►
From YouTube: IETF113-V6OPS-20220321-0900
Description
V6OPS meeting session at IETF113
2022/03/21 0900
https://datatracker.ietf.org/meeting/113/proceedings/
A
Yeah
and
as
you
can
probably
tell
I
am
neither
fred
nor
ron-
I
am
just
the
monkey
who's
sitting
in
the
chair
largely
to
try
and
keep
order
in
the
room.
So
you
know
if
anybody
misbehaves
at
the
mic
or
something
I'm
the
bouncer
for
for
ensuring
compliance,
he
will
comply.
D
This
is
fred.
Are
you
running
the
slides
as
well.
A
Okay,
let
me
see
if
I
can
get
that
logged
in
one
of
the
problems
is
I
don't
have
an
easy
way
for
people
to
change
to
the
next
slide,
because
I
do
not
have
a
clicker
thing,
so
I
can't
hand
people
a
clicker
or
if
they
need
to
change
their
slide.
If
they're
here
in
person.
A
E
D
F
A
I
mean
if
the
chairs
would
like
I
can
do
the
slide
sharing
from
in
the
room
so
that
people
can
say
next
and
I
can
push
the
next
button.
Otherwise,
if
anybody
has
a
usbc
to
usb
a
converter,
I
can
plug
in
a
clicker
thing.
I'm
only
slightly
terrified
about
plugging
a
random
does
that
have
a
converter,
converter.
D
Okay,
warren:
I
have
all
of
these
slides
up
at
the
meeting
materials
site,
but
for
some
reason,
when
I
choose
a
deck
to
share,
it
only
shows
two
sets
of
slides.
G
Hello,
can
you
hear
me
yes,
hi
juan
carlos
here,
I
I
attended
amitako
training
yesterday
and
there
are
a
couple
of
features
so,
first
of
all,
the
the
chairs
need
to
to
share
the
slides,
so
they
have
to
prepare
them.
Second
for
moving
the
slides.
G
If
someone
scans
the
qr
code
with
their
phone,
they
can
actually
move
ahead
and
back
so
that
you
don't
need
to
get
the
pointer
the
they
just
have
to
request
permission
to
control
the
presentation,
and
then
anyone
from
the
phone
can
actually
control
the
the
back
and
forth
of
the
slides
if
that's
more
convenient
so
that
we
don't
have
to
pass
around
stuff.
A
D
Okay,
I'm
trying
to
share
from
my
laptop
and
the
the
system
doesn't
want
to
do
that
for
some
reason.
D
Okay,
good,
the
first
set
of
slides,
is
in
administrative.
D
Yeah,
let's
see.
A
A
E
D
H
H
Okay,
if
you
open
that
page
and
scroll
down
to
slides.
H
J
K
L
D
D
Okay,
if
you
haven't
seen
a
note
well
screen
before
you
see
one
now
and
basically
you
you
should
remember
the
things
you
say
here.
You
have
legal
significance,
and
so
you
know
don't
be
stupid.
Next
slide.
D
Okay,
so
I
see
it
now,
meeting
pip
he's
playing
in
using
meat
echo
using
the
onsite,
do
and
use
it
to
join
the
mic
view
if
you're
remote,
like
I
am-
please
don't
share
unless
you're
intending
to
speak.
D
So
this
is
our
agenda
today
question:
does
anybody
have
any
changes
that
they
would
like
to
make
to
the
agenda.
M
N
N
So,
just
to
give
you
a
bit
of
the
history
of
the
draft,
you
see
that
we
submitted
as
the
personal
version
in
it
was
october
2020
in
april
2021.
It
was
accepted
by
the
working
group
as
a
work
group
draft
and
basically
we
completely
reviewed
the
draft
for
itf
111.
It
was
a
version
zero
two,
so
we
did
the
kind
of
major
transformation.
N
We
introduced
a
lot
of
changes,
we
reshaped
the
table
of
content
and,
let's
say,
since
version
zero,
two,
the
the
draft
can
be
considered
as
stable
enough
after
atf
111.
We
had
three
other
versions,
but
basically
those
versions
were,
let's
say,
prepared
to
address
the
comments
raised
by
you
by
the
working
group
and
we
are
now
at
version
zero.
Five.
By
the
way
you
see
they
are
the
list
of.
I
N
Who
contributed
in
the
discussion,
and
there
are
others
who
also
provided
comments
offline,
so
the
next
slide
is
good,
so
where
we
are
right
now,
as
I
said
version
zero
t
was,
let's
say,
done
to
address
the
comments
received
for
atf
111.
N
We
touched
the
table
of
content,
but
since
then
it
has
remained
stable,
stable
enough
and
the
other
changes
we
did
mainly
in
version
zero.
Four
and
zero
five
were,
to
let's
say,
update
the
numbers
about
the
deployment
and
the
adoption
of
ipv6,
and
we
touched
also
the
text,
for
example,
addressing
comments
asking
to
introduce
new
references
to
avoid
the
ambiguity
of
some
terms.
So
you
see
an
example
there,
for
example,
remove
migration
to
ipv6,
just
use
the
term
transition,
286
or,
let's
say,
expand
what
we
considered
as
the
base
text.
N
For
example,
please
explain
a
bit
more
the
behavior
of
a
happy
eyeball
so
that
those
were
examples
you
see
better
in
the
next
slides
what
we
have
done
so
far.
So
next.
N
So
you
see
the
table
of
content.
As
I
said,
it
has
been
stable
since
version
zero.
Two,
the
let's
say
the
list
of
the
topics
is
what
you
see.
They
are
what
we
touched
in
version:
zero,
four
and
zero.
Five
is
highlighted
there
so
version
zero.
Four
in
light
blue
and
version
zero,
five
in
yellow-
and,
as
I
said,
basically,
it
was
to
introduce
some
slide
modifications
or
just
touch
the
base
text
more
in
the
next
slide.
N
So,
basically,
I'm
not
going
to
give
you
the
full
list
of
details
across
the
table,
but
just
to
highlight
the
main
changes
you
see
version
zero
four
was
about
introducing
an
update
on
the
numbers
of
ipv6
adoption,
and
you
know
users
whatever
it
was
in
section
two
and
three
and
then
as
far
as
version
zero
four
is
concerned
also,
we
touched
also
section
five,
so
we
added
more
references
and
section
7.3.
We
basically
reviewed
the
test
about
the
operations
of
ipv6
for
version
0.5.
N
We
introduced
changes
in
section
3.3.6,
so
basically
it
was
about
you
know
to
expand
the
behavior
of
api
ball.
We
also
touched
other
sections,
for
example,
section
4
4.1
for
those
two
just
to
explain
better
and
thanks
for
the
comment
we
received,
that
introducing
ipv6
in
the
overlay
is
a
matter
of
having
let's
say
two
steps
or
two
phases.
N
We
named
them.
We
named
them
ipv6,
introduction
and,
let's
say,
full
transition
to
ipv6
pv6.
Only
basically,
those
phases
are
not
to
be
considered
as
a
sequence.
