►
From YouTube: IETF103-IPSECME-20181107-1350
Description
IPSECME meeting session at IETF103
2018/11/07 1350
https://datatracker.ietf.org/meeting/103/proceedings/
A
So
this
is
like
BJ
cami,
repeating
I'm
the
host
Terra
that
they're
equivalent
David
is
not
here
this
week.
I
don't
see
him
remotely
attending
so,
but
maybe
so
just
a
note
pal,
because
I've
seen
it
already
before,
not
just
read
it
very
quickly
because
I
hold
for
now
and
so
the
first
things
we
need
some
note
takers.
Maybe
any
volunteers
sign
some
people.
A
Alright,
so
blue
seats
are
already
going
around
and
then
we
have
the
first
things
is
agenda.
Do
I
have
any
modifications
on
or
things
that
people
would
to
add
to
attend
our
remove
or
change
the
order.
I
know
that
Daniel
is
going
to
be
in
other
meeting
for
first
half
an
hour
or
40
minutes
or
so
tonight.
So
so
we
have
to
take
slow,
it's
early
parts,
otherwise
we
might
run
into
presentations
while
he
is
not
here
yet
but
anyway,
so
we
have.
C
A
A
A
I
think
you
soon
so
it
should
be
out.
I
think
it's
13
s,
and
then
we
have
done
the
ready
drop
last
call
for
in
visit.
Ib
I
did
see
a
couple
of
reviews,
but
if
there's
any
people
wanted
to
do
more
and
do
people
think
that
that
document
is
actually
is
ready
for
going
out,
see
some
people
not
a
concern
with
it.
Robin
City,
so
I
will
start
getting
this
write-up
done
and
so
on
and
putting
it
out
there.
A
E
D
A
A
A
Few
physical
and
reason
for
it
was
accepted
because
the
outers
actually
contacted
me
and
say:
okay
I
would
like
to
get
this
one
done,
and
so,
if
other
people
have
documents,
they
won't
look
at
you
know
their
own,
they
should
start
working
on
them
and
conduct
the
chairs
and
a
stockade.
Should
we
started,
you
know
if
your
documents
ready
for
an
option,
call
or
simply
do
something,
I
mean
I'm
not
going
to
be
coming
and
hunting
people
and
say:
okay,
are
you
sure
draft
now
issue?
It's
your
individual
rough
now
ready.
A
Should
we
take
it
as
a
reputable
crop?
If
you
think
you
are,
it
is
ready,
come
to
us
and
say
that
okay,
this
is
ready.
Should
we
start
and
then
I
will
check
it
out,
and
you
know,
then
we
make
a
run
ups
and
call
us
also.
So
if
you
have
other
documents
that
we
want
to
probe
or
put
forward,
please
do
that
for
attic
for
exam.
The
auxillary
stuff,
like
these
ready.
D
A
All
right,
so
that's
that
group
status
for
that
drafts
that
we
have
as
a
working
group
drafts
now.
Okay,
then
we
have
a
couple
of
work
items
and
we
have
completely
new
stuff.
That
is
not
really
something
that
it's
not
you
don't
have
that
in
our
Charter,
it's
more
or
less
a
you
know,
part
if
occasional
stuff,
as
it
appears
PS
key
authentication
Hopis
both
motors.
Then
we
have
some
immigration
issues
coming
from
polar
II.
A
I
think
that's
actually
very
good
to
have
you
know
some
implementation
outs
and
noticing
what's
problematic
with
the
current
specification,
and
perhaps
we
may
need
to
start
working
on
peace,
talk
a
bit
of
that
or
do
something
else.
I
tell
you
have
a
couple
of
documents
about
over-under
bit
about
the
compression
mode
for
ESP,
and
then
there
I
question
two
notifications
for
ipv6
a
topic
for
those
last
two
ones.
I
think
this
is
the
all
right.
A
D
Okay,
so
we
saw
another
CV
eaters
year
about
week,
pre-shared
keys
so
another
attack,
they
optimized
it
slightly
and
they
we
thought
that,
but
only
IV
one,
a
crisis
mode.
They
could
attack
this
offline
that
ninety
came
up
with
some
new
technology,
so
now
they
can
do
other
modes
of
attacking
preciate
Keys
offline
result.
D
They
just
have
to
get
her
a
little
bit
more
data
and
then
they
can
do
another
dictionary
attack
against
Ike
view
on,
and
this
is
happened
a
lot
of
times
and
but
it
keeps
coming
back
to
us
so
so
we're
trying
to
tell
people
to
you
to
use
strong,
appreciate
keys,
but
they
don't.
We
tell
it
here
in
RFC,
72
96,
we
can
say
it
is
common
but
insecure
practice,
but
we
don't
tell
them
how
to
configure
it
anyway.
D
We
say
it
another
time
in
the
same
RFC
deriving
a
shared
secret
from
a
password
name
or
other
low
entropy
source
is
not
secure.
Please
don't
do
this
well,
people
do
it
anyway
and
then
in
the
new
PPK
drafty
of
course
also
say
well.
These
PPK
should
also
be
really
secure.
They
should
be
at
least
256
bits
of
entropy
and
we
know
from
experience
from
like
I
guess,
15-20
years
of
experience
that
people
just
don't
do
this-
we
have
v
198
section
3-
tells
us
some
pre-shared
key
strength
that
we
need
to
use.
D
So
this
one's
actually
friends
it
implemented
by
luis
von.
Only
if
you
run
in
ships
not
ironically,
because
the
library
crashed
and
has
happened
because
the
underlying
crypto
lie
we
actually
properly
implemented,
is
the
Nutley
respond
itself.
We
fix
that
and
then
it's
another
fifth
publication
that
I
quoted
in
here,
because
after
reading
this
I
actually
didn't
really
fully
understand
what
we
should
do
here.
So
it's
a
little
but
but
again
it's
its
guidance
to
people
on
how
to
use
strong,
pre-shared
keys.
D
But
users
are
not
implementers.
So
it's
nice
that
implementers
read
rfcs
and
infinitives,
but
they
don't
force
users
to
have
strong,
appreciate
keys.
So
why
do
people
still
use
preciate
keys?
The
biggest
reason,
I
think,
is
that
when
you
have
a
site
to
site
VPN,
there's
two
organizations
involved
and
you
do
not
have
a
share
trust.
So
you
cannot
use
certificates
because
who
will
be
the
root
CA
that
has
their
the
ultimate
key.
