the
university
of
chile
and
the
university
of
catalonia.
The
polytechnic
of
catalonia-
and
this
is
mainly
focusing
on
the
icon,
error
and
and
the
noack
modes.
The
idea
is
to
to
replicate
the
scenarios
and
then
start
moving
back
and
forth,
ideally
evolving
the
incline,
the
current
implementation
of
the
pi
chick
from
uchila
to
other
modes
like
noak
window,
downlink,
etc.
C
Regarding
the
drafts,
we
have
the
latest
updates
for
two
and
three
combined
here,
so
more
details
on
that
on
error,
data,
fragmentation
and
integrity
mode.
We
added
and
updated
the
references
and
terminology,
that's
already
from
revision.
Two,
the
usage
of
sequence
number
from
sigfox,
as
we
described
before,
to
optimize
the
retransmissions,
especially
around
the
all
one
fragment
and
and
then
we
added
a
recommendation
for
noac
usage
and
which
I
will
explain
in
a
in
a
further
slide.
C
We
can
move
on
to
the
next
place,
so
the
text,
testing
network
architecture
doesn't
change
how
we're
still
using
the
the
five
chic
software
on
the
device
right
now,
relying
on
a
micro
python,
running
on
on
a
pycom
low,
pi,
4
device
connecting
to
the
sigfox
network
and
the
in
the
functions
on
the
platform
is
based
on
a
cloud
native
implementation.
In
this
case
running
on
google
cloud
next
slide,
please.
C
So
the
no
ack
is
the
the
one
that
I
mentioned
before.
We
added
this
recommendation
to
use
the
fcn
to
indicate
the
packet's
size
and
if
you
remember,
we
had
that
discussion
at
the
moment
when
we
were
drafting
rfc
8724,
we
actually
requested
recommendation
of
fcn.
Originally,
we
were
planning
to
use
the
fcn
only
as
a
binary
number
to
indicate
either
the
the
middle
transmission
or
the
end
of
the
transmission
of
a
noack
packet,
but
we
suggested
use
recommending
the
the
use
of
fcn
to
indicate
the
size
of
the
packet.
C
So
that's
what
we
are
adding
here.
What
we
say
is
that
this,
in
combination
with
the
sig
fox
sequence,
number,
will
allow
implicitly
indicating
the
expected
number
of
fragments
and
or
the
size
of
the
packet
that
will
be
received.
So,
of
course,
there's
no
acknowledgement,
there's
no
date.
Integrity
in
case
there's
a
fragment
missing
too
bad.
We
lose
it,
but
at
least
the
network
will
get
an
indication
of
the
size
of
the
packet
and
and
the
integrity
of
the
packet,
even
though
it
cannot
recover
from
from
losses.
C
So
next
slide,
please,
by
the
way
we
are
planning
to
to
implement
this
and
test
it
in
the
in
the
next
version
of
the
of
the
pie
sheet.
C
So
as
next
steps
again
interpretability
test,
ideally
with
pacik
and
other
implementations,
we
had
planned
that
for
vancouver.
Of
course,
didn't
happen.
Madrid
didn't
happen,
so
we
don't
know
about
bangkok
right
now.
It's
a
question
mark,
I
would
say
a
big
question
mark,
so
we
need
to
be
creative.
If
we
want
to
do
some
interoperability
tests
this
year,
I
guess
otherwise.
The
offline
implementation
is
still
going
and
in
a
very
good
shape,
and
the
idea
is
to
continue
the
fine
tuning
of
timers
shift,
header
fields
etc
to
progress.
D
Yes,
for
the
test
we
are
now
running
or
we
are
trying
to
run
an
openshift
core
on
server,
so
it
can
be
a
way
to
test
interoperability.
C
Okay,
so
is
this
well
well,
we
can
talk
offline,
but
is
this
available
almost.
C
A
A
C
A
C
C
Okay,
so
I
see
the
question
is:
does
the
use
of
sigfox
sequence?
Number
means
you
cannot
interleave?
No,
in
fact
the
the
sigma
sequence
number
is
is
generic
it's
not
per
session,
so
we
need
to
be
a
little
creative
there
and
have
a
a
generic
or
or
how
can?
I
say,
a
counter
that
is
that
spans
over
multiple
sessions,
but
as
long
as
we
we
keep
track
of
that
that
counter
we
we
can,
we
can
use
the
same
sequence
number
independently
of
the
number
of
sessions.
A
E
B
E
Meeting
and
some
words
about
the
next
step,
if
you
can
go
to
the
next
next
slide,
please
so
we
did
the
last
call
on
draft
zero.