Clearly,
an
operator
can
jump
on
one
of
the
two,
according
to
the
airplanes,
so
clearly
from
a
deployment
perspective,
you
can
have
the
first
phase,
but
just
to
be
say,
the
preparation
for
the
second
one,
an
operator
can
also
decide
to
jump
directly
to
ipv6
only
and
if
I'm
not
wrong.
N
N
As
I
said,
the
the
draft
has
received
a
few
comments,
but
overall
the
content,
the
table
of
content
is
stable
enough.
So
we
like
to
ask
for
the
group
last
call
so
the
adoption
of
the
draft
in
any
cases,
any
comments.
Any
questions
are
more
than
welcome
and
this
is
it.
Thank
you.
D
So
paulo,
I
think
you're
done,
should
we
move
to
the
next
speaker.
Thank
you.
C
I
will
be
speaking
about
detection
of
nuts
foreign
through
the
srv
records.
Please
next
slide,
so
why.
C
You
this
uses
a
dns,
well-known
name
and
has
been
updated
by
rc
8880,
there's
also
72
25,
which
uses
support
control
protocol,
there's
also
dhcp
v6
option
and
array
option
and
xlips,
but
these
solutions
do
have
some
issues
connected
with
them.
The
70
50
is
pretty
hard
to
implement
right.
C
It's
easy
to
do
wrong
and
it's
easy
to
use,
even
if
it's
implemented
wrong,
but
in
this
case
it's
not
quite
secure,
it
doesn't
work
with
the
dns
six
dns
providers
other
than
the
one
of
the
isp.
C
7225
is
largely
ignored
same
also
for
the
hcp
option,
which
is
ignored
by
android
because
of
dhcp
and
there's
array
option.
But
there's
not.
I
haven't
seen
implementation
in
routers
so
far
and
concerning
the
the
time
in
which
rotors
are
replaced,
then
we
won't
see
it
for
some
time
in
some
networks.
C
C
So
the
rc
7050
requires
to
be
secure,
dns
signs
records,
corresponding
ptr.
These
are
obvious
and
then
there's
a
requirement
of
secure
channel
between
a
node
and
the
receiver,
which
is
not
always
the
case.
There's
also
the
trusted
domain
list
needed,
but
there
is
no
written
how
this
list
should
be
populated,
and
there
is
only
mentioned
that
there
should
be
no
user
input.
So
user
count
actually
says
that
this
domain
is
trusted.
C
Then
there
are
three
other
requirements
from
8880:
subrosal
must
distinguish
between
configuration
of
dns
because
it
has
to
use
auto
configuration
only
and
the
recursive
dns
resolver
is
interface-specific.
C
These
are
problematic,
at
least
what
I
have
seen
in
linux
there.
The
resolve
conf,
doesn't
include
the
source
of
the
configuration
of
this
information,
so
this
rfc
70
50
when
you
had
to
follow
all
the
prerequisite
requirements
it's
quite
hard
to
to
implement,
and
I
one
implementation
I
have
seen
on
linux.
C
Then
it
just
uses
the
well-known
name
and
the
ask
for
what
well-known
name
and
they
checked
the
vtr
record
and
nothing
more
and
it
didn't
follow
all
the
requirements
next
time.
C
It
should
use
only
white
super
widely
supported
protocols
like
the
application
already
knows
how
to
talk.
It
does
information
already
provided
so
the
note
can
know
what
information
should
know,
what
information
the
no
so
it
can
just
use.
The
information
it
already
has
must
work
with
foreign
dns,
so,
no
matter
what
operator
of
dns
would
be
used,
it
should
be
also
work
with
it.
C
No
synthesis
on
the
hose,
so
no
other
things
that
are
not
already
introduced
into
node
and
must
work
with
the
no
provisioning.
So
when
user
brings
its
own
device,
there
should
be
no
exchange
of
certificates
and
so
on
and
must
provide
detection
over
insecure
channel
and
must
run
in
the
user
space.
So
every
application
in
the
host
can
use
utilize.
This
method,
just
like
this.
C
Okay,
so
what
are
the
results
in
the
draft?
There
is
a.
C
The
srv
records
are
there
established,
the
application
should
be.
The
application
already
knows
how
to
talk
to
dns,
or
at
least
every
application
should
be
able
to
a
node
nodes
is
ip
address,
so
it
can
have
a
ptr
record.
C
C
So
I
think
the
srv
records
are
good
in
place
to
place
such
information
because
they
are
structured.
They
do
have
priorities
ways
and
they
do
have
a
time
to
live
so
they
can
be
updated.
Theoretically,
and
you
know
it
can
be
up
to
host
specific
level
so
for
every
ptr
there
could
should
be,
could
be
a
different
srv
record
and
different,
not
six.
Four
gateways.
C
So
how
the
srv
record
should
look
like,
at
least
from
my
perspective.
There
are
two
proposed
srv
records,
one
for
not
six
four
and
one
for
dns,
six,
four,
it
it
can
be
for
host
fully
qualified
domain
name
or
the
domain
itself.
C
It
utilizes
the
priorities
and
weights.
Only
the
port
number
can
be
utilized
to
transmit
other
information,
but
it's
not
mandatory
to
to
use
it
in
this
way.
C
C
So
what
are
in
the
news
from
the
previous
version
of
the
house
next
slide,
please.
So
there
is
version
zero.
One
daniel
has
been
just
detailed,
some
process
of
the
local
domain
detection
from
the
ptr
records.
There's
also
updates
from
the
previously
wrongly
submitted
draft
on
it
104,
and
there
are
some
changes
in
version:
zero,
zero.
It's
uses
a
ptr
record
instead
of
dns
search
list.
It's
because
the
dns
search
list
is
not
transited
through
throughout
the
routers
and
this
information
the
ptr
records
are
so
because
of
this.
C
C
There
are
also
specified
now
some
interaction
with
other
methods
and
forces
for
x
slot.
The
trans
there's
also
mentioned
that
any
transport
method
can
be
used,
so
not
just
the
plane,
dns
doh
dot.
Anything
actually,
there's
a
also
specified
how
the
negative
answer
should
be
should
be
counted
and
what
the
node
should
do
with
the
time
to
live.
C
There's
also
mentioned
there
can
be
also
the
multicast
support
and
there's
a
proof
of
concept
code.
On
github
repository
connected
with
the
draft
and
the
last
changes
I
was
spelling
in
grammar.
So
next
slide.
Please,
and
this
should
be
the
last
one.
So
thank
you
for
attention
and
here's
the
github
repository
with
the
proof
of
concept
code,
but
be
aware
that
I'm
not
programmers
of
the
code.
A
O
Hi,
I
think
some
questions
about
the
document
and
the
presentation
just
clarifications.
I
am
not
opposing
to
the
document
at
all
just
making
sure
that
some
some
things
are
correct.
The
first
one
is
that
rfc
7225
is
actually
being
used
by
some
isps,
so
it
should
not
be
ignored.
O
O
The
other
question
is
that
most
of
the
cpes
that
I
see
and
even
operating
systems
like
like
ios
or
or
android,
when
they
are
ready
for
using
964
or
464x
lat,
they
have
already
pre-configured
the
well-known
prefix.