So
so
people
fall
back
to
using
pre-shared
keys
and
the
same
for
their
friends
and
on
certain
operating
systems.
D
If
I
look
at
how
how
is
IPSec
been
attacked,
we've
actually
more
or
less
the
protocol,
as
withstood
any
attacks,
except
that
the
priests,
weak
pre-shared
keys,
the
the
sloth
attack
came
close,
but
we
had
some
random
is
in
the
spider
prevented
attack.
But
but
this
is
basically
our
biggest
problem
for
the
last
15
years
and
I-
don't
see
it
improving,
so
my
question
is:
should
we
do
something
and
this
sort
of
originated
about
I?
Think
two
years
ago,
I
did
something
in
our
implementation.
Leave
respond.
D
What
you
do
so
I
would
like
to
write
in
a
document
with
the
working
group
that
sort
of
gives
more
guidance
that
maybe
could
be
implemented
by
implementation,
so
that
the
implementations
start
enforcing
a
minimum
strength
on
pre-shared
keys.
If
you
think
this
is
a
bad
idea,
then
I
would
be
interested
in
how
otherwise
we
can
reduce
the
use
of
weak
pre-shared
keys.
F
D
F
F
Of
number
three,
then
the
answers
of
that
is
use
a
fake.
Now,
of
course,
we
went
through
this
a
long
time
ago
and
unfortunately,
the
the
previous
chairs
of
this
group
completely
mishandled
the
whole
thing.
It
turned
into
a
fiasco
and
it's
cuz
having
a
single
standards
track,
authentication
mode
for
Ike
v2.
We
have
three
informational
ones
and
some
stupid
guideline
framework
thing
that
nobody
implemented
any
of
that
stuff
right.
F
So
what
we
should
do
is
if
you
want
to
write
another
RFC,
we
should
read
an
RFC
that
does
a
Paik
and
we
can
deprecated
all
of
the
other
three
fake
things
in
the
guide
and
the
framework
guideline
thing
and
just
have
a
have
a
secure
way
of
using
PSK
is
then
we
don't
have
to
worry.
We
we
take
the
human
element
out
of
out
of
that
right.
G
G
Informational
takes,
but
if
we
want
to
do
something,
it's
very
easy
in
practice
to
generate
a
secure,
PSK,
just
take
16,
bytes
off
of
dev,
random
and
and
then
convert
into
hex.
So
it's
typable
and
people
don't
want
to
do
that
because
it's
inconvenient
and
they
don't
see
how
we
can
never
get
past.
That.
A
A
C
A
A
D
A
So,
but
but
go
back
to
this,
so
if
it's
a
password,
many
human,
readable
human,
remember
both
stuff,
then
we
should
be
using
the
you
know:
secret
password.
You
know,
methods
that
we
have
already
those.
So
you
should
just
you
know.
Perhaps
you
should
implement
those,
and
then
you
can.
You
know
say
that
if
anybody
wants
to
use
PSK,
it
must
be
a
strong
and
it
must
be
having
this
kind
of
you
know
randomness
in
it,
and
if
you,
if
you,
if
you
try
something
else,
it
says.
D
H
D
H
I'm
saying
that
if
now
we
have
some
problems
of
too
too
many
pegs
that
can
be
used
with
Ike
now
or
later,
the
maybe
is
a
ways
to
wait
for
staffer
g2
to
choose
one
or
use
any
pegs
that
we
have.
But
my
opinion,
yes,
is
that,
for
the
third
question,
is
that
if
we
don't
restrict
users,
not
to
use
user,
selects
the
space
case,
then
we
only
must
use
pegs
of
any
security
pegs,
as
we
have.
I
I
Sorry
so
I
think
it's
first
comment.
I
have
is
that
we
should
not
obsolete
the
PC
method.
It's
used
everywhere.
Everybody
know
how
to
use
it.
So
it's
not
going
away.
We
cannot
just
move
it
from
the
protocol
and,
if
it's
that
and
so
that
we
have
to
make
the
broker
stronger
than
then
the
audio
base
option
is
a
big.
E
Plus
so
first
I
think
that
option
is
not
an
option
because
actually
the
Pisgah
message
it
quite
secure
its
if
you
use
a
proper
is
key.
So
from
a
graphical
point
of
view,
it's
absolutely
sound,
so
is
there's
no
reason
to
drop
drop
it
from
IP
to
and
you
can't
you
can't
make
users
doing
stupid
things,
even
if
you,
if
your
implementation
will
require
some
I,
don't
know
checks
of
Biscay,
the
poor
user.
Just
we
will
write
it
down.
Put
this.
C
E
C
E
Some
previous
commenter
said
the
pake
is
a
good
option
for
those
people
who
vote
for
those
if
you,
if
it
seems
that
users
will
still
try
to
use
short
password
short
base
case
like
passwords
and
Paik,
is
a
good
option,
but
unfortunately
we
don't
have
a
standard
way
to
do
it
in
like
we
do
now.
We
have
a.
C
E
J
Sean
Turner,
not
an
implementer,
so
take
this
for
what
it's
worth
I
agree
with.
Dan
we
screwed
the
pig
thing
up
the
first
time
around
I,
don't
know
if
it's
gonna
get
any
better.
If
we
do
it
again,
I
think
right.
Everyone's
right,
we
can't
really
get
rid
of
PS
Ches.
We
have
this.
Should
we
write
an
RC
on
you
know,
PSK
links
and
stuff.
We
have
comparable
drafts
for
that
for
symmetric
keys
and
how
to
do
how
big
the
key
should
be.
A
So
I
went
to
go,
meet
Tommy
there's
there
were
several
people
saying
it.
Okay,
one
of
the
problem
is
that
you
know
the
current
framework,
for
you
know
different
secrets.
Without
the
submitters,
none
of
them
were
standard,
and
the
reason
was
that,
but
we
couldn't
agree
on
one
and
when
we
are
when
we
say
that,
okay,
if
we
pick
one
or
the
other
two
willing
to
you
know
not
to
publish
them
and
they
said
no,
they
are
going
to
be
coming
out
as
an
individual
RFC's
anyway.
A
So,
which
was
you
know,
would
have
really
ended
up
in
the
same
kind
of
situation.
Anyway.
The
question
is
now:
is
that
I
I
think
we
could
of
course
ask
for
you
know,
see
a
party.
Can
you
pick
me
one
fake
and
if
you
can
give
me
as
s1
fake
that
we
can
actually
you
know
use
for
at
they
would
actually,
you
know,
make
an
internet
quite
opaque.
