►
From YouTube: IETF102-V6OPS-20180720-0930
Description
V6OPS meeting session at IETF102
2018/07/20 0930
https://datatracker.ietf.org/meeting/102/proceedings/
B
A
B
A
In
v6
UPS,
if
you
think
you're
somewhere
else,
you're
in
the
wrong
place,
we're
in
a
room
the
size
of
Mammoth,
Cave,
I,
wonder
if
we
could
consolidate
into
a
smaller
part
of
it
and
specifically
I
have
two
blue
sheets
and
I.
Don't
know
how
they're
getting
over
to
that
section.
So
that
needs
to
happen.
A
A
Specifically,
we
have
two
blue
sheets:
I
have
no
idea
how
they're
getting
over
to
that
area.
So,
okay,
now
and
of
course,
I've
got
one
chair
in
three
laptops,
but,
okay,
here's
what's
going
on
yesterday,
we
got
more
done
than
the
agenda
actually
had
planned
and
which
is
a
good
thing,
because
a
number
of
us
have
gone
home
overnight
or
are
on
the
way
out
the
door
this
morning.
Well,
we
have,
but
we
have
left
this
morning
are
essentially
two
AOB
drafts.
A
A
A
C
Okay,
hello,
sorry
for
those
have
seen
us
in
the
interior,
one
on
Monday.
We
are
aiming
for
interior
adoption.
So
if
you
think
it's
useful,
please
have
a
chat
on
their
mailing
lists
and
discusses
things
there.
So
my
name
is
Richard
Patterson
I
work
for
a
quad
play
operator
over
in
the
UK
called
CI.
We
have
around
6
million
broadband
subscribers
on
our
fixed-line
network,
most
of
those
all
of
those
at
dual
sac.
C
C
This
draft
attempts
to
add
additional
functionality
and
there
to
to
help
sort
out
some
of
the
operational
issues
that
we
still
experience
with
IP
OE
the
catalyst
for
writing.
This
draft
was
having
yet
another
discussion
where
they
appear
in
the
industry
around
the
benefits
of
IP,
o
e
versus
pppoe,
so
I
went
through
all
the
obvious
ones.
Is
less
encapsulation
its
simplify
connection
to
establish
most
fixed
line
providers.
C
These
days
prefer
to
use
port
based
authentication,
which
makes
chat,
username
passwords
superfluous,
and
then,
after
all
of
that
connection,
establishment,
you
still
need
to
run
PA
dhcpv6
over
the
top
of
that.
The
one
detriment
to
IPO.
A
that
I
couldn't
really
argue
against
is
it
lacks
a
session
state
tracking
mechanism.
C
As
mentioned
my
POA
clients,
don't
really
have
this
functionality
built
in
so
what
some
PNG
vendors
have
done
is
they've
implemented
their
own
method
of
connection
tracking
using
up
and
neighbor
discovery
checks
now
effect.
If
this
customer
was
to
unplug
their
CPE
the
beans,
you
would
realize
this
tear
down
the
session
clear
out
the
DHCP
lease
and
allow
a
new
CPE
to
to
be
reconnected.
C
So
in
a
broadband
network,
was
a
layer
to
a
Metro
access
network.
If
the
bng
was
to
fail
or
go
into
plans
maintenance,
what
that
Metro
access
network
was
to
be
the
backhaul
was
to
be
rehomed
to
a
different
PNG
or
there's
another
network
element
in
the
path
that
is
behaving
as
a
stateful
property,
HP
proxy
or
a
relay
that
could
lose
its
state.
Then
the
CPE
could
be
left
with
a
stale
DHCP
lease
for
up
to
the
valid
lifetime.
C
It
still
thinks
it
has
a
prefix
delegation
or
an
ipv4
address,
so
it'll
be
still
sending
packets
to
the
internet
via
the
PNG,
but
the
bng
will
just
drop
it
on
the
ground
because
it
has
no
idea
who
that
subscriber
is
now.
In
this
example.
The
lifetime
was
set
to
one
hour.
So
what
would
generally
happen
is
that
half
of
that
lifetime
so
30
minutes,
the
CPU
will
send
the
dhp
Renu
to
the
look
to
the
previous
DHCP
server.
The
bng,
let's
try
and
renew
that
lease
the
PNG
has
no
idea
of
this
lease
anymore.
C
So
it
drops
that
renewal
on
the
ground.
The
CPU,
will
keep
trying
and
keep
trying
to
send
this
release.
I've
seen
the
renew
and
the
bins.
You
will
keep
dropping
it
until
such
time
as
that
valid
lifetime
completely
expires,
and
then
the
IPO,
a
the
DHCP
client
will
restart
the
discovery
and
the
solicit
phase.
The
Benjy
can
then
Rio
indicate
and
the
subscriber
comes
online.
C
Now,
all
these
parameters
can
should
be
signaled
via
DHCP
within
the
lease
itself,
so
we
specify
new
options
for
DHCP
v4
and
HP
v6
and
there
within
the
lease,
so
a
CPA
or
IP.
A
client
may
have
multiple
leases
with
multiple
health
check
parameters
on
going
at
once,
or
you
could
locally
configure
these
on
the
CP
itself.
There
could
be
a
default
configuration,
a
user
could
override
them
via
the
user
interface
or
we
can
push
it
down
by
a
TR,
69
or
Nick
on
the.
C
Drafts
this
fires,
three
different
types
of
health
checks.
Now,
first,
one
is
the
preferred
one:
we've
called
a
PFD
echo,
but
after
a
chat
with
Jeff
has
this
week,
we'll
probably
have
to
rename
this
a
little
bit,
because
it's
not
probably
won't
be
full
BFT,
so
50-acre
packets
style
would
probably
be
will
be
lightweight
it's
IP
based,
so
it
will
use
the
forwarding
plane
of
of
the
bng.
C
So
it
shouldn't
impact
the
control
plane,
and
this
is
something
that
the
ARP
neighbor
discovery
or
even
just
DHCP
renews
themselves
will
impact
on
the
BG
otherwise,
and
it's
also
a
type
II
based.
So
it
has
the
benefit
of
checking
the
the
lease
is
actually
still
valid
on
the
bng
itself,
as
a
banty
would
not
forward
the
IP
packet
back
to
you
if
it
had
no
lease.
C
If,
for
whatever
reason
the
IP
OE
clients
cannot
use
the
s--
PFD
echo
style
probe,
then
it
can
fall
back
to
our
poor
neighbor
discovery.