Okay
at
the
other
point
I
have
is
dns.
64
synthesis
is
actually
a
good
thing
for
being
done
in
the
host,
so
I
am
not
sure
how
we
should
recommend
not
doing
that.
If
that's
the
intention
of
of
the
document
or
or
it
was
a
confusion
in
your
presentation,
there
is
yeah
yeah.
C
O
Okay,
clear
then
there
is
another
problem
is
that
you
mentioned
about
some
operators
that
the
host
know
the
ptr
they
hit
his
address
because
the
ptr?
This
is
not
actually
actually
true
in
many
cases,
because
some
operators
do
not
configure
ptrs
for
enterprises.
O
I
mean
enterprise
connections,
even
unless
asa
specifically
regarded
and
even
less
for
residential
users.
I
I
know
this
is
not
good,
but
it's
the
reality,
and
my
last
point
is
that,
considering
that
the
most
of
the
deployments
are
using
well-known
prefix,
maybe
we
need
a
document
that
describes
practical
use
cases
for
the
network-specific
prefix
or
otherwise.
O
Do
we
really
need
more
procedures
to
to
to
to
to
learn
the
nav64?
I
am
not
really
convinced
that
therefore
makes
sense
again.
I
am
not
against
the
document,
but
I
am
not
really
convinced
that
it
will
make
sense.
Thank
you.
C
So
we
just
our
act
on
the
ptr
record.
I
know
that
it's
not
the
usual
thing
to
make
the
ptr
record,
but
there
are
implementations
that
allow
dynamically
generating
those
records,
so
it's
possible
to
use.
I've
also
tried
to
configure
it
myself.
It's
not
not
really
a
problem
to
to
do,
but
I
also
understand
that
it's
not
the
typical
use.
C
I
P
Okay,
so
you
say:
ptr
is
the
entry
point
into
the
secure
delegation.
Yet
in
the
draft
I,
I
might
have
read
the
wrong
version.
There's
no
example
of
the
ptr
and
what
are
you
going
to
do
with?
I
mean
people
in
non-public
ip
space,
so
how
the?
How
are
you
defining
how
to
get
the
actual
public
ip
space
because
you
might
be
rfc,
1980
or
euler
or
what
have
you
that
is
not
or
that
is
not
routable.
So
how
are
you
going
to
get
the
ip
and.
P
The
we
in
add,
we
sort
of
had
a
similar
problem
with
upgrading
dns
servers
and
we
decided
against
the
ptr,
because
it
was
a
hard
what
jordy
said
to
get
delegations
in
pdr
and
also
you
might
not
have
public
routable
ptr
available.
So
you're
saying
that
you
have
to
have
public
router
with
ptr
available
for
that
to
work.
C
P
Okay
and
when
you
say
from
the
ptr
again,
there
was
not
an
example:
you
get
a
domain,
but
that
domain
is
not
what
you're.
Looking
for
into
the
srv
record,
you
are
walking
it
kind
of
label
by
label
to
get
something
that
is
working.
Is
it
correct
understanding.
P
C
For
the
whole
specific
people,
srv
record,
so
we're
a
full
qualified
domain
name
and
then,
if
there's
no
record,
not
even
a
dot
record
srv,
then
you
moved
one.
P
Step
up
you
remove
one
label
right,
yeah
and,
and
the
draft
says
you
are
stopping
by
the
second
level
domain
yeah.
Now
there
are
third
level
domains
that
are
kind
of
actually
second
level.
Women
like
co
dot
uk,
so
the
drop
should
specific,
maybe
define
more
where
to
stop,
because
otherwise
you
could
have
a
rogue
actor
doing
something
bad
to
that
yeah.
This.
C
Q
Ted
lemon
apple,
so,
first
of
all,
the
document
seems
to
be
using
the
ptr
record
to
discover
the
local
domain,
but,
as
ralph
was
saying,
it
doesn't
actually
say
how
that's
done.
Are
you
aware
of
rfc
c6763
the
dns
service
discovery,
not
from
top
of
my
head?
Okay,
it
would
be
worth
reading
that
because
rsv6763
actually
specifies
how
to
discover
services
using
the
ip
address
to
look
up
a
ptr
record,
and
I
think
you
know
just
referencing
the
mechanism
that
it
uses.
Q
In
fact,
you
could
probably
just
use
dns
service
discovery
directly,
because
it's
doing
almost
exactly
what
you
describe
in
the
document.
The
other
issue
that
I
well,
I
would
concur
with
the
point
that
was
raised
earlier,
that
doing
validation
in
the
stub
is
better,
and
so
therefore,
synthesis
should
also
happen
in
the
stub,
but
the
sorry
my
brain's
a
little
foggy
from
being
up
too
early.
Q
Q
The
document
doesn't
actually
provide
any
additional
security,
so
it
would
probably
be
worth
taking
that
out
if
you,
if
you
aren't
able
to
explicitly
say
what
security
is
provided
and
if
you
want
to
say
what
security
is
provided,
you
should
describe
explicitly
what
security
you're
providing
because,
as
far
as
I
can
tell
from
reading
the
document,
this
is
no
more
secure
than
any
of
the
other
solutions.
I
mean
if
you,
if
you
bootstrap,
based
on
on
a
non-secure
dns
lookup,
for
example,
essentially
you're
just
trusting
the
network.
So
it's.
C
You
actually,
when
you
do,
have
a
secure
delegation
of
a
ptr,
and
you
do
have
and
you're
supposed
to
have
the
address
of
the
note
valid
and
provided
by
some
secure
way,
at
least
a
guard
or
something
like
this.
Then
you
can
go
from
that
and
use
the
dns
sec
route
to
actually
validate
the
information.
C
Q
Every
every
step
of
the
detection.
I
understand
that,
but
the
problem
with
that
is
that
that
assumes
that
you
have
a
secured
delegation
for
the
ptr
record
and
for
almost
every
ip
address
that
any
actual
node
is
going
to
get.
You
can't
establish
a
security
delegation
because
it's
going
to
be
an
rfc
1918
address,
so
for
the
for
the
vast
majority
of
use
cases,
you
don't
get
any
security
now
you
you
could,
for
example,
talk
about
a
ce
router
that
does
validation
of
the
rfc
1918.
Q
Sorry
of
the
of
the
ptr
record,
because
that
ptr
record
could
have
a
secured
allegation
because
it
was
it's
a
public
ip
address
assigned
by
the
isp
we
hope,
but
but
in
the
general
case
you
can't
make
that
assumption
and
yeah
okay.
So
if,
if
you
have
a,
if
you
have
an
ipv6
address
on
the
network,
then
yeah
you
can
have
a
secure
delegation
for
that.
So
so
that's
that's.
Sort
of
useful
it'd
also
be
useful
to
connect
that
to
vrpki
validation.
Q
C
Q
J
Jen
lincoln,
first
of
all,
I'm
a
bit
confused
about
the
differentiation
between
application
and
a
node.
You
started
your
draft
with
an
application,
might
not
have
access
to
array
information,
and
then
your
draft
continues
saying.