Then
we
can
actually
check
your
password
out.
Then
we
could.
Actually,
you
know
start.
A
A
H
C
K
H
Assassins
live,
so
I
am
afraid
that
even
so
I
assumed
that
even
if
we
can
choose
one
Pig,
then
at
least
there
can
be
selected
at
most.
Two
is
always
really
different
properties.
So,
for
example,
one
balance
take
one
augment
Peck
so
with
strict
recommendations
or
where
to
use
which
so
I
believe
that
clear
recommendations
can
be
made
and
can
be
used.
I
believe
that's,
fair
G
will
do
this
actually.
A
I
think
I
think
location
or
at
the
best
position
actually
to
pick
one,
because
if
he
has
an
implementation
for
the
you
know
for
the
framework,
and
he
implements
one
and
it
says
that
okay
I
have
implemented
this.
If
you
want
to
interrupt,
try
me
try
this
average
with
this,
then
probably
the
next
one
who
is
implementing
it
probably
going
to
pick
exact
the
same.
Then
we
have
exactly
one
so
I
hope.
H
Yes
again,
yes,
we
had
the
same
discussion
on
psychology
and
there
was
a
lot
of
opinions
that
we
should
select
either
0
or
1.
And
yes,
this
will
be
most
envious
for
everybody.
But
I'm.
Just
saying
is
that
at
least
I'm
sure
that
there
will
be
one
clear
recommendation
for
each
set
of
protocols
at
least.
F
Dan
hardens
again
so
getting
back
to
its
tennis
off,
you
said
yeah
there,
the
Seaford,
you
might
come
up
with
one
balanced
one
augmented,
but
if
we
were
to
do
a
bake
in
for
Ike,
V
I
think
it
should
be
balanced
because
either
side
can
be
the
initiator.
If
we
did
an
Augmented
Paik,
we
basically
shoot
ourselves
in
the
foot,
so
the
C
FRG
can
standardize
a
an
Augmented
take
if
they
want,
but
I,
don't
think
it
should
be
used
here
now,
regarding
you
know,
could
we
agree
on
one
peg?
F
A
H
After
so,
there
was
a
discussion
in
cipher
G
and
the
process
just
started
on
Monday,
so
Alexei
Melnikov
asked
for
questions
which
to
ask
for
each
peg
and
then
the
plan
is
that
panel
will
review
all
existing
pegs
to
select
or
to
make
some
recommendations
and
try
to
select
one.
We
if
I
am
not
mistaken,
we
don't
have
an
exact
time
frame
now,
but
I
hope
that
the
process
won't
take
years.
So
I
believe
that
we
can
be
more
or
less.
H
A
So
if
gr
gr
party
is
going
to
be
picking
one
anyway,
then
I
think
we
would
like
be
actually
coming
back
to
this
question,
but
is
your
fellow
probably
also
going
to
be
require
some
kind
of
chartering
change?
So
it's
we
can't
do
it
now
completely,
but
I
think
that's
something
we
probably
would
come
back
when
they
have.
There
are
a
little
bit
further
on
in
the
process.
Well,.
F
A
Actually,
in
in
this
is
actually
coming
back
to
this
question
comment
that
was
coming
out
earlier,
that
if
you
are
actually
having
balanced
connections,
meaning
either
one
can
be
in
each
other,
which
means
that
it's
going
to
be
this
kind
of
VPN
side
to
side.
There
is
no
excuse
of
not
using
strong.
You
know
wrong
purposefully.
Take
the
cases
where
we
actually
use
the
weak.
Be
excuse,
mostly,
is
where
you
have
somebody
typing
in
the
password,
or
somebody
actually
has
to
remember
it
or
you
have
you
know
this
kind
of
you
know
one-sided.
A
F
Slightly
disagree
with
you
on
that
I
think
the
only
people
are
really
strong.
Ps
KS
are
the
kind
of
people
that
you
know
drive
around
in
armored
cars
and
have
people
with
guns
that
deliver
CDs
to
to
people.
You
know,
that's
the
you
know.
Those
are
the
people
who
are
worried
about
quantum
computers.
Basically,
so
everybody
else
that
types,
the
stuff
into
a
router
is
not
doing
256
bits
of
entropy
at
rest
that
it's.
A
Actually
also
gives
a
good
question
should
be
actually
the
PPK
require
more
strict.
You
know
length
restrictions
say
that,
for
example,
that
then
it
must
be
always
at
least
this
long
and
put
it
in
our,
but
we
can't
do
that
in
a
PS
key,
very
easy
because
it
people
are
using
it
already,
but
for
PPK
bit
a
quantum
stuff.
Mister
all
the
whole
reason
there
is
to
be
secure
Agosto.
So
we
have
rather
have
it
be
long
enough
that
you
could
actually
put
forces
iran.
L
G
A
Got
egged
agree
on
that
I.
Think
our
test
system
did
allow
that
all
right,
so
no
more
comments
on
this,
so
I
think
we
should
put
down
in
there.
You
know
for
the
cheers
to
that
I
think
by
PI,
integral
on
this
course
that
we
should
ask
for
to
a
party
to
have
two
big
preferred
balanced
bake,
so
we
can
actually
use
it
here
and
we
can
probably
make
some
kind
of
you
know
request
to
them
for
from
from
the
chairs
about
that.
A
So
if
you
can
put
that
down
into
you
know
minutes
as
a
action
point
for
our
chairs,
so
we
try
to
make
kind
some
kind
of
email
that
will
first
put
send
to
the
list
and
then
ask
if
this
is
okay
and
then
you
can
bash
it
and
then
we
can
fix
it
and
do
a
loop,
a
loop
it
around
a
couple
of
times
before
we
can
set
it
out
all
right.
So
I
think
that's
all
for
that.
Let's
move
the
next
one,
then.
E
Hello
once
more,
very
responsive,
so
I
will
take
you
about
some
experience
of
employment
and
I
quit
200
HP,
so
t-spin
capsulation
was
defined
in
RFC
82
29.
It
was
more
than
a
year
ago
as
it
was
published,
and
the
idea
is
quite
simple:
just
use
another
transport,
and
so
it
modifies
I
behave
in
some
ways.
For
example,
original
Intuit
is
responsible
for
restoring
TCP
connection
if
it
is
broken,
and
there
are
some
special
behavior
with
my
bike
with
net
apply
this
like
fragmentation.
E
So
anyway,
the
RFC
is
not
long
and
seem
to
be
quite
simple,
but,
as
usual,
material
is
in
details.