This
upper
level
discovery
will
be
for
the
default
gateway
which
would
be
residing
on
the
PNG,
and
this
could
also
be
forced
in
the
words
that
are
with
that
l2
flag.
Before
and
likewise,
if
my
pov
clients
still
cannot
use
our
poor
neighbor
discovery
for
whatever
reason
or
it
is
forced
to
use
the
passive
flag,
then
it
could
not
fall
back
to
just
passively
monitoring.
The
are
poor.
C
Indeed,
cache
tables
on
the
device
itself-
probably
quite
a
lot
there.
So
as
Lawrence
I
mentioned
on
Monday,
perhaps
maybe
we
can
find
tudents
and
and
remove
some
of
the
additional
options
that
we
think
were
nice
to
have,
but
make
it
a
bit
overly
complex
the
actions
to
take.
Then
we
define
for
in
this
current
document.
So
the
first
one
which
is
the
default
is
the
renew.
It
would
be
considered.
Probably
the
softest
approach
of
the
four.
C
C
The
next
option
would
be
route
bind,
and
this
would
skip
the
renew
phase
it
would
in
multicast
or
broadcast
these
rewinds
messages
to
any
available.
B
ng
or
DHCP
server
within
that
layer
to
access
again
it
still
has
the
same
issues
as
before.
Where
you're,
not
you,
may
not
be
able
to
REO
indicate
subscriber,
but
it
could
help
and
topology
is
where
there
are
multiple
servers.
C
The
third
option
is
solicit,
and
this
is
the
one
that
I
see,
probably
that
we'll
be
using
on
our
network,
because
this
one
skips
the
previous
two
phases
move
straight
to
the
solicit
and
discovery
phase,
allowing
that
PNG
to
re-authenticate
the
subscriber
and
premium
online.
It
doesn't
explicitly
expire
the
lease,
so
the
lifetime
for
the
lease
will
still
be
present,
which
means
the
CPU
won't
withdraw
any
locally
bound
roots
on
the
LAN
or
anything.
So
it
should
hopefully
facilitate
a
renew
of
the
lease
without
actually
having
to
renumber
as
well.
C
Ideally,
the
fourth
of
final
approach
would
be
what
I
call
the
scorched
earth
approach.
That's
explicitly
expiring
the
lifetime
of
the
lease
completely
withdrawing
the
lease
from
the
local
device
and
sending
a
DHCP
release
through
the
network
to
clear
out
any
potentially
stale
state
live
in
the
network
that
could
be
preventing
that
client
from
coming
back
online
and
that's
it.
A
D
D
There
was
I
pee
over
Ethernet,
which
is
connectionless,
and
then
there
were
modems
that
you
had
to
dial
up
and
you
had
to
make
a
session
and
if
it
failed,
you
had
to
hang
up,
and
you
had
to
redial
and
30
years
ago
was
the
battle
between
connection-oriented
networking,
ATM
and
virtual
circuits
and
connection
less
an
Ethernet
is
connection
less
and
if
you
don't
have
a
connection,
you
can't
disconnect
there's
one
of
the
great
benefits
and
white
packet
switching
one.
So
all
this
mechanism
to
manage
connections
over
an
inherently
connection,
less
technology,
seems
our
peace.
D
Emissary
odd.
To
me,
you
talked
about
cases
where
your
devices,
using
our
poor
neighbor
discovery
to
tell
if
the
client
is
still
responding
and
reclaiming
its
DHCP
lease.
That
is
an
explicit
violation
of
the
deep
in
protocol.
There
are
no
liveness
requirements,
so
if
you
reclaim
a
lease
that
you've
given
to
a
client
then
and
give
that
address
to
a
different
client,
there's
a
reason.
The
DHCP
spec
disallows
that-
and
this
is
the
reason
so.
C
I'll
pause
right
there,
I
didn't,
say
it:
reclaims
lists
it'll,
clear
out
the
lease
and
it
could
still
keep
a
binding.
Well,
a
curb,
a
note
within
its
local
table.
For
this
clients.
That's
no
longer
there
had
this
lease,
and
so
it
could
reserve
it
for
when
that
client
comes
back
online
and
that's
what
most
of
them
seem
to
do
so.
D
So
I
think
the
real
question
is,
it
seems
like
an
or
thought
of
work
was
done
to
create
this
problem
so
that
we
could
do
more
work
to
fix
it.
If
you
just
don't
do
that,
if
you
just
don't
care
whether
the
clients
there,
if
you
don't
expire
or
retire,
whatever
terminology
you
want,
if
you
give
it
a
two-hour
lease,
just
let
it
be
a
two-hour
lease,
do
not
kill
it
prematurely,
and
then
you
don't
have
this
problem
in
the
first
place.
So
yeah
there's
there's
a
couple
of
things.
C
D
C
This
is
a
fixed
line,
broadband
network
will
we
need
to
authenticate
a
subscriber
for
data
retention
and
compliance
reasons.
We
need
to
keep
track
of
their
radius
accounting
records
to
make
sure
that
we
know
that
that
subscriber
had
the
site
he
address
or
prefix
at
this
given
time.
So
when
the
le
a
has
come
to
us
so.
D
I
guess
I
had
two
other
points.
One
is
the
way
the
ITF
works
is
rough
consensus
and
running
code.
If
this
is
for
clients
of
IP
over
Ethernet,
that
sounds
like
Linux
Mac
Windows,
which
vendors
have
signed
up
to
implement
this,
because
with
no
vendor
support
you
can
write
an
RFC,
it
doesn't
really
matter
so
and
then
the
final
comment
is:
if
you
want
some
kind
of
liveness
check,
if
that's
what
you
really
want
is
to
be
hitting
the
client.
Are
you
there
are
you
there?
Are
you
still
there
I?
C
D
D
C
D
C
Have
our
own
in-house
CP
development
team
we
will
be
building
us.
France
télécom
also
have
this
issue
they've
implemented
something
similar
before
this.
So
we've
had
a
lot
of
feedback
from
mid
to
try
and
ratify
this
focus
in
New
Zealand.
They
don't
have
their
own
CP
in
Australia.
They
don't
have
their
own
CP
development
team.
They
have
this
issue
on
their
network.
They
can't
convince
the
vendors
to
implement
something
like
this,
because
there's
no
document
to
reference.
C
D
Which
client
devices
do
you
imagine
this
running
on,
because
I
was
I
was
talking
about
Mac
Linux
windows
and
your
bookmarks
EPS
I'm
talking
about
cities
as
a
fixed
line,
broadband
problem?