J
Oh,
this
method
should
be
used
before
any
other
method
as
soon
as
address
is
configured
so,
and
it
keeps
talking
about
a
node
which
I
read
as
an
operating
system
right,
so
are
you
suggesting
that
the
host
should
ignore
information
received
an
array
until
this
process,
which
means
sending
a
number
of
dns
packets
is
completed,
and
only
then
you
can
start
using
array
information
which,
to
be
honest,
it
looks
a
bit
confusing
to
me
because
you
introduced
a
lot
of
delay
in
configuring.
A
host
and
by
the
way,
some
routing
vendor
are
implementing
carry
option.
J
C
Yeah,
I
hope
that
I
will.
We
will
see
this
ra
option
implemented
soon,
but
what
I
can
see
from
the
operators
are
what
I'm
working
for.
They
will
not
change
the
routers
in
in
a
few
years,
so
between
the
option
would
be
introduced
into
the
firmwares
of
the
routers.
Not
every
router
would
have
a
firmware
update
and
not
everyone
would
be
able
to
to
produce
this
r
option.
J
So,
in
this
case,
the
text
is
even
more
confusing
because
if
your
system
or
node,
whatever
you
called
in
the
draft,
could
not
even
access
array
option.
How
can
you
recommend
using
this
method
and
then
based
on
preference,
number
wait
for
another
method
to
complete
if
you're
claiming
you
can't
even
access
array.
C
If
you
can't
use
other
methods,
then
you
are,
you
have
to
stick
with
the
methods
you
can
access.
So
basically.
J
I
think
I
think
the
draft
is
very
confusing
about
how
it's
supposed
to
work
between
application
and
the
operating
system,
especially
when
draft
is
saying
doing
something.
If
you
don't
have
c
lot
or
doing
something
when
address
is
configured,
your
application
might
not
even
have
access
to
that
information
right
at
all
and
by
the
way
you
say
when
an
ip
address
is
configured.
C
Yeah
it
it
might
it
if
it
comes
from
the
same
prefix.
That's
there
there's
no
point
to
to
actually
ask
for
every
single
ip.
But
yes,
it
should
generally
ask
for
ask
for
every
id
and.
J
I
think
the
last
comment.
I
think
it
might
work
clarifying
what
to
do
if
there
is
ipv4
address
present,
because
I
don't
think
the
draft
even
mentions
that
yeah.
L
Thomas
eckhart,
so
if
I
understand
it
correctly,
the
goal
is
to
make
it
easy
for
applications
using
a
simple
library
on
top
of
supposedly
existing
dns
libraries
to
utilize
the
scheme.
So
let
me
not
comment
more
on
the
the
feasibility
of
the
specific
choices
you
made,
but
given
how
the
long
list
of
you
know
not
seemingly
easily
adoptable
options
we've
already
built.
It
seems
to
me
that
the
one
thing
missing
would
be
to
also
have
and
abstract
api
specification
about
how
you
know.
L
Application
should
discover
and
utilize
the
64
mapping
information
so
that
well,
we
don't
necessarily
have
to
get.
You
know
hundreds
of
applications
trying
to
read
that
draft
and
trying
to
do
their
own
implementation
and
also
us
ending
up
getting
a
lot
more
freedom.
L
When
we're
basically
saying
that
the
ultimate
specification
against
which
applications
are
written
is
not
all
the
details
of
any
specific
dns
encoding,
but
that
that
can
be
changed
and
enhanced
over
time,
without
necessarily
changing
the
api
right,
but
just
being
able
to,
in
an
agile
fashion,
improve
on
the
on
the
on
the
libraries
keeping
the
apis.
And
then,
if
these
dns
things
are
not
secure
enough
or
need
to
be
improved,
then
hopefully
we
could
do
that
a
lot
better
right.
L
So
just
don't
think
about
the
protocol
elements
that
you
want
to
do
with
dns,
but
also
think
about
what
is
the
you
know,
abstract
api,
maybe
go
to
tabs
with
that.
I'm
not
sure
if
they're
more
experienced
in
in
something
like
this,
then
v6
ops
would
be,
but
certainly
I
think
it
would
help
a
lot
to
move
this
forward.
If
there
was
also
that
api
aspect
covered.
D
D
Let's
take
this
off
the
list
and
speakers
in
general,
I've
asked
you
to
commend
yourselves
to
10
or
15
minutes.
We
are
now
48
minutes
into
the
first
hour.
K
K
Thank
you.
So,
in
the
last
version
I
had
some
let's
say:
warm-up
results
about
iptvs,
but
now
they
are
real
ip
version.
Six
transition
technology
results.
So
it's
a
result
of
dual
facebook
and
86
for
implementation.
K
It
was
recommended
at
the
last
iedf
meeting
and
it
was
an
open
source,
any
d64
implementation
and
as
last
time
I
again
did
two
kind
of
tests,
calibrate
agents,
the
number
of
cpu
cores
and
there'll
be
scalability
the
number
of
concurrent
sessions.
So
the
method
is
this
guide
in
our
draft,
which
was
presented
already
twice
at
bmwg
working
group
and
the
session
shared
like
target.
K
So
the
the
problem
is
that
if
you
use
let's
say
facebook
l864,
you
cannot
send
traffic
from
the
version.
Four
to
version
six
direction,
just
the
random
port
numbers
would
be
required
by
rsd4814.
K
K
K
But
of
course
we
use
a
methodology
of
rsd,
82
and
19
for
iq
version.
Signature
technology-
and
here
is
the
description
of
the
hardware
used
for
the
test.
You
can
jump
it
over
and
put
your
results
for
next
type
d,.
K
Unfortunately,
you
cannot
see
my
my
mouse
pointing,
but
in
the
first
row
you
can
see
the
number
of
cpu
ports
and
the
second
third
and
fourth
line.
You
can
see
the
media
and
minimum
maximum
values
of
connection
establishment
rate.
So,
as
you
can
see
from
one
to
two
chords,
it
is
increased,
let's
say
for
220
to
350
kilo
connections
per
second,
so
it's
quite
okay
and
that's
not
linear,
but
quite
okay,
and
even
with
full
port,
it's
significantly
higher.
So
it's
not
linear,
but
quite
okay.
K
However,
for
eight
on
16
cores
it
doesn't
scale
up
well.
So,
unlike
last
time,
we
I
showed
the
results
of
ip
tables,
which
showed
about
more
than
10
times
performance
at
16
cores
compared
to
single
core.
Now
it's
about
less
than,
let's
say
four
ports
with
fixing
cores,
so
it's
really
bad
scale
up
and
also
you
you
can
see
a
throughput
is
very
similar.
It
gives
up
somewhat
up
to
four
cores,
but
after
that
it
doesn't
get
up
bad
we've
got
the
next
slide.
Please.
K
K
Both
the
source,
force
and
destination
for
cambridge
was
usually
doubled,
and
this
time
I
couldn't
do
any
tuning
like
with
ip
last
time.
So
I
measured
three
kinds
of
quantity,
maximum
connection,
establishment
rate
and
screw
put,
and
there
was
some
new,
which
is
the
connection
t's
down
rate,
which
has
different
has
been
defined
lately
and
on
wednesday.