So
having
implemented
it
record
some
experience
and
we
think
that
some
things
I
knitted
from
RFC
or
not
clarified
quite
well,
and
some
of
them
are
just
for
optimization
and
probably
some
optimal
behavior
of
the
protocol
and
a
few
things
are
really
important
for
reliability
or
internet
interoperability.
E
So
well,
I
will
mock
those
which
are
less
important
and
just
for
only
four
teams:
Asian
in
blue
and
the
real
imported
scenes
in
red.
So,
first
all
transmissions
so
I
could
to
define
some
transmission
policies
at
exchange
or
original
extend.
Exchange
initial
is
always
responsible
for
the
transmission
packets.
In
case
it
didn't
get
response
in
time,
so
this
piece
reliable
protocol,
so
we
don't
need
to
transmit
actually
because
TCP
takes
care
of
it,
and
if
you
returns
meat,
it's
in
some
real
situation.
It
can
even
get
worse
for
some
congested
networks.
E
At
least
people
try
to
send
all
the
transmitted
messages
that
is
just
not
needed,
so
it
will
make
congestion
even
worse.
Well,
it's
probably
extreme
for
taken
into
account.
The
check
messages
are
quite
small,
but
still
so
there
are
still.
There
are
some
situations
when
you
need
to
retransmit
in
case,
for
example,
when
TCP
is
broken
and
then
restored,
each
party
master
transmitted
all
outstanding
requests,
because
there's
no
guarantees
that
previous
requests
reach
reached.
E
The
other
party
before
the
connection
just
by
connection
is
broken.
So
there
are
some
changes
well
using
cookie
and
puzzles
well,
cookie
allows
responder
to
verify.
Initiate
recipe
address
without
creating
any
state
is
defined
in
I
could
to
Cora.
I
have
seen
I
go
to
RFC
and
we
still
see
big
cookie
is
less
useful,
because
TCP
can
shake
already
verifies
that
initiated
IP
address
is
real.
Well,
there
are
some
attacks
on
sequence.
Number
won't
talk
about
it.
I'm
talking
in
general
in
general,
TCP
verifies
that
you
are
talking
with
a
real
IP.
Well,
there.
E
Tcp
extensions
that
also
allow
us
to
disappear
responded
to
become
stateless
to
speak
cookies,
I
mean
but
anyway,
for
ike
a
cookie
is
not
really
useful,
with
disappea
and,
moreover,
chesapeake
rates
a
state
before
well
before
the
sim
back
can
shake
is
a
despair
upon
the
cry
state
before
seen
a
cream
shake
is
finished,
so
cookie
is
meaningless
at
this
time
because
it
is
intended
for
Ike
responder
to
be
stateless.
It
is
stateful
on
this
piece
back
and,
and
still
using
puzzle
still
make
sense,
because
it
it
allows
responded
to
a
fraud.
E
Some
to
make
initiated
do
some
work
to
prove
that
it
is
really
want
to
connect
well,
and
there
is
some
Corrections.
There
must
be
some
correction
generation,
not
not
correction,
but
some
qualification
generated
cookie.
It
is
still
a
local
matter,
but
with
TCP
responder
must
not
include
port
number
initiated
point
number
in
the
cookie,
because
it
will
be
different
with
every
TCP
connection.
So
well,
then,
error
handling
is
like
a
saying
it
again.
Echo
to
cord
rod
says
it
is
a
in
it.
E
All
responses
are
not
authentication
are
not
yet
so
initiate
and
need
to
wait.
Some
time,
even
if
it
receives
error
notification,
for
example,
no
proposal
chosen,
it
will
need
to
wait
some
time
before
some
more
packets
arrive,
because
this
notification
can
be
fought,
probably
with
UDP.
It's
all
true
is
disappear,
not
well.
If
you
receive
error
notification
with
TCP,
it
means
that
either
it
is
a
response
from
a
real
responder
or
the
TCP
connection
is
hijacked.
If
it
is
hijacked,
you
will
never
receive
a
response
from
dannion
responded,
so
there
is
no
way.
E
There's
no
need
to
wait.
You
can
just
immediately
act
and
the
error
notification,
so
it's
just
an
abbreviation
or
coffee.
Of
course
you
can
arrange
if
you
want
just
to
pose
a
user.
Well
interaction
with
my
bike
again,
I
will
see.
82
29
have
sections
that
describe
how
my
bike
should
handle.
Who
is
this
P
connection,
and
it
says
its
first.
You
must
try
if
IP
address
changed
is
changed.
E
You
must
recollect
meditation,
not
the
TechNet
detection
society,
not
detection
and
estimation
identification,
because
ours
will
be
difficult.
It
will
be
different
in
any
case
from
which
is
appearing.
A
UDP,
TCP
and
message
idea
of
TCP
based
exchange
must
remain
the
same.
So
it's
important
for
interoperability
so,
and
probably
one
of
the
most
difficult
problems
which
disappear
based
high
co2
is
interaction
with
a
chi
clusters,
so
there
is
an
RFC
62
and
63
11.
E
Sorry
that
defines
a
way
how
cluster
and
client
need
to
communicate
in
case
in
cluster
for
lower,
and
so,
if
fellow
takes
place,
the
new
active
not
immediately
sends
informational
exchange,
which
contains
a
new
message.
Id
and
optional.
Musicans
number
is
P
sequence
number
so
that
in
case
any
messages
are
lost
during
flora.
The
si
can
be
used
after
follower.
So
the
problem
is
that
in
TCP
after
fellow
you
will
probably
not
have
to
speak
on
action.
It
is
broken
because
well
either.
E
It
is
possibly
this
up
that
then
you
need
to
synchronize
this
p
state
constitutional
dissipate
between
pickup,
not
an
active
node.
It
is
not
always
practical.
So
if
you
synchronize
only
iock
state
TCP
connection
will
be
broken
after
follower
and
at
this
case
in
this
situation,
then
you
active
node
cannot
send
anything
because
it
is
not
an
original
initiative.
It
need
to
wait
until
the
super
connection
is
restored,
but
the
client
is
not
aware
the
connections
broken.
E
So
if,
if
there
is
nothing
to
send
it's
just
Waits
and
seats
and
backs
so
we
have
some
some
some
kind
of
clutch
did
lock.
So
I,
don't
know
what
device
in
this
situation,
probably
a
client
need
to
periodically
send
people
a
package
just
in
case
every
year.