Okay,
so
this
is
not
Ethernet
within
the
home.
No.
This
is
this.
Is
the
Ethernet
cable
between
the
CPE
and
Sky
Chris,
yep,
okay,
and
that
and
that
is
Ethernet?
That's
not
DSL!
That's
not
coax!
That's
actually
Ethernet!
It's
video
cell,
which
season
it
framed
its
Jeep
on
also.
E
So
Barbara
stark
AMT.
This
has
been
a
problem
since
we
started
doing
IP
over
Ethernet
20
years
ago.
The
is
this
is
not
a
new
problem
and
it
has
been
a
problem
and
the
BFD
echo
that
he
pointed
out
is
what
everybody
pretty
much
decided
to
do
and
it
is
implemented
in
most
every
residential
gateway
and
bng.
That's
out
there.
C
E
So
I
guess
my
question
is:
does
the
BSD
echo
in
fact
nature
need
and
if
it
meets
your
need,
I
mean
the
broadband
formas,
as
you
noted,
in
TR,
146
and
in
other
places
and
in
the
residential
gateway
requirements,
actually
has
the
BFG
echos
sections?
What
do
you
really
need
for
IETF
to
document?
That's
not
already
there
in
a
place.
That's
much
friendlier
to
service
providers.
I'll,
say.
C
C
What's
the
other
part
of
the
questions
right,
there
was:
oh
yes,
or
what
else
is
this,
so
the
signalling
of
the
the
actual
parameters
themselves
via
HP
lease
so
that
we
don't
have
to
have
something
manually
configured
or
run
on
default
so
that
that's
signaling
between
the
network
and
the
CPE
I
think
will
also
be
quite
useful
and,
and
probably
the
main
one
is
the
behavior
types.
So
whilst
here
a
one
four
six
specifies
the
health
check
mechanism
would
prefer
to
use
PFD
echo
style.
C
F
Lorenza
community
I
think
you
should
just
write
down
that.
That's
a
bug
and
needs
to
be
fixed
right.
Then
then,
basically,
you've
got
everything.
You've
got
Tier
one
four
six,
which
is
presumably
a
bunch
of
people
already
have
implementations
of.
If
you
don't,
then
you
can
add
an
implementation
of
it
right
and.
F
Yeah,
like
I,
said,
like
I,
said
in
in
Terry
I
mean
you
used.
I
think
you
just
have
too
many
options
here
and
the
more
options
you
have:
the
lower
quality
of
the
U
at
power
of
like
implementation
resources,
the
lower
quality
of
the
implantation
and
they
wide
a
variety
of
bugs
that
you
have
to
do.
I,
don't
think
it's
good
for
you!
I
I
did
want
to
circle
back
to
Stuart's
point:
hey
you're,
saying
that
you
know
five-second
five-minute.
F
We
used
don't
scale,
but
it's
not
clear
to
me
why,
in
the
sense
that
you
know
if
you're,
if
you're
doing
it,
if
you're
using
bhp,
that's
just
sent
up
as
a
relay
right
or
if
the
bng
is
a
DHCP
server,
then
I
think
if
you're
saying
five
minute,
things
won't
scale,
but
then
you're
saying
that
we're
gonna
use
mud,
nubs
expensive.
If
it's
on
the
CPU,
so
so.
C
C
I
mean
there
are
providers
that
I've
heard
of
that
I
think
focus
and
also
mentioned
that
they
they
use
a
DHCP
least
time
of
like
five
minutes
of
10
minutes,
so
that
it
mitigates
this
problem
somewhat
because
they
not
doing
any
state
checking
but
it'll
still
come
up
within
yeah
five
minutes
and
that's
better
than
say
half
an
hour
or
an
hour.
We
couldn't
really
scale
much
much
quicker
than
an
hour
on
some
of
our
platforms
without
it
actually
impacting
the
canal
and
CPU
and
control
plane.
C
F
C
F
C
You
have
to
manage
your
layer
to
domain
a
bit
smarter,
so
this
could
be
a
simple
easy.
Well,
it
could
be
an
e
VPN.
It
could
be
point-to-point
Sutor
wires.
It
could
be
with
you
know
your
handling
with
fast
reroute
and
the
suit
was
coming
in
different
interfaces.
You've
then
got
even
further
problems
where,
but.
C
F
F
F
C
You
know
multi
yeah,
multi
homes,
sort
of
multi,
beam,
G
layer,
topology
sure
this.
This
isn't
an
issue
if
it's
just
a
temporarily
path
failure,
and
then
the
Bengie
still
keeps
the
state.
If
the
past
comes
back
up,
then
it's
not
a
problem.
You
only
unlike
PPP
the
see.
If
you
won't
even
know
there
was
problem,
it
just
dropped
some
packets
and
then
carried
on
forwarding.
I.
C
Think
your
points
on
this
I
think
are
perfectly
valid,
but
don't
think
it
solves
the
problem
that
it
would
still
require
the
ditch
police
to
be
quite
short,
so
this
is
still
generally
half
of
the
least
time,
so
it
would
have
to
be
if
we
fix
this
problem
and
I
still
have
one
hour.
Lisa's
CPU
could
still
be
down
for
up
to
a
half
an
hour.
F
Sounds
like
you're,
trusting
the
client
to
do
the
right
thing,
though
I
wouldn't
I
wouldn't
design
it
like
that.
If
I
were
you
even
if
you
write
the
CPU
software
it
you
know
those
things,
the
things
on
the
right
are.
Basically,
if
you,
if
the
thing
on
the
left
is
able
to
corrupt
the
state
of
this
thing
on
the
right,
then
that
thing
is
serving
hundreds
of
thousands
of
users,
and
you
don't
want
that
to
be
unhappy,
that
that's
be
my
advice,
but
yes,
so
I'm
not
gonna.
F
C
C
I
think
the
point
is
that
neither
this
bi-directional
checking
hasn't
existed.
Some
BMG's
have
implemented
a
unidirectional
check
which
actually
exacerbates
the
problem
from
a
CP
point
of
view,
because
now
the
beam
G
will
clear
the
state
even
quicker,
then
say
one
hour
less
time,
so
being
G
is
trying
to
do
the
right
thing
in
that
direction
have
made
it
worse
in
the
upstream
direction.
So
all
I'm
trying
to
do
is
get
from
that
unidirectional
state
to
that
bi-directional
state,
so
I'm
just
trying
to
implement
what
PPA
has
done
from
both
sides.