Seven,
thanks
to
all
reviewers,
we
had
a
lot
of
good
comments.
We
integrated
in
draft
eight
that
we
published
about
two
weeks
ago
so
now
we'll
see
the
challenges
in
that
draft.
E
E
We
use
crfc87837
terminology,
so
there
is
no
more
any
device
on,
but
its
devices
and
so
on
and
so
on.
So
every
draft
should
be
consistent
and
easy
to
understand
on
all
technologies.
E
We
installed
my
possibility
in
the
czech
acknowledged
example.
We
add
some
more
examples,
fix
some
needs
and
typos
and
we
spent
a
lot
of
time
specifying
the
retransmission
timer.
We
we
already
discussed
that
a
lot
in
the
interim
meeting.
There
was
no
retransmission
timer
and
now
there
is,
there
is
restoration,
the
transmission
timer
and
they
are
optimized
for
lower
one
case
and
they
also
take
into
account
the
network
server
that
can
not
manage
their
queues.
E
So
next
steps
we
published
a
draft
eight
two
weeks
ago.
We
didn't
receive
any
comment,
yet
I
don't
know
if
it's,
because
you
guys
are
missing
time
or
everything
is
okay
for
you.
So
I
don't
know
if
we
should
end
the
last
call
somewhere
in
the
future
or
it's
my
first
documentary
atf.
So
maybe
the
chess
can
help
us
here.
A
Classical
is
finished,
it's
just
that
we
were
solving
the
comments
that
were
given
during
the
last
call.
Now
clearly,
if
one
comment
was
was
given
after
the
the
end
of
the
last
car,
it
was
an
interesting
command.
We
would
still
process
that.
But
yes,
officially
following
the
process,
we
are
done
with
the
last
call.
We
are
just
solving
the
issues
and
once
we
are
happy
that
the
issues
are
solved
then
ready
to
go
right.
So
it
seems
to
me
that
version
8
could
be
pushed
for
for
the
ietf
review
now.
A
So
I
guess
the
first
thing,
which
which
will
happen
is
the
shepherd.
Does
the
shepherd
write
up?
I
guess
the
minute
you've
started
that
and
presses
the
button
you're
used
to
it
dominic.
If
not
just
I
can
work
with
you
to
see
what
button
exactly
it
is
to
be
pushed.
But
it's
publication
request
button
on
the
data
tracker
and
once
this
is
done,
then
normally
the
eric
and
the
rick.
A
A
G
Okay,
I
will
press
yeah,
okay,
thank
you.
Okay.
This
graph
gives
a
recommendation
for
using
chick
on
the
3gpp
architecture.
The
complete
work
must
be
done
at
the
etsy.
As
you
can
see
in
this
slide,
the
the
3gpp
architecture
is
more
complex,
the
the
one
we
have
used
in
80
in
the
rfc
8376.
G
H
Yes,
no,
I
just
wanted
to
say
that
we
have.
We
are
a
little
bit
ahead
of
time.
So
initially
you
had
five
minutes
for
a
presentation.
I
think
that
you
can
do
ten
minutes.
So
it's
okay.
We
have
some
time.
Okay,.
G
Thank
you.
So
this
is
the
first
time
we
present
this
draft
completely
and
today
we
have
also
finished
the
fragmentation,
configuration
and
parameters
next
slide,
please
so
as
okay,
this
one.
So
as
I
was
telling,
the
architecture
is
more
complex
and
we
divide
the
work
in
different
uses
cases,
and
I
present
each
use
case
because
it
will
be
different
for
the
network
gateway.
G
G
Here
we
are
using
the
pdcp
layer
as
the
3gpp
defined
all
the
header
compression
in
this
layer
as
rock
or
or
band
jacobson.
So
here
we
can
put
a
chic
compression
in
this
configuration.
We
don't
need
fragmentation
because
rlc
layer
makes
segmentation,
so
we
don't
need
chick
com,
fragmentation,
so
the
for
the
rule
id
we
can
get
dynamically
a
number
and
it
could
be
up
to
nine
bits
next
slide,
please.
G
We
have
a
very
little
packets,
so
we
will
see
at
the
end
of
the
configuration
in
this
use
case
also
we
will
use
or
we
can
use
control
link
that
has
a
transfer
block
from
2
bytes
to
125
bytes,
and
we
can
also
have
data
link
where
we
can
have
1
300
bytes.
So
we
have
very
big
big
bandwidth
and
very
small
bandwidth
next
slide.
Please
the
third
use
case
is
when
we
can
use
the
application
server
as
the
network
gateway.
G
G
C
Well,
yeah,
you
finished
with
the
use
cases
I
just
wanted
to
mention.