I
don't
know
five
seconds
at
ten
seconds
so
that
if
there's
no
traffic,
of
course,
if
there's
no
traffic
from
client
to
server,
so
the
connection
is
there
so
that
it
get
known
that
connections
broken
anticipate
connections.
E
You
started
follower,
oh,
the
resorption
is
to
make
cross
the
wait.
Don't
immediately
send
don't
immediately
initiate
information,
information
exchange
and
wait
for
I.
Don't
know
a
predictable
amount
of
time
before
client
wakes
up
and
restores
the
connection.
So
I
don't
have
immediate
opinion.
What
to
do
well
and
if
you
worry
about
this
be
proxies
so
I've
got
some
respondents.
E
So,
for
example,
if
connection
is
broken
display
and
if
originally
initiated
is
a
modified
ike
behind
this
proxy
and
the
connection
is
broken.
This
initiate
is
not
known
about
it
so
to
understory
it
in
time
facially
time
in
fashion.
It
will
restore
it
on
me.
It
sends
some.
It
has
some
new
data
to
send,
and
so,
if
a
regional
responder
is
a
modified
like
different,
is
by
proxy,
so
responder
will
try
to
send
package
if,
in
the
leaves,
there
is
no
connection
to
speak.
Electric
connections
broken.
Of
course
all
the
situation
happen.
E
If
connection
is
broken,
so
everything
if
everything
is
up
and
so
but
sometimes
Cap'n,
so
question
comment.
Some
probably
probably
there
are
some
more
issues
in
specific
installations
that
I
don't
didn't
yet
come
across.
Probably
somebody
else
come
across
and
my
question
my
my
creation:
easy.
Isn't
it
time
for
eighty
two,
twenty
nine
beasts
or
not
yet
or
is
it
ever
needed.
M
M
I
agree
that
there
are
things
to
improve
upon.
I
would
probably
prefer
to
wait
a
little
bit,
maybe
for
more
of
like
these.
If
we
have
other
issues
or
we
lot
more
deployment
experience
to
put
in
there,
I
think
none
of
these
I
believe
are
like
critical
enough
to
require
something
show-stopping
on
it.
But
that's
just
my
opinion:
could
we
jump
back
some
slides,
just
kind
of
go,
give
individual
responses
to
some
of
the
issues,
maybe
just
start
at
the
beginning,
yeah,
so
I'm,
not
sure
exactly
why
you
mark
this.
M
There
may
be
an
implant,
that's
not
directly
attached
to
the
TCP
connection
and
I
think
in
the
case
of
retransmits,
the
fact
that
we
are
often
tunneling
TCP
connections
in
ESP
over
this
is
far
more
of
a
concern
than
the
IKE
retransmissions,
because
the
AI
packets
themselves
are
not
going
to
be
really
contributing
all
that
much
to
the
ingestion,
which
is
why
the
document
has
a
whole
bunch
of
considerations
for
why
congestion
inside
congestion
is
potentially
problematic.
It's
certainly
at
high
reordering
or
high
loss
rates.
That's
why
this
is
a
fallback.
M
A
M
On
this
next,
one
I
think
that's
good
to
bring
up
I,
don't
think
we
really
considered
cookie
and
puzzle
totally
with
the
cookie
calculation,
including
the
port
number,
because
it's
often
gonna
be
different.
Isn't
that
also
the
case
for
UDP
behind
Nats
and
the
reason
that
we
switch
over
to
for
your
500
is
that
we
can
be
behind
an
ad
and
your
port
number
is
going
to
look
different.
M
E
Not
uni,
actually
it's
a
local
matter,
so
I
will
probably
mistaken,
painted
it
in
red,
because
cookie
calculation
is
a
local
matter
and
an
example.
Example
given
an
RFC
doesn't
include
port,
but
some
implementation
may
include
because
it's
local
Mecca,
so
we
should
just
make
again
advise
that
ode
must
not
be
should
not
must
not
be
included
actually.
A
It's
usually
not
a
problem
because
I
made
the
nuts
map
the
port
number,
always
the
same.
So
from
the
responders,
the
server's
point
of
view.
The
source
port
is
always
going
to
be
this
random
number.
It's
not
the
same.
That
was
in
the
sender's
end,
but
because
this
cookie
is
calculated
by
the
server
and
it
verifies
that
it
comes
back
from
the
same
client
from
the
same
source,
port.
So,
okay.
D
M
M
Yeah,
that's
so
thank
you.
Next
yeah,
this
is
just
optimization.
Yeah!
That's
fine!
So
again,
here
with
the
interaction
with
mobike,
I
I
think
certainly
was
my
assumption
that
you
would
be
recalculating
the
nat
detection.
Upon
doing
this,
it's
kind
of
like
whenever
you
do
a
mobike
packet,
it
should
be
identical.
What
you
do
over
TCP
and
UDP
do
not
agree.
E
Well,
anyway,
what
will
be
difficult
and
not
detection
include
spot.
So
if
you,
if
you
understand
over
TCP
unmodified
packet,
then
not
will
Aaron
arrows
be
detected,
even
if
it
is
no
not
I,
don't
know
whether
it
is
good
or
bad,
but
well
to
be
to
be
so,
sounds
that
not
detection
to
work
correct?
You
must
be
calculate.
A
So
so,
whenever
you
are
sitting-
and
if
you
are
sitting
IP
addresses
in
mobike,
you
always
must
send
the
packet
exactly
as
a
saying
you
never
ever
can
modify
it.
You
can't
recalculate
the
net
but
detection
IPS
so
because
that
will
break
the
system.
The
problem
is
that
you
don't
know
if
your
original
UDP
packet
ended
at
the
other
end,
it
might
have,
he
might
have
the
response
might
have
get
lost
and
then,
if
you
retransmitted
in
TCP
with
a
different
content,
he
does
he
doesn't
know
what
to
do.
So.
A
That's
why
you
never
ever
modified,
but
on
the
other
hand,
mobile
says
that
if
anything
goes
wrong,
when
you
are
switching
the
addresses,
if
you,
if
you
switch,
the
address,
is
presented
for
the
same
thing
in
a
mobile.
If
you
are
doing
update,
address
IP
address,
if
the
IP
address
you
are
trying
to
use,
you
can't
use
it.
You
move
the
next
one,
you
still
Centaurus,
you
know
saying
oh
I'm
moving
this
address,
but
actually
losing
different
address.
A
A
E
N
Canaussie
Google
I'm
confused
in
my
understanding.
Please
do
correct
me
if
I'm
wrong.