H
Try
to
remove
the
mic
methods
working
anyway,
so
for
the
people
that
were
interim,
the
slides
will
be
different
because
I'll
restart
from
scratch.
So
you
will
get
explanation
about
what
the
PVD
is
and
what
is
the
problem
statement
that
we
are
trying
to
solve.
So
my
name
is
Eric.
Mink
I
am
one
of
the
fifth,
the
five
authors
and
many
of
them
he
appear
is
iam
Tommy's.
Here
DeWitt
is
maybe
in
the
building
somewhere
and
when
she
is
back
in
Paris.
H
So
when
the
internet
started
and
we've
seen
a
previous
slide,
we
looks
like
we
had
a
single
connection,
but
is
no
more
the
case.
If
using
right
now
on
your
laptop,
you
must
have
EF
Wi-Fi
at
home.
You
have
wide
as
well,
and
you
may
even
have
using
bluetooth
to
Tetra
with
a
mobile
phone,
so
you
have
at
least
like
on
this
slide
three
interfaces.
Now
on
some
interfaces,
such
as
the
Wi-Fi.
You
have
maybe
two
upstream
routers.
H
So
basically,
here
you
can
go
out
through
the
mobile
service
provider
is
P,
1
is
P
2
and
VPN,
and
you
get
three
physical
interface
honestly
I
have
only
one
is
P
at
home
at
all
the
rest,
I
F
so
I
have
this
problem
sometimes
two
times
at
home,
so
now
train
a
few
are
connected
and
but
well,
how
do
you
route
this?
For
instance,
assuming
you
select
the
wide
interface
which
interface
you
go
out
ice
p1i,
speed,
routing,
well,
tell
you
which
address
you
use
in
the
case
of
ipv4
the
legacy
protocol
right.
H
We
have
only
an
IP
address,
so
if
you
exit
to
is
p1
using
an
address
from,
is
p2
and
is
doing
anti-spoofing
as
you
should
do,
it
should
be
blocked
so
combining
times
making
press
the
only
way
to
make
is
working
is
to
use
the
n-word.
Not
you
do
not.
If
you
exit
to
is
p1
by
using
a
nice
p1
global
address
and
then
the
traffic
is
coming
back
good,
but
we
are
in
v6
ops,
not
envy
Lagasse
ops.
So
how
do
we
do
it
there?
H
Let's
one
torrent,
SEC
Tournament,
to
do
easy,
min
filming
again,
large
cooperation
do
BGP
peering,
you
use
your
own
P
eyespace,
not
a
problem.
Talking
about
P
I,
here
right
now,
boss
for
SMB,
multiple
draft
I
tend
to
use
this.
That's
what
I
say:
pa
OPA
address.
That's
fine,
gen
and
address
has
got
draft
as
well
to
select
which
router
to
exit.
They
are
part
of
the
solution.
I'm
at
what
can
we
do
more?
H
Another
issue
was
explained.
My
one
previous
author
Markus,
keen
at
that
point
of
time,
was
working
for
Microsoft
IT
and,
if
show
me,
this
slide
showed
us
this
slide
the
ITF
99,
but
a
year
ago
I
am
basically
on
the
left
hand
side.
This
is
Microsoft
corporate
IT,
where
they
got
a
different
prefix
in
us
compared
to
Europe
one
airing
and
one
ripe,
prefix.
H
There's
a
blue
greenish
cloud
on
the
left,
but
they
want
to
look
at
breakout
so
meaning
if
they
want
the
traffic
to
go
to
the
Internet,
they
need
to
use
local
address
and
not
go
to
the
corporate
network.
So
that's
the
blue
and
the
blue
is
giving
them
and
a
prefix
like
2001,
bb-8,
obviously
the
example
and
as
well
as
they
want
to
go
to
Azir,
but,
as
here
in
europe,
is
using
part
of
the
prefix
of
microsoft,
ripe,
prefix
right
I
mean
it
does
make
sense.
H
If
you
want
to
do
an
addressing
plan
now,
people
at
this
office,
the
pc
at
the
bottom
basically
receive
to
prefix
by
using
erase
one
from
corporate
EMEA
and
one
from
the
local
ISP
look
at
this.
We
know
that
the
saucer
dress
selection
will
only
always
prefer
the
longest
prefix
match.
If
you
don't
cheat
with
it.
So
I
want
to
go
to
the
azure
in
the
cloud
which
address
I
will
select.
Oh
I
receive
a
to
a
zero
one
address
from
my
layer:
3
VPN,
the
greenish
part.
H
So
if
I
go
directly
to
us
you're
on
the
right,
the
ice
beam
blue
will
do
BCP
to
t8.
Anti-Spoofing
will
drop
the
packet
right
bad
now.
I
want
to
go
to
say
a
bookkeeping
website
to
manage
my
vacation,
which
is
on
the
green
left
cloud.
So
2001,
which
address
do
I
select
the
ISP
address,
which
is
2001
the
longest
prefix
match
so
I
go
out,
maybe
by
the
left
part
because
I'm
rotted
there
to
the
right
destination,
but
the
reply
will
be
sent
through
the
ISP
right.
H
So
obviously
it
will
not
fly
that
lemon
has
got
another
if
you
had
the
same
IETF
in
a
different
group.
If
you
have
to
interface,
for
instance,
on
my
phone
here,
I'm
from
Belgium,
if
I
ask
the
address
of
a
newspaper
over
the
Silla
link,
it
will
exit
in
Brussels
and
I
will
go
to
Mecca.
My
website
will
close
to
Brussels
and
now,
if
I
use
for
the
data,
the
wide
from
the
ITF
I
will
travel
back
to
research,
which
is
not
optimized,
but
even
worse.
H
If
repair
interface
working
group
on
RC,
75
56,
they
define
provisioning
domain
provisioning
domain
is
a
consistent
set
of
network
information.
As
I
said
interface
address
next
stop
route
cause,
you
can
get
sits
with
specific
routes,
DNS
servers
and
so
on
and
so
on.
So
all
draft
has
come
for
goal.
Oh
no
identify
those
PVD
because
you
can
get
multiple
of
them
at
least
one
per
interface
in
a
PVCs.
You
can
get
multiple
routers
and
multiple
prefix
or
multiple
addresses.
H
So
in
the
case
of
v6,
multiple
PVD
per
interface,
as
well
as
giving
you
more
information
about
the
network
itself,
it's
already
good
for
connectivity
to
have
a
bender.
So
you
don't
get
those
issue
mentioned
before,
but
you
need
to
know.
Oh
I
can
select
this
specific
ipv6
address
compared
to
the
other
one.