I
will
talk
about
this
methodology,
but
I
will
shortly
mention
this
one
too.
Could
you
go
to.
K
Next
time,
thank
you
so
here
you
can
see
that
in
the
no
no,
no
just
okay.
Thank
you
so
for
the
first
line
you
can
see
the
number
of
connections
coming,
starting
from
1.5
million
and
then
6
million
and
25
million
and
100
million
400
million
and
up
to
1.6
billion.
So
you
see,
the
number
of
source
ports
were
always
doubled,
except
the
last
case
when
I
couldn't
double
the
source
port
but
increase
four
times
the
destination
point,
and
you
can
see
here
the
connections
per
second
median
and
the
support
medium
values.
K
K
Just
something
new
the
connection
here
downright
unfortunately,
we
have
no
better
method
than
an
aggregate
measurement
that
we
load
a
given
number
of
connections
into
the
connection
tracking
table
using
the
first
phase
of
the
test
and
then
delete
the
entire
connection
tracking
table
and
of
course,
we
perform
the
measurements
with
different
kind
of
number
of
state
states
or
connections
in
that
connection
tracking
table.
K
K
K
Yes,
thank
you,
so
you
can
see
that
we
use
the
number
of
connections
and
you
can
see
that
the
duration
time
of
the
pool
table
in
the
number
four
roll
number
four
was
increasing.
It's
about
four
times
increased
and
if
I
subtracted
the
duration
time
of
the
empty
table,
then
it
was.
It
was
always
0.46
a
second.
K
K
I
retroactively
performed
the
connection
team
around
the
great
mathematics
I
could
evolve
and
it
was
funny
that,
if
tables
produced
a
very
high
throughput
and
also
very
high
financial
establishment
rate,
because
in
the
order
of
millions
per
second
and
its
connection
here
down
rate
is
quite
low.
So
it's
just
in
the
order
of
400
or
300
kilo
connections
per
second,
it's
really
low.
D
It
looks
like
we'll
need
to
take
comments
with
the
list
and
and
and
again
you
know
this
will
this
will
all
be
discussed?
We
will
have
the
weekly
discussion
periods
and
part
of
that,
of
course,
will
be.
D
If
you
feel
that
we
should
adopt
the
draft,
then
you
need
to
say
so,
okay,
so,
mr
shu,
I
believe
your
neck
and
pardon
me
if
I
mispronounced
your
name.
R
So,
first
of
all,
it's
not
easy
to
to
understand
all
the
issues
and
solutions.
So
in
our
draft
we
summarize
all
these
excels
and
solutions
into
a
single
place,
so
it
makes
it
much
easier
to
reference,
because
nd
is
a
very
important
practical.
You
know
it's
a
critical
part
of
the
first
half,
so
this
makes
things
a
little
easier,
and
another
thing
is
that
we
we
clearly
clarify
that
host
isolation,
meaning
like
putting
all
the
holes.
R
You
know
isolated
from
others
is
very
effective
in
preventing
some
of
the
energy
exceeds
preventing
some
of
the
anti-issues.
We
also
provide
a
guideline
say
where
to
isolate
holes
and
when
not
to
isolate
holes,
for
example.
If
you
need,
for
example,
multi-class
dns,
then
then
you
don't
want
to
isolate
the
hose
because
they
will
break
the
cervix
isolation,
and
the
overall
result
is
that
we
believe
that
this
will.
You
know
when
you
applying
these
guidelines,
you
can
simplify
the
first
hop
deployment
next
slightly.
R
So
I
would
like
to
talk
about
one
of
the
key
contributions
of
this
draft.
Is
that,
as
we
said,
you
know
if
multicast
or
you
know
you
cannot
trust
other
hosts,
then
you
know.
If
you
isolate
the
host,
then
obviously,
if
there's
only
a
single
host,
then
the
multicast
exceed
will
go
away.
You
don't
have
the
choices
you
have
other
holes
because
other
holes
are
separated
into.
You
know
different
subnet.
R
So
this
idea
first
came
from
rfc
8273
and
it's
called
a
unique
prefix
hose,
but
when
8273
was
before
it's
published
as
an
rfc
there's
actually
a
long
debate
in
the
here
in
whether
you
know
it's
a
it's,
a
good
idea
to
publish
it
as
an
rfc
and
some
of
the
concerns
is
pressed,
including
well.
There
are
certain
corner
cases,
because
ifc
8273.
R
It
basically
proposed
this
subnet
isolation
and
it
automatically
assumed
that
you
know
the
holes
are
also
isolated
in
l2,
but
you
know
there
are
certain
corner
cases.
For
example,
if
you
have
a
multimedia
access,
this
may
not
be
true,
so
there
are
certain
corner
cases,
and
although
there
are
some
opposition
that,
oh
you
know,
if
you
do
this
approach,
then
the
router
becomes
stateful
and
some
other
use
too
many
projects,
because
you're
basically
giving
each
host
a
prefix.
R
R
When
you,
you
know
kind
of
like
constrain
the
application
scenario
of
ifc
8273,
then
you
know
you
avoid
the
corner
cases.
This
is
the
first
one
in
you
know.
Regarding
the
stateful
router,
we
believe
that
it's
a
good
price
to
pay,
because
when
you
do
this
isolation,
then
you
know
you
the
holes,
you
don't
need
to.
You
know,
put
the
mechanism
in
the
whole
to
avoid
this
problem,
for
example.
Alternatively,
for
example,
the
wind,
the
wireless
nd
rfc
6775,
they
propose
some
changes
to
the
horse,
which
we
believe
that
you
know
is
more
complicated.
R
R
You
know
everybody
can
get
us
less
29
and,
as
less
29
contains,
you
know,
32
billion,
that's
64.,
so
it's
it's
more
than
enough.
Really,
the
the
the
scenario
where
we
are
where
we
have
the
most
number
of
codes,
for
example,
in
mobile
broadband,
we
are
already
giving
you
know
each
each
host,
meaning
the
the
mobile
phone
slash
64
anyway.
So
really
you
know
this
is
we
believe
that
use
too
many
profits
is
not
an
issue?
Actually.
You
know
this
takes
advantage
of
the
plaintiff
address
of
ipv6
to
simplify
the
deployment.
R
So
we
basically,
we
is
basically
discussed
this
idea.
You
know
this
whole
isolation
where
it's
booked
to
be
applied.
We
identify
that.
You
know
if
it's
a
public
access
network,
so
you
know
you
may
not
be
able
to
trust.
Other
hosts.
Think
you
know.
Host
isolation
is
is
a
good
idea
or
in
a
while
it's
especially
low
power
environment.
R
Where
you
know
multi
cars
can
consume
a
lot
of
power
unnecessarily
and
also
you
know,
may
not
be
that
reliable.
So
in
this
kind
of
scenario
it
can
be.
You
know
a
good
idea
to
apply
this
whole
isolation.
But
again
you
know
if
you
need
multicast,
dns
or
your
environment
is
very
safe
and
it's
wired.