The
purpose
of
NAT
detection
is
to
know
if
we're
sending
over
ESP
or
if
we
enabled
UDP
encapsulation.
If
we're
encapsulating
over
TCP,
there
is
no
benefit
whatsoever
in
that
detection.
We
can
just
take
the
results
of
it
and
throw
them
on
the
floor
right.
So.
E
As
its
true,
but
so
that's,
okay,
we
first
try.
Edp
is
mobike
always
so
in
this
case,
we're
not
detection
must
be
present
and
currently
well,
it
should
be.
Present
must
be
present
by
RFC
by
the
way
well
for
TCP.
Of
course,
we
not
more
tolerable.
So
it's
not
important
whether
it's
not
so
important,
whether
you
are
inherently
detect
and
said
there
is
not
when
there
isn't
a
yes,
but
okay,
probably
you
can
make
it
market
in
blue
to
simplify.
N
M
M
Right
yeah
I
think
this
is
this
is
a
good
one.
No
no
I
I'm,
not
sure
what
the
cleanest
way
to
handle
that
is
yeah.
So
I
think
this
is.
This
is
one
of
the
areas
where
I
think
it's
will
be
good
to
have
new
text.
I,
don't
think
it's
a
you
know
a
block
or
a
critical
thing,
but
we
should
definitely
roll
it
into
updates
and
then
yeah
proxies
I.
M
D
D
O
So
my
friend
I'm
also
going
to
do
a
plus
one
on
it's
not
ready
to
do
a
this
document,
but
I
actually
think
the
results
that
you've
documented
here
are
really
excellent.
In
and
listening
to,
the
the
experience
that
apples
had
I
think
is
really
helpful,
and
so
what
I'd
suggest,
instead
of
having
a
dis
trying
to
do
the
best
work
at
this
point,
is
to
take
your
document
and
adopted
as
a
working
group
item
and
move
it
forward
as
an
implementation.
O
G
So
the
of
nehrim
well
I
also
think
it's
too
soon
for
a
best
document,
but
rather
than
rather
than
write
that
out
as
and
is
it
right
now
as
an
RF
informational
RFC,
we
could
have
a
working
document
that
lasts
for
years.
I
mean
there's
no
rush
to
publish
it
and
eventually
it
might
become
80
229
bits,
but
it
can
start
with
just
a
documentation
of
the
implementation
experience
and
then
it
might
make
sure
either
way
one
way
or.
A
We
had
in
equation
two.
We
had
this
clarification
document,
which
was,
you
know
then
finally
published
as
an
RFC,
and
then
we
rolled
it
into
there
too
and
I
was
but
I
didn't
know.
If
it
actually
was
very
useful
to
actually
publish
it,
doesn't
you
know,
and
this
kind
of
RFC
it
could
have
been?
Also,
you
know
just
being
out
as
a
you
know,
informational
or
just
a
draft
to
say.
Okay,
this
is
a
draft
where
we
collect
all
the
problems.
A
We
have
a
TCP,
you
know
encapsulation,
and
then
we
think
we
are
ready,
meaning
we
don't.
They
are
not
going
to
get
any
more
of
those.
Then
we
would
start
working
on
the
you
know
TCP
and
copy
and
paste
document,
and
then
we
would
roll
or
all
of
these
information
from
the
craft
to
test
its
new
RFC
and
I.
A
That's
all
for
us
actually
something
we
don't
have
a
discipline
cup
source
and
piece
in
our
charter,
so
that
will
be
requiring
shattering
change
anyway,
but
that's
something
that
we
have
time
to
do
because
I
mean
I
understand.
This
would
be
at
least
out
first.
You
know
you
know
such
like
that
and
and
and
if
you
feel
that
it's
actually
important
to
publish
this
a
clarification
document
as
an
RFC
we
can
do,
but
so
what
watch
people's
opinions
on
these
two
degrees?
It
would
be
a
good
option
to
go
forward.
A
So
so,
let's
so,
let's
make
a
very
basic
hum
who
thinks
this
is
think
that
we
should
work
on
this
document
even
in
this
circuit
group,
even
if
it's
not
actually
in
our
current
charter.
So
if
you
dig
through
the
I
make
a
harmony,
the
first
person
is
to
take
these
super
context.
The
well
now,
if
you
think
so,.
C
A
A
A
Even
if
you
are
you
know,
I
said
I
think
we
can
actually
take
it
and
I
said
you
know
this
kind
of
item,
because
it's
still
a
minor
clarification
that
is
kind
of
you
know
things
that,
because
we
are
supposed
to
be
doing,
we
are
not
changing
anything
or
adding
new
stuff
that
were
just
clarifying
the
old
stuff,
so
I
think
it,
but
ok
doodle
it.
Actually,
let's
ask
if
we
okay.
A
A
A
I'm
not
sure
about
that,
because
I
mean
it's
it's
sometimes
it's
actually
much
more
easy
effort.
The
good
thing
about
the
clarification
document
force
that
they
like
versatile
was
that
it
was
only
concentrating
on
those
items
that
were
important.
For
you
know
it
was
actually
you
know
unclear
in
the
previous
document.
Did
it
have
a
hundred
pages
of
you
know?
Extra
stuff
that
was,
you
know,
would
have
been
hiding
the
basic
you
know
core
of
the
issues,
so
I
think
that's
I
think
that
we
should.
A
We
should
start
on
working
on
and
it's
actually
much
easier
to
work
on.
If
you
could
just
add
an
issues,
there
is
issue
issue
and
not
think
about
what
would
be
actually
there.
You
know
Excel
so
loose
on
how
to
write
the
draft.
You
know
in
the
way
that
they
access
all
of
this
issue,
but
in
some
cases
we
are
going
to
just
have
a
list.
You
I,
we
don't
have
a
solution
yet
all
right.
Okay,.
E
A
E
E
D
Cycle
Clara,
so
I'm
read
that
is
helping
the
Linux
kernel
people
in
adding
this.
There
was
about
half
a
year
ago
a
original
patch
submitted.
They
thought
I
was
too
intrusive
into
the
TCP
stack,
so
that's
now
being
worked
on
at
rat
had
to
make
it
less
intrusive.
So
we
do
hope
to
have
this
in
the
Linux
kernel
soon.
Q
Q
Q
Q
So
ESP
is
not
so
much
respect
between
well.
The
reason
we
put
different
layers
is
to
to
differentiate
the
different
steps
of
the
process
of
an
ESP
packet,
so
the
different,
the
compressions.