Yes,
all
right,
two
questions,
the
first
one
on
the
number
two.
I
guess
you
would
need
to
transfer
the
context
to
the
visited
plm.
In
this
case,
you
need
some
sort
of
coordination
between
the
device
owner
and
the
roaming
network.
That
seems
a
little
complicated.
G
No,
it's
it's
already
done
for
rock,
so
it's
already
done
for
rock.
We
are
going
to
do
the
same
thing.
It's
in
the
the
home
network
will
send
the
hash
alert
for
the
home
networks
and
the
information
of
the
context
to
the
actual
network,
or
I
don't
know
how
to
say
it,
but
virtual
network
and
the
one
you
are
now
so
the
two,
the
two
hash
alerts
will
manage
to
send
the
information
one
to
the
other.
C
Okay,
that's
that's
nice
and
probably
good
to
to
explain
because
at
least
to
me
it
wasn't
obvious,
but
if
there's
a
solution
that
would
be
interesting
to
know.
And
the
second
question
is
you:
you
didn't
consider
over
the
top
chic
like
end-to-end,
like
running
transparently
over
the
3gbp
network,
in
the
sense
that
you
could
the
network
the
3gbp
network
could
be
completely
not
aware
of
the
shake
being
used
on
top
of
mbiot.
G
So
I
have
chick
here
on
the
top
of
the
network,
so
it's
the
3gpp
network
is
clear.
I
cannot
do
anything
inside
and
I
have
to
manage
the
the
rules
and
the
compression
context
and
everything
from
the
device
to
the
application
server
is
that
your
system
yeah.
G
So
for
fragmentation
configuration
we
are,
we
want
to
use
the
ack
on
error
mode
and,
as
I
explained
before,
the
transfer
block
is
between
16
bits
to
1000
bits
and
they
vary
for
reach
16
bits.
G
A
G
G
G
The
size
of
the
transfer
blocks
it's
a
multiple
of
16
bits.
B
I
guess,
and
I
just
said,
transfer
block
range
from
16
to
1000,
with
an
increment
of
16.
C
G
Okay,
so
in
this
configuration
we
we
try
to
get
the
best
header
size
for
for
this
transfer
blocks
and
we
decided
to
make
two
configuration
one
of
eight
bits
and
one
for
16
bits.
G
The
one
for
eight
bits
perhaps
could
be
used
for
the
transfer
blocks,
less
of
than
300
bits
with
the
rule
id
of
three
bits:
the
the
tag
of
one
by
width
and
fcn
of
three
bits
and
the
window
of
one
bit.
G
A
So
in
particular,
I
would
really
appreciate
that
people
who
worked
on
the
lorawan
specifications
who
got
used
to
this
kind
of
document
could
see
if
there
is
anything
any
section,
any
kind
of
text
that
is
globally
missing
in
this
document
or
if
we
have
the
same
coverage
as
we
have
with
laura
one
so
review
and
others
I
mean
evil.
If
you
could
have
a
look,
that
would
really
really
be
nice.
A
H
Okay,
so
he
says
that
he
will
he'll
look
at
it
and
to
give
you
as
well
for
asking
for
the
deadline.
I
I
wouldn't
expect
olivier
that
we
have
any
hard
deadline.
So
if
you
can
do
that
by
in
the
next
in
the
following
weeks
or
or
a
month,
that
would
be
really
great
to
have.
First
things.
A
Yes,
I
mean
you
should
if
you
could
have
a
look
before
now.
On
september
1st
I
mean
some
of
us
have
been
in
vacation
that
for
several
weeks
I
don't
know
not
me,
but
so
enjoy
your
vacation
do
your
stuff,
but
you
know
when
they
start
over
in
september.
It
would
be
nice
that
you,
if
you
could
have
this
lock.
F
D
Okay,
so
that's
my
time,
I
think
yes,
so
I'm
going
to
talk
about
the
data
model,
so
the
last
version.
So
this
now
we
have
something
that
is
quite
stable.
You
can
have
access
to
it
on
the
lp1
repository
on
github,
and
so
the
you
have
the
chic
young
file
with
the
date.
This
way
we
can
see
which
one
we
talk
about.
D
So
what
what
is
new
in
this
young
document
is
that
we
have
merged
this
just
before
we
had
one
young
file
for
the
id
and
another
one
for
the
rules,
and
now
both
of
them
have
been
merged
in
a
single
file.
We
added
this
also
some
identifier
for
check
parameters,
mostly
all
of
them.
We
have
also
defined
the
fragmentation
parameter.