You
know
you
don't
have
the
issue
we
talked
about,
then
you
don't
need
to
to
apply
this
post
isolation.
R
R
R
So
in
this
version
we
clearly
will
add
a
paragraph
to
address
the
concern
of
this
multi-linked
subnet
and
then
the
other
change
is
like
last
time.
Yangrinkova
talked
about.
Well,
if
you
do
host
isolation,
you
may
use
many
more
interfaces,
which
is
true,
so
you
know
in
this
version.
We
add,
you
know
we
address
these.
You
know
the
impact
of
many
interfaces,
but,
generally
speaking,
this
is
a
in
a
public
access,
wireless
environment.
R
I
Hi
jared
moch-
normally,
I
would
say
akamai,
but
in
this
case
I'm
going
to
put
on
my
I
actually
own
a
fiber
to
the
home
isp.
So
in
this
case
I've
just
started
to
enable
v6.
For
my
end,
subscribers
would,
in
this
deployment
case,
would
I
now
need
to
allocate
a
lot
more
v6
space
to
them,
because
in
their
home
network
they
would
need
to
allocate
a
slash,
64
per
device.
R
R
I
think
my
short
answer
to
your
question
is:
if
not
from
our
job's
perspective,
all
we
care
is
that
each
whole,
its
own
project,
these
projects
can
be
six,
people
can
be,
you
know,
can
be
longer
or
shorter,
you
know
so
so
our
girls
perspective,
we
don't
really
care,
I
think,
but
you
you
probably
care,
I
think,
like
you
know
the
the
how
long
or
how
short
the
project,
I
think
that
depends
on
your
own.
I
Yeah
I
mean
I'm
specifically
thinking
about
the
embedded
devices
where
the
user
doesn't
have
a
lot
of
interaction
with
them.
So
you
take
something
like
a
printer
and
a
phone
which
might
suddenly
find
themselves
on
a
different
ip
subnet
from
each
other
and
the
the
end
user
is
not
going
to
be
able
to
print
anymore
potentially,
which
is
that's,
that's
the
type
of
thing
that
in
some
cases,
that
they
actually
phone
the
isp
for.
R
I
I
R
R
We
said
that
you
know
you
give
this
cpe
and
two
different
cpe
on
different
projects,
but
what
we
are
not
saying
is
that
inside
your
home,
you
also
put
it
really
into
computer
into
a
different
subnet
again,
you
know
what
we
are
proposing
here
is
that
this
host
isolation-
it
has
a
certain
scenario
it's
suitable
for,
and
it
has
a
certain
scenario.
It's
not
suitable
for.
So,
if
you
are
talking
about
the
holes
inside
your
home,
we're
clearly
saying
our
journey.
You
know
this
is
not
a
scenario.
D
E
I
R
Well,
okay,
so
if
I
understand
your
question
correctly,
you
are
asking
well,
there
are
two
holes
they
now
they
are
already
separated
in
different
subnet
and
also
separated
in
different
l2,
and
if
they
have
duplicate
address,
you
know,
is
it
okay?
Our
answer
is
it's
okay,
because
it's
not
only
in
difference
of
that,
it's
also
in
different
links.
So
you
know
if
the
address
is
duplicating
it's
okay,.
M
Hey
learn
something:
what
do
you
mean
there
if
their
address
is
duplicated?
It's?
Okay?
If
a
packet
comes
from
the
internet,
which
one
of
the
two
gets,
it.
R
The
router
maintain
the
state
information,
so
we
are
talking
about
launcher.
We're
talking
about
link
local
address
can
be
duplicated.
M
I
think,
mostly
with
regard
to
wind
right,
so
the
drop
states
that
it
when
basically
allows
allows
that
solves
most
of
the
problems,
because
there's
no
there's
no
attacks
that
can
be
mounted,
but
I
think
you
should
note
that
the
hosts
themselves
conduct
the
infrastructure
by
applying
for
many
ip
addresses
right.
M
So
that's
one
thing
right:
if
I,
if
I'm
a
host-
and
I
create
a
million
ip
addresses-
that's
pretty
easy
for
me
and
then
the
router
will
fall
over
and
it
run
out
of
nd
cache
space
right
and
it
run
and
because
a
host
has
to
register
the
address
right.
So
that's
something
where
the
host
can
attack
the
infrastructure,
which
is
not
possible
in
in
a273,
it's
not
possible
in
prefix
per
host
because
the
the
reader
can
just
throw
away
all
those
entries.
It
doesn't
care
right.
M
R
I
don't
fully
understand
you
because
in
this
dual
we're
actually
kind
of
like
you
know
making
it
even
more
stricter
than
ap273.
So
if
827c3,
you
know,
if
you
can
accept
that,
then
it's
definitely,
you
know
in
our.
We
will
not
have.
M
I
R
The
only
difference
is
that
you
know
in
your
configuration
in
your
configuration
you
make
sure
that
you
know
you,
for
example,
using
either
a
private
v-lang
or
something
you
make
sure
that
the
holes
are
not
only
on
different.
You
know
with
different
projects,
but
also
you
know
isolated
in
l2,
so
the
router
doesn't.
M
N
M
You
don't
have
it,
but
I
just
read
it:
okay,
if
the.
If,
if
the
host
doesn't
need
to
register
it's
ip
address
with
the
router,
that's
fine,
then
I
have
an
objection.
I'll
I'll
go
back
there.
D
So
we'll
have
more
comments
on
the
bus
in
a
few
weeks
eric,
I
believe
or
I'm
sorry,
mr
g,
I
believe
it's
your
turn.
M
R
Okay,
I
will,
I
will
check
with
you
offline,
because
again
you
know
from
my
perspective,
the
host
only
register.
So
maybe
the
text
is
not
clear
and
we
will
clarify.
A
A
B
A
S
Thank
you,
hello.
Everyone
thank
you
for
giving
me
the
opportunity
to
talk
about
requirements
to
multi-domain
activities
on
the
network.
This
is
the
first
time
to
present
it
in
this
ops
next
slide.
Please.
S
Yeah
first,
I
think
some
basic
viewpoints
about
ipvcs
only
firstly,
fpv
only
is
the
ultimate
stage
of
ipv6
development.
Second
seminar
ipv4
service
should
be
considered
seriously
for
this
deployment
to
show
to
ensure
the
service
of
the
service
continuity
of
to
accept
the
global
activity
for
internet.
Suddenly,
the
deployment
is
ready
to
activate
traffic
ratio.
When
fpvc
usage
will
increase
to
a
certain
limit,
it
would
be
better
to
consider
activities
only
last
night,
please.
S
S
That
is
yes,
different,
es
mists
of
different
scenarios
such
as
metal
network,
backbone
network
4g,
network
authority
of
biology
map,
mobile
call
cloud,
be
the
central
network
and
for
for
the
same
operator
different
as
most
likely
are
augmented
by
different
departments
or
institutions
and
using
different
routing
and
security
policies.
S
So
the
question
here
is
why
we
talk
about
advances
only
for
multi-domain
network.
Actually,
we
understand
issues
when
planning
activities
only
in
our
skill
network,
although
there
are
many
activities,
only
techniques
available.