Okay,
all
we
do
it
so
with
the
draft
is
revision
zero
right
now
there
was
it's
a
routine
trip
item
in
inter
area
before
it
was
personal
draft.
Basically,
what
we
do,
we
add
a
PVD
option
into
the
router
advertisement.
H
We
have
a
typology
allocated
by
an
ax
which
is
twenty
one
couple
of
bits.
I
shall
come
back
on
this.
The
main
thing
in
the
middle
is
the
PVD
ID,
fully
qualified
domain
name.
We
use
a
fully
qualified
domain
name
for
two
reason:
one
we
can
be
sure
they
can
be
unique.
Like
free,
qualified
domain
name
can
be
unique,
even
if
they
refer
to
multiple
servers
as
well
as
we
will
use
it
to
access
more
information
by
forming
an
ear
I
based
on
this
relief,
a
qualified
domain
name.
H
So
two
purpose
here
then,
after
we
may
repeat
a
router
advertisement
in
this
option,
I
shall
explain
to
you
later
why
some
flags
h,
it
means
HTTP.
It
means
please
do
make
a
new
ride
based
on
the
fully
qualified
domain
name.
Is
there
to
fetch
more
information?
If
it's
not
there,
don't
do
it.
You,
the
flag,
is
there
you
may
do
it
L
means
for
legacy
ipv4
we
can.
We
are
not,
if
I
with
this
PVD
all
options
into
the
array.
H
But
what,
if
you
are,
you
are
stuck
what
about
the
ipv4
information
that
I'm
receiving
by
the
HCP
before?
If
you
put
the
bit
L
it
mean
that
this
PVD
is
associated
as
well
with
the
D
ipv4
information
of
you
see,
you
can
only
get
one
array
with
one
PVD
with
Al
bit.
It's
up
to
the
operator.
To
do
it
correctly
and
arm
means
that
typically,
there
is
a
repeated
rotting
advertisement
below
explaining
this
later
sequence.
Number.
What's
this,
it's
simply
to
tell
you:
hey
I
have
changed
the
PVD
somewhere
I
have
changed
the
information.
H
How
do
I
notice
is
a
different
one?
It's
a
new
one.
That's
when
we
change
the
sequence
number,
then
it
can
trigger
a
refresh
of
all
information
in
only
oath.
Of
course,
if
you
have
a
large
conference
with
twenty
thousand
people,
let's
say
or
thirty
thousand
on
the
Wi-Fi,
you
do
not
want
everyone
refreshing.
H
The
information
at
the
same
time
fetching
the
additional
information
of
SSL
right,
because
this
SL
server
will
be
a
little
bit
overloaded,
there's
a
reason
why
we
put
in
the
top
right
corner
delay,
which
is
an
exponential
delaying,
looked
into
the
draft
to
see,
but
is
basically
something
between
one
seconds
and
fifty
days,
which
is
random.
So
you
delay
a
random
time
between
now
and
up
to
50
days
to
refresher
information,
okay
and
by
the
way
we
are
here
to
a
get
a
personal
feedback,
so
feel
free
to
ask
questions
and
good
comment.
H
H
So
no
additional
information
and
no
rotting
at
a
repeat
it
and
nothing
to
do
with
ipv4
sequence
number
could
have
put
any
number
there
and
then
we
express
the
fully
qualified
domain
name
in
the
way.
Dns
requests
are
written,
so
it's
not
with
dot
in
ascii.
It's
really
always
ask
for
a
error
or
a
or
whatever
under
dns,
some
padding,
and
now
we
have
insert
it
later
to
normal
array
options,
one
for
the
DNS
server
and
one
for
Pio
for
some
address
and
if
you
look
about
the
length
length
is
12
now
look
upon
this.
H
If
I
am
a
PVD
aware,
oast
I
know
how
to
handle
those
options
right
right
now
there
is
none
except
some
experimental
code
on
Linux.
I
know
that
it's
a
PVD
I
know
the
length
on
the
structure.
So
I
know
exactly
where
the
sequence
number
is.
I
know
the
length
of
the
DNS
field,
the
fully
qualified
domain
name,
because
the
structure
I
know
there
is
some
padding
to
align
on
the
eight
byte
boundary
and
then
I
continue
the
passing
of
the
recursive
DNS
server
and
the
PIO
I.
H
H
Now,
if
I
am
a
non
PVD
aware,
oast
notice
an
option,
I
have
no
clue
all
to
pass
it,
but
I
skip
it.
So
I
used
the
length
and
go
directly
after
I.
Don't
look
what's
inside
okay,
so
if
we
are
sending
now
pivot
in
all
our
oast
here,
you
will
ignore
completely
the
PIO.
So
you
don't
use
an
address
from
this
one
and
you
will
not
use
a
DNS
server,
so
it
means
that
we
can
deploy
PVD
they
to
ed
everywhere.
It
will
break
nothing.
You
bring.
H
H
Ok
so
think
about
it
very
simple
case:
you
use
one
array
with
a
single
option
in
it,
the
PVD
that
includes
Pio
and
DNS.
So
if
I
am
a
PVD
aware,
oast
I
will
use
this
address
and
this
DNS
server
if
I
am
normalized
I
will
ignore
it
because
there
is
no
address
inside.
Ok,
you
have
put
on
the
outside
enter
them
in
the
normal
routing
header
rotting
advertisement
had
a
lifetime
zero
for
the
router.
H
We
have
one
use
case
with
a
big
service
provider
in
Europe
that
want
to
use
this
to
advertise
a
specific
Pio
and
specific
route
for
the
TV,
for
instance
the
setup
box,
because
them
there
will
be
P
video
where
they
will
accept
the
PVD,
fully
qualified
domain
name
and
basically
use
this
Pio
and
this
exit
route
a
very
clean
way
of
doing
a
walled
garden.
Ok,
if
you
remember
what
was
happening
in
Japan
years
ago,
with
the
mix-up
of
the
wall
garden
and
an
ISP,
we
avoid
this
and
multiple.
H
At
the
use
case
now,
remember,
I
said
there
is
an
H
bit.
If
there
is
the
H
bit
is
there?
The
node
may
have
a
case
may
form
this
year,
I,
yet
GDPs,
slash,
slash
free
qualified
domain
name,
dot,
well-known
slash,
DVD
and
basically
use
HTTP
or
TLS
right
to
access
it.
What's
there
inside
this,
a
JSON
file
where
we
repeat
the
name
that
could
be
displayed
in
a
better
way
that
a
fully
qualified
domain
name,
that's
optional.