We
add
also
description
to
explain
what
are
the
meaning
of
each
fields
and
also
we
will
work
a
lot
of
the
mandatory
optional
fields
in
the
young
models
to
to
optimize
things.
D
D
So,
to
give
you
an
idea
of
how
we
define
identity
ref
for
all
the
element,
so
you
see
the
first
line.
We
define
a
base
type
which
say,
for
example,
this
is:
will
this
will
be
used
for
rcs,
algorithm
and
so
it
says
identity?
Then
we
defined
each
identity
for
all
the
lcs
algorithm
that
derive
from
this
base
type,
and
then
we
define
a
type
def
that
will
be
used
in
in
the
young
model,
so
next
slide.
D
So
this
leads
to
the
definition
of
different
types.
One
is
for
the
field
id
one
is
for
the
length
which
is
an
union
of
a
number
unidentity
ref.
We
have
one
for
direction
indicator
one
for
matching
operator,
compression,
decompression
action,
etcetera
exterior.
You
have
the
list
here.
So
what
we
try
to
do
is
to
be
to
take
all
the
fields,
for
example,
for
a
field
id
all
the
field
that
can
exist
and
even
in
the
future,
we
include
here
field
that
can
be
taken
from
the
aom
draft.
D
Okay
next
slide,
so
here
is
the
definition
of
rules.
So
definition
of
rules
is
it's
a
list
of
rules
that
is
that
are
identified
by
rad
and
the
rule
length.
We
also
add
rule
version
that
can
be
used
for
management,
but
this
is
a
value
that
is
not
defined
in
the
chicago
c
and
then
we
have
two
nature
of
rules.
D
D
Then
we
have
the
lcs
algorithm,
as
we
saw
before,
and
the
max
window
size
that
is
normally
compute
from
the
fcn
size.
So
it's
two
power
fcn
size,
minus
two,
but
we
can
reduce
this
value,
and
so,
in
that
case
it's
the
maximum
window
size
that
is
set.
We
have
a
transmission
timer,
so
this
is
a
mandatory
timer,
so
it's
a
value
between
one
second
and
the
maximum.
D
A
Just
curious,
but
when
we
say
that,
how
do
we
indicate
the
unit
of
time,
for
instance,
for
a
transmission
timer,
I
mean:
do
we
base
it
on
the
specs,
so
people
have
to
know
this
and
then
the
spec
or
is
there
a
way.
H
Once
we
we
decided
okay,
well,
that's
the
way
to
when
we
are
happy
with
the
model
and
the
the
data
model
is
pretty
stable.
We
can
ask
for
the
young
data
doctors
to
to
review
it
and
they
will.
You
know
they
can
indicate
if
there
is
a
better
type
to
put
here
or
just
leave
it
as
it
is
and
with
with
the
command.
D
Then,
for
the
okay
for
the
last
field,
we
have
a
fragmentation,
mod
type
that
can
be
no
accuracy
always
and
for
each
of
them
we
can
define
parameters
and
right
now
we
just
define
parameters
for
account
error,
and
these
are
questions
that
you
for
interoperability
issues.
So,
for
example,
do
we
put
tiles
in
all
one,
so
it
can
be
yes
or
no,
or
we
can
say
since
the
protocol
or
it
that
the
sender
can
say
choose
if
he
use
it
or
not,
and
we
have
the
act
behavior.
D
So
next
slide
is
for
the
compression,
so
the
compression
is
not
new,
so
I
will
not
go
into
much
details.
The
only
thing
that
I
think
it
has
been
done
just
before
it.
That's
a
target
value
is
a
union
of
binary
and
string.
So
we
can
easily
carry
any
size
of
binary
information
or
string,
for
example
like
in
uri
path
in
in
in
co-op,
and
you
have
all
the
other
parameters
on
the
rules.
So
next
slide,
so
here
I'm
quite
happy
with
the
rule
now
or
the
young
model.
D
D
So
for
me,
it's
okay,
so
we
can
do
this
work
before
doing
a
last
call,
but
last
call
may
be
a
way
also
to
to
push
and
to
do
the
work.
So
I
asked
the
chair
to
do
this
item
and
by
the
way
I
received
some
comment
from
evo
this
morning,
but
are
quite
interesting.
We
have
thank
you
for
that.
It's
so
I
will
discuss
of
them
tomorrow,
but
I
don't
think
it's
big
big
changes.
A
A
Yeah,
so
if
we
do
it
in
september,
and
you
see,
then
why
don't
we
try
to
get
some
young
doctor
review
in
the
meantime.
G
H
I
I
I
would
probably
also
so
that's
a
question
or
maybe
an
observation-
that
it
might
be
good
to
wait
for
the
for
the
shake
the
co-op
over
chic
draft
to
be.