If
you
compress
the
inner
packet,
you
have
to
compress
it
before
encrypting
it
and
most
likely
before
ESP
fields
are
being
applied.
Q
The
description
of
the
compressions
is
provided
with
what
we
called
rules,
EHC
rules,
and
so
we
defined
a
number
of
rules
and
then
how
this
the
compression
occurs
is
a
coordination
of
all
these
rules
and
it's
what
we
call
the
eh-eh
strategy.
So
the
se
strategy
is
mostly
saying
or
if
we
are
in
this
situation,
I'm
gonna,
compress
this
field
is
fill
this
field,
which
means,
if
I
am
in
this
situation,
I'm
a
graph
come
I'm.
Gonna
apply
this
rule.
This
rule
and
this
rule,
so
that
is
not
in
some
cases
you
need.
Q
Q
So
you
have
a
field,
you
have
an
action
associate
to
the
field
and
some
parameters
so
that
basically,
how
we
describe
the
rule,
the
actions
it
should
be.
Well,
you
send
the
value
completely
or
you
just
remove
that
value
or
you
apply.
Well,
you
take
this
value.
You,
for
example,
you
you
just
send
the
least
significant
bit
of
that
value,
and
usually
the
compression
occurs
you
take.
The
value,
derive
the
complete
value
from
the
essay
or
from
yeah.
So
checksum,
you
don't
put
it
so
there
are
a
limited
number
of
factions
and
it's
pretty
obvious.
Q
Q
Q
Q
So
in
some
time,
for
example,
we
privilege
IP
disease
or
ipv6
over
ipv4,
so
some
time
we
had
to
make
some
choices,
but
in
most
cases
we
try
to
to
make
that
compression
useful
and
not
to
create
too
much
burden
and,
for
example,
ease
of
configurations
as
when
we
could
derive
fields
of
the
inner
packet
from
the
outer
packet.
That's
the
rule.
We
we
provided
so,
which
means
that
we
avoid
to
agree
on
each
fields
of
the
inner
packet
and
yeah.
Q
Q
Q
Q
Q
Think
it's
not
even
oh,
no
I'm,
not
sure
I'm,
including
the
the
user,
implicit
IV
or
not
I'm,
not
sure,
but
still,
if
you
use
that
for
a
traditional
VPN,
I
think
it's
I
previously
cannot
do
this
I,
don't
know
how
traditional
it
is,
but
you
still
have
a
32-bit
per
packet
compression,
so
still
a
significant
gain
and
yeah.
So
so
we
had
an
implementation
on
Kentucky
and
we've
run
that
over
so
I
think
it's
an
m3
device.
Q
So
the
the
funny
thing
is
that
so
we
we
we
had
different
type
of
compressions,
so
we
we
have
no
compression
at
all.
So
it's
in
red,
so
we
measure
the
number
of
bytes
you
can
send
and
the
corresponding
energy.
So
the
reason
you
have
a
stayer
is
because
usually
what
matters
is
the
number
of
radio
frame
you
are
sending
and
if
you're
sending
a
radio
frame
with
one
byte
or
a
radio
frame
that
is
full
I'm,
not
saying
the
number,
because
I'm
gonna
get
it
wrong,
it's
the
same
cast.
Q
So
basically
what
needs
to
be
optimized
is
the
number
of
radio
frames.
So
this
is
why
you
have
some
stairs
and
if
anything
is
that,
because
you
can
compress
the
inner
packet
it's,
you
can
send
more
data
by
using
a
die
TSP
then
not
encrypting,
the
data
and
actually
we
so
the
fooled
I
TSP
it's
a
little
bit,
but
it
only
applies
in
very
specific
cases.
But
if
you
have
a
reasonable
compression,
the
cost
is
very
close
to
clear
text
the
cost
associated
to
the
care
tax.
Q
So
we
we
use
a
notify,
payload
and
well.
Basically,
the
strategy
is
to
say
the
initiator
says
this
is
the
range
of
value
I'm,
accepting
the
other
ones
so
accepting
and
there
are
some
default
values.
So
unless
you,
if
you,
if
you
specify
nothing,
it
means
you're,
ready
to
accept
any
kind
of
value,
but
you
can't
provide
some
specific
size
of
value,
you're
ready
to
accept
on
specific
fields.
Q
Q
E
What
is
responsible
for
this
plus
so
I
assume
that
idea
to
compress
too
heater
compression
is
very
excited
and
well
I
like
it.
But
what
strikes
me
is
the
complexity
of
the
number
of
parameters
and
there's
a
complexity
of
negotiation
so
for
pure
per
user
or
poor
implementer.
It
I
think
it
will
be
not
easy
to
select
all
these
to
configure
these
parameters,
so
that
probably.
B
E
Q
E
Q
It's
half
lens
if
I
could
do
so
it's
true.
The
thing
is
that
we
took
a
lot
of
time
defining
the
rules,
but
it's
more
I
think
it.
The
code
should
be
lighter
than
the
descriptions,
because
the
I
mean
the
implementation
is
straightforward
and
the
rules
is
to
define
formally
what
we
do
and
to
structure
the
things,
but.
Q
Q
So
that's
an
yeah
yeah
I
mean
any
think
that
that's
a
goal
we
would
like
to
achieve.
I
think
I
had
we.
We
were
close
to
achieve
one
because
we
well,
the
intent
was
to
say
basically
say
what
you
want
and
anything
that
is,
you
don't
need
it's
gonna
take
a
default
value
and
the
reason
for
that
is
that
some
of
the
parameters
would
never
matter.
So
we
don't
want.
Then
we
don't
want
that.
The
end
user
configure
something
that
is
not
gonna
be
useful.
Q
So
that's
one
of
the
reason,
so
we
just
mentioned
configure
what
you
need
and
the
reminding
is
gonna
be
taking
and
default
values
or
never
gonna
be
used,
but
yeah.
So
that's
good,
you're,
more
focused
on
the
itv2
extension
sure,
okay,
yeah.
We
would
appreciate
that
feedback.
D
Parker's,
so,
even
though
you
keep
saying
that
this
is
really
simple
and
straightforward,
I
read
both
the
drafts
and
your
presentation
and
I
listened
to
it
and
I,
don't
understand
anything
so
I
think
it's
really
complicated.
Now
I
am
a
little
biased.