H
There
is
an
expired
date
pretty
much
like
expires
on
any
HTTP
content,
so
you
should
refresh
it
before
this
time
and
again
we
do
some
randomization
there
to
avoid
a
wee
one
fetching
at
the
same
time,
the
prefixes,
which
is
there.
This
is
the
list
of
prefix,
typically
slash
14/32,
that
are
allowed
to
get
this
information.
It
basically
I
have
many
more
signs
regarding
the
privacy
and
the
security
of
it.
I
don't
present
in
here.
We
can
do
it
offline.
H
If
you
want
it's
to
raise,
the
level
of
security
of
this
can
still
be
spoof
right.
It
relies
on
router
detachment
that
can
be
spoofed,
so
nothing
really
magic
there,
but
you
try
to
raise
the
bar
now
inside
them.
This
additional
information
beside
those
three
information
attained.
The
draft
we
have
at
least
another
draft
in
captive
portal.
This
specifies
this
key
captive
API
and
that's
a
draft
from
the
two
gentlemen
there
to
me
and
peer
in
short,
to
summarize
it
captive
portal.
H
One
issue
is
that
the
redirect
that
we
all
know
in
the
hotel,
in
the
station,
in
the
train
and
so
on,
does
not
always
work
because
of
TLS
and
HSTs,
so
they
need
to
find
something
else.
So
getting
the
URI
where
you
need
to
connect
one
to
username
and
password
or
credit
card
number
and
blah
blah,
it's
very
important
for
them.
Okay,
and
we
can
get
other.
We
were
talking
about
the
people
I'm,
not
providing
information
as
well
about
the
upstream
connectivity
right.
You
can
send
very
easily
or
fast.
H
Your
Wi-Fi
link
is,
but
you
don't
know
the
speed
of
the
bandwidth
between
your
access
point
and
the
access
network
could
be
all
two
megabit
per
second
line
could
be
a
fast
it
and
could
be
a
gigabit,
so
just
may
be
important
for
your
application.
The
goal
of
this
JSON
file,
just
to
be
clear,
is
not
to
change
the
layer
tree.
H
The
layer
tree
is
communicated
by
the
router
of
attachment
is
to
provide
additional
information
to
application,
and
we
sincerely
believe
it
could
be
very
useful
for
many
many
use
case,
because
then
network
can,
with
some
sense
of
security,
but
not
full
security,
again
communicate
to
the
application.
What
are
the
services?
What
are
the
characteristic?
What
are
the
properties
if
you
use
this
Pio
that
could
be
routed
differently
and
this
interface?
It
could
be
routed
differently.
H
We
can
receive
as
well
to
pvd's
with
the
same
name
from
two
interface.
Why
not?
Okay
could
be
considered
sensible
exactly
the
same,
for
instance.
So,
as
I
said,
we
are
making
big
progress
on
this.
We've
got
in
spring,
the
the
number
from
Ayane.
So
that's
this
original,
but
in
future,
as
long
as
you
don't
get
an
RFC
its
return
in
future,
but
its
options
21
implementation
status.
If
you
go
and
get
app,
you
will
get
many
many
implementation
so
sending
this
array
with
those
bits
and
this
structure
is
kind
of
easy.
H
So
we
have
modified
array
DVD,
typically
for
Linux
and
or
DHCP
as
well,
which
is
also
used
in
open
wrt
and
working
for
cisco.
We
also
add
some
extra
metal
code
in
iOS
there,
but
is
very
simple
right
is
sending
an
array
with
with
additional
bits
on
the
receiver
side.
We
have
another
easy
one,
which
is
why
shark,
so
we
can
decode
all
the
options
there,
which
kind
of
an
easy
one
which
is
more
complex.
Is
we
are
working
for
many
months
on
open
source
implementation
of
the
landings?
H
It's
not
so
much
about
getting
the
array
and
getting
the
options
right.
I
mean
that's
easy,
it's
more
being
able
to
make
a
bindle
in
the
Linux
code,
independent
of
the
array
of
the
PVD,
so
linking
the
route.
The
address
the
DNS
server,
which
is
in
the
user
space
in
Linux
right,
is
in
Lib
C.
So
it's
working
now,
so
we
can
download
the
code,
and
basically
you
are
more
than
happy
and
twelve.
Today,
for
instance,
in
in
London
we
had
a
curtain
with
all
those
people
over
there.
H
Kapoor
dalip
Allah
here
are
basically
two
testing
the
captive
Porter
with
the
ladies
code
there
and
it
was
running
fine,
okay
and
basically
so
I
came
here
to
present
to
you
so
the
the
intend
of
this
draft.
We
are
pretty
good
shape
in
Korea
our
intend.
We
received
some
comments
to
change,
making
the
wording
to
be
clear.
For
instance,
what
do
we
do
when
we
debtor
a
phone
to
the
rest?
Obviously
we
should
propagate
the
PVD
option
on
the
to
detect.
H
Phones
do
the
different
clients,
the
similar
thing,
nothing
really
dramatic,
and
we
intend
to
go
on
last
call
after
a
Bangkok.
I
still
need
to
present
this
to
six
men
because
we
had
we
changed
the
ipv6
protocol,
but
it
wasn't
month,
which
will
be
too
short.
Fred
was
kind
enough
to
accept
me
here
so
comments
review
as.
H
Whatever
we
want
to
commence,
it's
better,
if
you
put
in
theory
and
CC,
of
course,
but
I
know
the
pain
as
well,
but
you
don't
follow
a
working
group.
You
don't
want
to
subscribe.
Just
for
one
thing:
I
am
following
the
six:
absolutely
I
am
for
the
main
six
man
I
am
for
during
interior
post
it,
whoever
you
want.
Okay,
that's
basically
it
or
go
to
the
mic,
or
we
have
basically
one
hour
and
a
half,
so
we're
attorney
to
go
bit
or
around
the
beard.
Oh
yeah,.
A
I
Hello,
Tom,
Jones,
University
of
Aberdeen
I,
say
my
name
there,
so
we
use
the
the
PVD
information
in
the
the
e
uni
project,
which
is
a
project
to
build
the
next
generation
replacement
for
the
socket
API,
and
this
sort
of
information
that
could
be
signal
from
the
network
turned
out
to
be
really
powerful
to
use
and
it
allowed
us
to
build
applications
which
could
make
decisions
about
the
network
at
a
high
level
on
the
network
could
provide
a
reasonable
level
of
guarantee
and
I
would
say,
say
you
know
any
operators
that
this
is.