Let's
not
please.
S
I
will
give
an
example
to
show
the
problem.
The
first
one
is
partial
division.
Only.
I
will
use
for
six
foot
to
illustrate
the
problem,
but
I
do
not
mean
that
it
is
not
a
good
solution,
other
country.
On
the
contrary,
I
think
it
is
very
successful,
for
it
has
been
adopted
by
several
operators
or
phone
network
which
is
composed
of
mobile
coils
and
back
bonus.
A
general
model
for
activities
only
is
to
use
four
six,
four
isolating
the
mobile
call
and
the
peanut
giveaway
is
known
as
the
egress
of
the
mobile
call.
S
S
In
order
to
extend
activities
only
to
cover
backbone
network,
another
option
is
to
develop
internet
at
the
edge
of
the
background
network,
with
this
model
the
fpv
from
uee
to
the
egress
of
the
backbone.
But
the
problem
is
unnecessary,
static.
Winding
all
the
traffic
to
ipv4
networks
will
go
through
a
connected
way
and
will
goes
a
long
way
wrong.
The
data
paths
likely
become
less
optimal
compared
with
relative
routing,
moreover,
session
level,
states
of
the
users.
S
Another
possible
approach
is
to
deploy
for
464
accelerated
mobile
club
and
deploy
other
activities
only
techniques
such
as
activity
for
overactivity
towering
in
the
backbone
network.
We
all
know
that
ipv4
overactive
stun
is
defined
in
rc5565,
where
the
network
core
is
where
the
network
core
is
activated
only
for
another
networks.
S
S
So
in
this
case,
ipv4
embedded
rpvs
packs
need
to
be
tracked
from
the
back
to
actually
four
packs
at
the
egress
of
the
mobile
coin,
at
the
peanut
and
then
transfer
transform
to
activistic
in
the
next
to
me,
the
number
of
conversion
gateways
will
increase
along
with
the
numbers.
Yes,
for
instance,
we
can
see.
We
can
see
that
there
are
three
ipv4
vc
convert
gateways.
S
Obviously
too
much
ipv4
at
least
leads
to
complexity
of
the
network,
and
the
keypad
is
increasing.
Therefore,
additive
6
only
should
spike.
I
think
I
think
it
should
be
urgent
need
for
marijuana.
Is
his
own
solution
to
eliminate
unnecessary
conversion
functions
and
improve
the
date
forwarding
efficiency
next
slide.
Please.
S
Okay,
so
here's
a
general
procedure
when
ingrass
pe
received
ipv4
packet
from
client
facing
interface,
that's
turned
towards
remotely
full
network.
The
ingress
key
must
transform
the
ipv6
package.
It
might
be
the
fpv
for.
S
Pe
another
case
is
that
ipv4
package
may
be
transformed
into
a
v6
pad
in
ue,
as
done
by
464
athlete
before
the
before
we
reach
the
edge
of
the
network.
In
this
case,
the
ingredient
pe
received
activist
package
will
look
up
the
tax
destination
address
and
forward
the
packages
in
the
egress
pe.
So
the
two
model
here
next
one
please.
S
Okay,
thank
you.
So
much
network
should
meet
following
scenarios:
ipv6
user
to
ipv4
servers,
hosting
cloud
data,
centers
ipv4
user
to
activate
four
servers:
ipv6
user
to
ipv6
servers
and
the
dc
to
data
center
for
data
center
communications
and
also
provide
service
for
traffic
for
the
old
network
and
ipvs
is
only
epgp
at
peering
in
internet
chinese
pond
and
the
5g
transport
service,
sd1,
etc.
Various
scenarios
should
be
considered
left
side.
Please.
S
S
The
second
one
is
ipv4
services.
It
should
provide
activity,
service
delivery
and
there
should
be
no
passive
degradation
of
current
parents
accessing
the
remaining
fpv4
services.
The
third
one
is
energy
to
end
for
any
given
rgb
for
press
flow.
There
should
be
no
ipv4
fpv's
conversion
point
in
the
middle
of
the
ipv6
in
the
past
when
traversing
multiple
domains
domain
network,
in
other
words
ipv4
package,
should
not
appear
in
the
middle
of
the
activity
pass.
S
At
the
end,
at
the
egress,
the
pe
can
recover
rtv
fighters
by
reading
the
next
title
field
of
the
packet.
So
the
translation
that
you
have
this
later
mode
in
pvc
only
have
the
same
ability.
This
address
mapping
algorithm
the
fifth
requirement
is
controller
dependent.
We
all
know
that
computer
has
been
deployed
by
some
operators
to
provide
sd9
functions
in
the
case
of
ipv6.
S
S
S
The
next
requirement
is
high
scalability
and
it
should
achieve
scalability
simplicity,
high
availability
and,
when
p
precise
ipv4
features
at
the
item
at
the
edge
of
the
network,
the
positive
ipv4
releases
not
increased
linearly
or
exponentially
along
with
the
quality
of
human
traffic.
It
is
better
to
adopt
algorithm-based
imaging
approach
to
avoid
excessive
status
queries
at
the
eyes.
It
should
not
overload
the
existing
activities
resulting
table
requirement.
8
the
srvs
is
applicable,
as
services
can
be
supported
by
exerting
sih
in
translated
activism
package
or
increment
encapsulated
package.
S
S
So
this
is
the
job
about
the
requirements.
It
does
not
intend
to
design
a
new
transition
mechanisms
compared
with
two
years
ago.
The
situation
of
the
network
has
changed
a
lot,
especially
due
to
the
rapid
development
of
5g
and
cloud
computing,
so
we
should
examine
activities
from
new
perspective.
For
this
draft,
more
work
needs
to
be
done.
We
are
looking.
We
are
looking
forward
to
receiving
more
comments
and
suggestions.
We
also
hope
it
can
be
adopted
in
the
future.
D
Question
for
you
are
you
planning
in
one
of
your
slides?
You
mentioned
that
you
need
a
framework
for
how
to
handle
multi-domain
the
ipp
fixed
deployment.
Are
you
planning
to
propose
such
a
framework.
S
Yes,
well
so
actually
we
want
to
hear
the
feedback
from
vixx's
ops.
First
then,
based
on
the
feedback
comments,
and
then
we
maybe
consider
to
provide
a
framework
or
something
like
that.
S
S
Okay,
okay,
so
from
the
from
standpoint
of
operator
from
china
telecom,
we
hope
there
should
be
a
framework
to
give
the
guideline
of
ipv6
only
deployment
in
our
skill
network.
S
Actually,
we
have
done
some
12
in
some
field
trials
during
last
one
year
to
tell
the
result
some
cases
future.
H
H
H
Why
don't
we
cue
this
up
for
a
week's
discussion
I'll
make
sure
that
the
issues
are
tracked
in
the
tracker
and
during
that
time
we'll
ask
the
author?
If
he
wants
this
to
be
a
call
for
adoption.
E
Okay,
thank
you.
So
that's
about
a
presentation
about
a
measurement
about
ipv6,
essential
headers
over
the
internet
and
the
first
things.