D
I
was
hoping
to
kill
IP
comb,
and
now
you
come
here
with
something
that
it's
like
much
more
complicated
and
I
become
paws
in
in
bells
and
whistles
and
features,
and
things
and
I'm
not
sure
if
it's
really
justifiable
to
say
well,
there's
a
few
battery-powered
devices
that
could
benefit
from
having
a
few
bytes
shaved
off
here
and
there,
and
especially-
and
if
you
start
talking
about
this,
has
to
be
simple
to
configure
for
the
end
user.
Well,
the
end
users
just
going
to
say:
oh
look,
I
can
now,
instead
of
once
per
five
minutes.
D
I
cannot
do
my
temperature
measurement
once
per
minute.
I
can
get
five
times
more
value
out
of
my
device
and
then
we've
got
all
of
this
complexity
and
and
I.
Don't
think
that
would
be
really
worth
it
in
the
end
for
the
protocol,
like
people
already
say
that
IPSec
and
I
are
really
complicated
and
I
think
this
just
adds
so
much
more
complication.
I
would
rather
have
us
work
on
removing
IP
compton
and
to
add
this.
Q
Yeah
we
so
the
thing
is
the
main
difference
between
rock
and
and
and
the
thing
is
that
you
have
no
symbol
signaling.
So
here
we
just
defined
rules,
which
means
you
don't
have
the
the
agreement
in
of
the
rules
within
the
data
pass.
You
don't
have
these
learning
phases
so
because
rock
well,
the
idea
of
rock
is
basically
you
have
a
learning
phase.
So
the
first
packet
is
unkind,
not
compressed,
and
you
have.
N
N
Use
do
during
IQ,
say:
hey
we're,
always
going
to
use
this
I
dressed
as
you
to
be
port,
and
then
you
always
use
them
good
sexually
and
then
the
second
one
about
like
battery
port
devices,
aren't
just
a
light.
Switch
like
my
phone
is
a
battery
powered
device.
My
watch
to
buy
a
powered
device
they're
all
performance
sensitive.
If
we
can
really
improve
the
throughput
on
this
and
the
battery
life
I
think
that's
really
important
for
users,
so
that
I
think
there
is
a
viable
use
case
for
this.
A
Q
H
Q
Can
configure
your
interface
with
and
I
think
4
or
9
36,
or
both
or
only
one,
but
you
have
to
choose
it's
a
trap.
So
then
the
address
family
failure
reasons
could
be
that
an
initiator
asked
for
one
ipv6,
which
is
not
supported,
in
which
case
the
message
should
be
unsupported,
address
family
and
you
have
a
case
of
an
initiator
that
requests
both
side
piece,
but
only
one
is
supported.
Q
You
specify
that
I
paid
before
only
supported
or
ipv6
only
supported,
and
you
have
the
case
where
an
active
for
N
and
ipv6
is
supported,
but
so
either
one
can
be
provided,
but
you
have
to
choose
one
I
think
the
use
case
is
really
two
3gpp.
So
that's
my
understanding
is
that
it
is
how
it
is
negotiated
with
the
radio
access
network,
and
so
they
want
to
bring
that
to
violence.
C
C
Q
C
A
So
they're
given
I
think
actually.
Yes,
there
is
too
many
other
cases
here.
I
think
we
can.
Actually.
We
have
I
think
we
already
have
that
particular
dress
families,
or
we
had
that
in
currently
already
there,
which
means
that
the
if
it's
if
it,
if
it
can't
give
you
anything,
you
should
fail.
Okay.
Now
the
question
is:
if
it
supports
both,
if
he
sends
both,
why
don't
you
just
return
one?
You
know
written
controller
parameters
for
if
you
asked
for
ipv4
and
ipv6
and
only
returned
ipv4
addresses.
A
You
know
that
it
doesn't
support
v6,
because
it
didn't
give
you
anything
and
the
same
thing
with
the
next
one,
but
the
last
one
is
the
different
one,
because
there
will
be
okay,
you
asked
both
ipv4
and
ipv6
and
you
could
get
either
one
of
them,
but
not
in
the
same
essay
so
or
not
in
the
piece
you
know
at
the
same
time,
so
the
single
AF
support
it
is
or
or
required
is.
Actually,
you
have
the
same
thing
with
the
you
know:
the
traffic
selectors
that
we
have
a
single.
A
K
Aren't
these
essentially
CG
PP
error
codes?
Yes,
so,
like
you
know
that
they
like
their
error
codes,
is
they
I
mean
they?
They
there's
all
these
elaborate
requirements
around
them.
You
know
so
it's
like!
Well,
if
you
do
this,
but
then
you
do
that
and
then,
like
this
other
flag,
is
set
and
these
the
bits
3
3
to
7
or
1,
then
you
need
to
return
this
error.
So
it's
is
there
a
reason
like
other
than
this
doesn't
make
a
lot
of
sense
at
all.
Is
it
actually
harmful
to
do
this
because.
A
Q
Q
Option
per
notify
fellow
I,
think
he's
looking
for
for
assignment
of
so
for
for
new
status
type,
so
I
I
think
there
are
some
objections.
That's
my
understanding
and
I
think
there
are
also
some
agree.
I
mean
and
I
also
think.
The
message
is
that
we
should
reduce
the
number
of
such
a
start.
That's
the
tier
and
well
one
question
I
had
is
am:
should
we
go
to
status,
type
or
I
say
we
better?
Q
A
I,
don't
think
so
that,
but
anyway,
so
we
are
running
out
of
time
now
and
I
think
that
you
would
actually
like
to
see
this.
Is
we
already
adopted
this
draft?
That's
working
to
document
errors
already
comments
on
that
saying
that
okay,
this
is
this
could
be
simplified
and
now
I
have
to
want
people
to
go
and
really
the
cane
and
sent
a
little
proposal
to
the
mailing
list,
and
you
know
solitaire
that
is
working
table
to
to
get
it
out
soon,
because
I
mean
what
are
the
problem.
A
Is
that
okay
tell
good
to
be
shipping
stuff
anyway,
and
they
ought
to
be?
They
actually
copying
our
stuff
in
their
their
standards
anyway.
So
if
we,
if
we,
if
you
are
not
ready,
they
will
put
some
numbers
there
anyway,
so
I,
that's
why
I
would
rather
them
to
put
our
numbers
and
our
meaning
of
them
than
to
invent
their
own.
One
more
comment
on
them
all
right,
okay,
so
anything
else.
That's
that's!
For
that.
I
think
we
are
going
to
be
getting
one
comment,
doesn't
something
else,
and
then
we
are
going
to
be
finishing.