I
J
Eric
Nygren,
well
I'm.
One
thing
that
might
be
worth
looking
at
our
kind
of
looping
in
with
is
the
discussion
that
happened
in
DRI
you
earlier
this
or
yesterday.
Whether
look
because
one
of
the
things
it's
being
discussed,
there
is
in
particular
on
DNS
server
discovery,
and
there
is
a
discussion
on
when
do
you
trust
the
network,
if
you're
going
to
be
using
DNS
privacy
or
denotes
over
TLS?
When
can
you
trust
the
network
or
dhcpv6
to
go
and
give
you
a
here's?
J
K
H
K
H
K
Think
that's
actually
two
different
things.
I
think
creating
an
eye
on
a
registry
is
one
thing
mm-hm
and
the
means
by
which
you
update.
It
is
a
separate
thing.
Is
this
true?
One
possible
thing
is
that
every
new
option
needs
a
new
document
and
the
other
option
is
everything
needs
like
well,
it's
in
a
new
document,
but
it
doesn't
have
to
be
I
mean
it
needs
a
expert
review.
The
way
DHC
used
to
run
a
run.
K
L
Generator
I
really
like
it,
because
from
experience,
for
example,
at
ATF
and
v6
only
network,
the
main
issue,
I've
seen,
was
DNS
split
horizon
between
VPNs
and
v6
on
your
network.
If
you
are
v6
on
the
network
and
you
own
the
VPN,
even
if
it
works
I
mean
getting
established,
then
you
either
start
using
your
corporate
DNS
and
everything
like
breaks,
because
it's
not
dns64
years
so
basically
yeah.
We
need
to
send
my
multi-homed.
L
Multihoming
draft
also
mentions
that
it's
actually
probably
the
biggest
unsolved
problem,
because
you
can
solve
the
route
increase
you
using
as
IDR,
but
as
long
as
you
start
seeing
dns
split
horizon
and
you
could
not,
you
need
actually
to
use
the
property
unless,
when
you
combine
it
with
network
information.
Yes,
this
is
a
great
other
console
than
this.
Okay.
M
So
Warren
Kumari,
following
up
from
Eric's
thing:
it's
not
only
adding
new
things,
it's
also.
What
should
you
do
with
ones?
You
don't
understand
right.
Most
of
these
I
think
would
be
optional,
but
possibly
having
some
like.
If
you
don't
understand
this,
please
don't
use
it
so
breaking
this
search
space
up
into
okay,
thank
you
that
could
do
both
okay.
A
N
Brian
carpenter,
those
of
you
who
were
in
the
RF
C++
buff
on
Monday
evening,
might
have
heard
Martin
Thompson
make
a
remark
about
the
awfulness
of
STD
25
standard,
I,
hate,
EF,
standard
number,
25,
well,
actually,
internet
standard
number
25,
which
is
also
RFC.
Eight
six
seven
I
just
wanted
to
give
an
implementation
report
on
that
standard
for
ipv6
it's
twelve
lines
of
Python.
That's
my
version.
Martin
wrote
one
which
is
one
line
of
JavaScript
and
my
client,
which
is
actually
only
six
lines
of
Python,
interrupts
interoperates
with
his
server
I.
N
Don't
know
if
he's
got
a
client
which
would
interoperate
with
my
server
anyway,
the
they
interoperate
very
nicely
using
ipv6
I
didn't
even
bother
to
implement
an
ipv4
version,
so
this
is
a
standard
from
1983
I
think
which
probably
hasn't
been
used
for
last
20
years,
which
was
criticized
for
being
badly
written,
which
we
both
managed
to
implement
for
ipv6
I
join
working
group
sessions.
Of
course,
okay.
A
I
should
mention
we
had
an
outcome
from
the
discussion
yesterday
of
prop
64
internet
draft
Warren
has
taken
the
cheshire
draft
on
ipv4
DARPA
ipv4,
only
DARPA
as
a
Navy
sponsored
draft
and
basically
saying
please
make
it
so
and
as
as
we
discussed
yesterday,
Lorenzo
and
Jen
and
Eric
file
the
drafts
saying
what,
if
we
use
an
RA
option
and
I
understood
the
working
group
that
people
here
to
be
saying.
Yes,
let's
do
that
so
this
morning
I
sent
a
note
to
the
six-man
chairs,
saying
your
six
men.
A
M
M
I'm,
not
a
huge
fan
of
70/50
I
think
that
many
people
view
it
with
a
somewhat
jaundiced
eye,
but
I
think
that
what
the
document
that
I
am
sort
of
ad
sponsoring
does
doesn't
say
that
7050
is
brilliant
or
bad
or
anything
all
it
tries
to
do
is
make
it.
This
is
already
deployed,
let's
try
and
take
some
of
the
sharp
pointy
edges
of
it,
so
that
people
who
are
using
it
don't
shoot
themselves
in
the
foot.
M
So
I
think
that
you
know,
regardless
of
what
happens
with
an
RA
option,
making
70/50
less
dangerous
or
less
tricky
temperament
is
still
worth
doing.
So,
if
anybody
thinks
that
the
document
that
I'm
ad
sponsoring
is
really
dangerous,
I'd
love
to
know
that.
But
unless
I
hear
that
I
think
you
know
making
it
suck
less
would
be
good.
M
C
A
D
Just
wanted
to
say
thanks
for
and
that's
a
good
summary
I'm,
not
especially
a
fan
of
70/50,
either
when
we're
implementing
at
Apple.
We
we
ran
into
problems
where,
when
users
of
overridden
DNS
or
they
using
VPN
with
the
discussion
this
time
about
DNS
over
TLS
and
over
HTTP
I,
think
it's
it's
it's
becoming
more
into
the
spotlight.
I
think
a
couple
of
years
ago,
when
we
identified
this
problem,
a
lot
of
people
hadn't
seen
it
I
think
more
people
are
seeing
this
problem.
M
D
So
if
we
are
going
to
use
70
50
there,
our
draft
points
out
some
of
the
problems
and
at
what
you
have
to
do
to
make
it
work.
I'm
not
especially
advocating
there
and
I'd
be
fully
supportive
of
RA
or
other
suggestions
that
people
have.
But
if
we
are
gonna
we
should
either
deprecated
70
50
and
say:
don't
do
it.
Who
should
say
how
you
should
do
it
so
that
it
works
so.
M
Thank
you
guys
actually
can
follow
up
from
that.
You
made
my
point
kinda
for
me.