You
know
to
be
accepted
by
the
isg,
because
you
know
I
don't
expect
there
for
there
to
be
any
changes
on
the
data
model,
but
just
you
know
to
be
on
the
safe
side
to
to
know
that.
H
Okay,
well,
everything
that
is
in
the
chico
co-op
over
chic
draft-
can
be
expressed
with
the
young
data
model.
You
know.
A
A
A
H
Yes,
so
actually,
pascal,
we
need
to
to
find
this
out
and
probably
also
with
the
help
of
eric.
At
some
point,
it
was
the
the
yes.
I
Yes,
so
basically
it's
usually
you
ask
for
review.
There's
a
pretty
much
the
same
thing
as
you
ask
for
a
review
by
in
directorate
or
iot
directorate.
You
ask
a
review
by
young
doctors.
I
can
do
it,
I'm
not
sure
whether
the
chairs
can
do
it,
but
in
worst
case
you
can
send
an
email
right
and
people
will
do
it.
A
I
H
Are
busy
these
days?
So
probably
you
know
maybe
they're
all
french,
so
we
should
you
know,
probably
in
in
in
terms
of
timing.
We
if
we
aim
for
review
in
september.
That
would
be
really
good.
Of
course
we
can,
if
you
have
that
before
that,
then
that
will
be
really
good.
You
can.
A
Ask
now
and
we'll
see
yes
sure,
okay,
so
we.
D
H
It
would
be
really
amazing
if
you
have
both
documents
at
the
same
time
or
not
surprised.
I
don't
think
that
we
need
that
we
have
to
put
a
dependency
on
core
conf,
so
the
document
is
self
contained
as
it
is,
it
works
really
well,
it
works
also
with
core
conf
okay,
but
I
I
don't
think
we
we.
We
have
to
put
any
dependency
because
it
might
take
some
time
for
qualcomm,
and
so
you
know
let
no
don't
let's
ship
this
one
and
we'll
push
on
kirk
offside
as.
A
G
Okay,
now
I
will
present
the
status
of
the
club
static,
header
compression,
so
next
slide.
Please,
the
status
is
that
we
have
only
one
discus
to
solve
the
other.
Two
discus
have
been
solved
for
the
for
the
last
version.
G
Thank
you
to
all
the
reviewers,
all
the
inputs
and
the
discussion
we
have
about
this
draft
next,
like
this,
so
the
what
has
been
modified
from
version
13
to
version
15
is
so
we
have
divide
in
section
two.
The
first
figure
has
been
divided
in
three
to
better
explain
the
different
scenarios
we
can,
where
we
can
use
co-op
and
chic
with
co-op.
G
G
G
What
we
need
to
explain
is
that
to
make
compression
every
directional
field
could
be
divided
in
two
depending
on
the
direction
and
the
target
value.
So
when
an
earth
fill
id
is
bi-directional,
but
the
target
value
change
in
each
direction
to
make
the
compression
we
need
to
define
it
twice,
one
for
each
direction
up
and
down,
and
that's
what
we
call
unidirectional
and
we
directional.
G
G
G
What
we
have
done
is
to
put
a
definition
where
any
file
that
is
defined
as
as
now
the
co-op
options
with
type
length
and
value
format,
dt
lv
format
could
be
compressed
with
chick,
where
the
type
will
be
the
file
id.
The
length
gives
the
field
length
of
co
of
cheek
and
the
value
will
be
used
as
the
target
value.
G
G
G
What
is
the
dashed
under
the
lines
that
we
forget
forgot
to
to
change
the
unidirectional
direction
in
section
3.1?
G
So
we
need
to
check
again
this
part
and
it's
not
very
clear
how
we
can
we
that
we
reflect
the
way
to
compress
any
option.
So
we
need
to
rewrite
or
reflect
this
better,
and
he
had
the
discussion
about
the
security
consideration
where
he
sends
some
inputs
that
we
need
to
to
add
to
this
section
next
slide.
Please
that's
it
thanks
questions,
quick,
some
questions.
G
A
Exactly
at
the
at
the
end
of
this
meeting,
so
I
believe
we've
done
a
lot
in
particular
and
on
on
this
squad
draft
a
lot
was
was
done
during
this,
not
this
quarter,
but
yes,
this
quarter
between
atf.
So
thank
you
very
much
for
the
hard
work
that
you
discover
going
through.
A
H
H
And
thank
thanks
mini
takers,
thanks
dominique
for
leading
the
the
taking
effort
and
thanks
everyone
and
hope
to
see
you
all
soon
in
person.