I've
heard
some
people
say
you
know.
The
70/50
thing
does
not
work
hugely
well.
How
about
we
just
deprecated
that
well
I
heard
some
people
I,
don't
think
we
can
do
that
it's
widely
deployed,
if
somebody
does
want
to
do
that,
that
should
probably
happen
in
a
different
document
or
somewhere
else.
Not
here
separate
discussion.
L
Okay,
channeling
Cola,
so
first
of
all,
we've
just
uploaded
the
0
1,
because
0
is
0
version
was
a
bit
incomplete.
So
it
addressed
some
of
the
comments
received
yesterday.
So
we
basically
clarified
all
those
load-balanced
ed
crouching
like
give
some
examples
of
how
it
should
be
done
and
what
is
what
is
in
scope?
What
is
out
of
scope
to
avoid
confusions,
so
everyone
who
commented
guest
today
we
I
appreciate
if
you
can
have
a
look
and
let
us
know
if
new
text
is
better
and
I
will
send
update
on
6
mine.
L
My
link
list,
so
I
think
we
should
actually
have
discussion
in
one
place
because
sorry
I
did
yesterday,
we
uploaded
the
draft
word,
but
we
did
not
send
notification
on
the
list
and
I
wonder
yeah
7050,
so
I
read
at
70/50
that
looks
like
actually
a
two-parter
is
how
you
get
the
prefix
and
they're.
A
very
interesting
part
of
what
host
is
expected
to
do.
What
get
the
prefix
and
currently
our
draft
is
said
when
host
get
the
prefix
it
is
a
behavior.
Is
it
should
be
similar
to
what
70/50
does
especially?
L
It
applies
in
situation
when
you
have
more
than
one
prefix
what
was
supposed
to
do
and
so
on.
So
if
people
talking
about
duplicated
70,
50
or
saying
it
dangerous,
we
need
to
probably
make
it
clear
if
we
talking
about
discovering
the
prefix
or
host
behaviors
in
terms
of
what
is
expected
from
the
host.
O
O
There
are
some
discussion
which
is
ongoing:
I
think
that
is
too
too
early
to
have
that
in
this
kind
kind
of
profile
document
risk
and
he
did
by
majority
I
think
that
discussion
can
can
be
separated
from
from
the
current
the
can
profile
so
that
the
profile
can
continue
in
its
own
route
and
have
the
error
discussion
in
six-man
with
other
people
change
their
mind.
Since
the
consensus
that
we
have
behaved
working
up
when
we
really
produce
the
RFC,
1750,
51
I
think
the
Arab
Nia.
O
It
means
him,
I
would
say
intuitive
and
so
on,
but
it
has
a
lot
of
issues
there.
That
need
to
be
to
be
to
be
solve
and
discussed,
and
the
other
point
is
that
we,
as
in
the
if
you
see,
for
instance,
the
the
text
proposal
from
Jody
there.
There
is
this
text
about
your
put
into
to
the
option,
and
there
is
the
this
new
text,
which
is
pure
things
in
the
way.
In
case
you
are
discovering
the
prefix
in
multiple
ways.
O
There
so
you
start
with
72
to
25,
which
is
form
is
superior
to
the
other,
because
it
is
really
determines.
You
can
use
all
the
problems
there
and
then
you
go
to
the
DHCP
option
and
then
the
third
option,
with
the
array,
and
then
you
end
up
with
70
50
50.
So
we
have
already
multiple
options
and
ways
to
discover
to
discover
the
prefix
level.
O
It's
what
was
the
point
whether
to
add
in
another
one
and
so
on
this
kind
of
discussion
to
separate
it
from
the
the
profiles
that
I
don't
think
that
it's
for
me
in
general
I
would
I
don't
like
to
have
individual
brats
in
stable
profile
document,
because
it
doesn't
make
sense
at
all
there.
So
that's
that's
that's.
My
solution
to
the
working
group
is
to
remove
this
discussion
from
from
your
drives.
F
Other
entity
as
an
author
of
70/50,
sorry
as
an
author
of
the
draft,
sorry
as
an
implementer
of
70/50
I,
agree
with
Warren
7050
is
hard
to
implement
it's
it's
with,
particularly
with
its
interactions
with
DNS
over
TLS
and
VPN.
It's
the
wrong
way
to
do
it.
That
said,
Mohammed
by
himself
is
in
a
position
to
block
our
new
draft
until
the
ITF
closes,
and
so
there
is
no
guarantee
that
this
truffle
ever
survives.
Six
men,
therefore,
from
a
pragmatic
perspective,
sure
fixed
70/50.
F
F
We
have
a
bunch
of
stuff
that
we
standardized
earlier
I,
looked
at
7:51
and
reasons
not
to
do
array
we're
tenuous
or
basically
invalid
at
best,
and
but
of
course,
if
we
say
well,
we
already
have
too
many
ways
of
doing
this,
even
though
only
one
of
them
is
implemented.
That
was
like
yeah
I
mean.
If
that
position
stands,
then
we
go
nowhere
and
so
yeah
fix
70
50
by
all
means.
F
And
then
you
know
people
will
be
unhappy
with
it,
but
until
we
get
a
replacement,
I
can
tell
you
for
one
thing:
it's
not
going
to
be
72
25,
because
I'm
never
gonna
implement
that
because
it's
way
to
complicate
it
so
we're
just
gonna.
We
could
just
you
know
so
yeah,
let's
fix
70
50,
but
that's
what
we
that's,
how
we
can
primarily
proceed.
O
7225
is
not
only
to
were
to
solve
some
of
the
issues
there,
but
also
it
solved
the
problem
if
you
have
a
net
64
and
if
you
want
to
to
cover
the
incoming
connection,
you
need
to
solve
this
problem
too
and
then
white.
It's
it's
something
that
you
say
that
you
are
solving,
not
only
the
the
discovery
of
the
prefix
there,
but
you
also
another
issue
that
you
are
encountering.
We
didn't
s
64,
so
it's
one
tool
that
will
solve
your
multiple
issues
so,
but
that's
from
its
yet
it's
we.
O
We
have
the
PCP
server
which
is
implement
in
ask
for
devices.
Today
we
have
multiple
vendors
who
are
offering
Oh.
Today
we
have
an
issue
with
the
handset
to
have
the
PCP
client
to
be
support,
at
least
for
the
mobile
handsets.
So
if
you
think
it
is
complex
to
to
support
PC
Pacific
client
at
the
handset
yeah,
what
should
I?
What
should
I
say
more?
So
it's
it's.