I
hope
that
you
like
the
acronym
that
we
used
james
kind
of
a
fun
one,
at
least
so
my
name
is
eric
wink.
I
am
supervising
basically
a
student
in
the
master
computer
science,
which
is
raphael
and
another
supervisor
guy
is
justna,
which
is
a
phd
students,
also
at
the
university
of
the
ash.
So
the
next
slide,
if
you
don't
mind
so
thanks
to
jan
fernando
and
tim
right,
there
is
this
rfc
7872.
E
E
Right.
We
don't
use
the
same
system
and
no
need
to
say
in
this
room
that
the
suitability
of
extension
headers
over
the
internet
is
important.
If
you
look
about
the
six-month
working
group
and
we've
got
only
in
the
room,
we
know
that
many
drafts
are
talking
about
from
there
up
and
it's
not
only
of
biops.
E
E
We
were
unable
to
find
a
vm
provider
in
africa,
which
is
ipv6,
enabled
we
try
as
well
to
get
one
in
china,
but
it
was
a
little
bit
more
complex,
but
this
one
we
get
good
hope
as
we've
got
bruno
zeres.
Thank
you
fernando.
If
you
are
listening
to
me,
we
got
one
in
slovenia,
thank
you,
yan
and
one
in
the
land.
Thank
you.
Stefan.
I
sent
thank
you
sander
on
this
one
and
we
got
also
bangalore
and
others
next.
E
E
E
E
We
also
use
all
the
routing
other
times
from
the
zero
which
has
been
obsoleted
up
to
six
and
just
to
please
run
right
on
this
one
I
mean
anyway.
We
want
to
check
it
so
fragmentation,
both
atomic
fragments.
So
when
there
is
only
one
fragment
in
in
the
fragments,
I
would
say
and
non-atomic
so
where
we
have
the
first
fragment
offset
zeros
with,
but
with
the
more
bits
on
we
use
as
well
authentication
editor
marksman,
says
hey.
E
We
should
try
as
well
esp
that's
for
the
next
one
and
then
two
kind
of
new
ip
protocols
right,
the
no
next
header
which
is
defined
in
8200
and
one
which
is
defined
in
network
programming,
which
is
the
ethernet
payload,
and
we
just
want
to
check
what
they're
going
through
all
those
packets,
always
as
a
payload
in
this
payload.
I
am
using
the
students
actually
encoding.
E
E
But
the
most
important
is
not
so
much
about
which,
as
is
dropping,
is
more
about
the
result
themselves.
So
the
only
case
where
hope
biop
was
behaving
like
specified
is
when
we
were
using
a
unknown
option
with
the
bit
discard.
If
you
don't
know
it,
because
in
this
case
100
of
the
packets
were
discarded,
which
is
expected,
behavior
they're,
the
one
a
few
of
them
were
going
through
destination
option.
E
If
you
use
addition
option,
which
is
only
eight
bytes
long,
it's
basically
go
through
they
com
in
all
the
13
meshed
thing
they
were
always
traversing
the
internet
without
any
drops,
which
is
good
as
soon
as
you
go
to
16.
That's
still
good.
You
move
to
32
you're.
Only
93
percent
of
the
packets
are
going
through,
and
so
on.
It's
getting
worse
and
worse.
E
If
you
do
the
math
40
bytes
of
a
pv6
header,
plus
16
56,
this
is
less
than
64..
If
you
do
it,
40
32,
the
first
one
we
get
this
year
drop
is
72,
which
is
larger
than
64..
So
we
can
mostly,
yes,
no
proof,
no,
not
a
fact
right.
My
only
educated
guess
is
that
some
routers
have
got
a
look
right
of
64
bytes
and
if
there
is
larger,
they
drop
it
because
maybe
they
are
applying
a
layer,
4
access
control
list
on
the
transit
traffic
and
they
don't
know
so
they
drop
okay.
E
E
E
So,
obviously,
in
this
case,
a
couple
of
providers
were
using
internally
segment
routing
and
were
dropping
packets
coming
from
the
outside
with
it,
so
it
kind
of
works
and
all
the
rest.
One
two
three
and
ron
you'll
be
pleased
that
your
cr8
5
and
6
are
going
through
as
expected.
So
it
is
perfect
atomic
fragments
they
are
heavily
dropped
right.
30
percent
of
the
drop
happens
there,
but
normally
atomic
fragments
are
no
more
used
right.
So
that's
kind
of
okay.
E
E
E
Isp
that
drop
it
so
wait
and
see
on
this
more
details
will
be
coming
because
currently
the
heuristic
that
we
are
applying-
I
mean
I'm
not
that
confident
in
them.
So
we
are
trying
to
get
something
much
more
accurate
for
dropping
and
there
are
well-known
papers
from
caidah
and
address
on
this
next
slide.
E
E
Of
course
again
it
would
be
a
sampling,
so
nothing
really
scientific
but
always
said
if
you
know
about
a
vm
provider
or
you
are
happy
to
host
a
vm
running
linux
and
python
and
scrappy
on
your
rpv6
site
in
africa.
Please,
let
me
know:
okay,
while
we
are
here,
can
even
pay
you
a
bill
glass
of
wine
on
this
and
then
the
other
one
is
maybe
redo.
E
Something
like
jan
fernando
and
tim
did
not
only
sending
traffic
between
the
vantage
point,
but
also
sending
traffic
to
alexa
1000
or
some
bgp
prefix
and
see
what
has
changed
compared
to
six
years
ago.
That's
the
next
step,
so
stay
tuned
for
philadelphia.
We
should
have
more
results
there
and,
as
I
said,
we
really
need
to
get
more
information
about
the
s.
E
So
there
are
some
techniques
used
by
notably
about
kaida
and
others
to
do
it
and
that's
the
next
step
and
the
last
slide
before
going
to
bq
of
questions
which
looks
like
v6
extraction
that
the
chain
with
so
many
excession
there's
one
to
the
other.
That's
the
git
up
for
under
the
code
right
master
student
quality
code
right
so
but
it
works.
Okay,
I
see
some
questions.
D
I
actually
see
quite
a
few
questions,
but
I'm
afraid
we're
going
to
have
to
take
this
list.
D
H
Okay,
we
wanted
to
outline
the
procedures
for
this
working
group.
We
had
some
discussion
about
what
the
role
of
a
chair
is
when
something
can
proceed
when
it
can't.
So
we
just
wanted
to
take
some
of
the
ambiguity
out
of
it.
Next
slide,.
H
H
Posting
an
internet
draft,
the
file
name-
should
be
draft,
your
name
v6
ops
and
your
topic
post,
a
message
to
the
v6
ops
mailing
list,
inviting
review
and
the
chairs
will
determine
whether
the
draft
is
in
charter
and
our
charter
can
be
found
at
this
website.
H
H
Well,
you
must
have
a
new
or
updated
in
charter
draft
and
you
must
have
at
least
some
discussion
on
the
mailing
list
meeting
time
allocation
rules
drafts
that
have
been
adopted
by
the
working
group
get
time.
First.
Individual
submissions
next
chairs
may
deny
a
request
for
meeting
time
if
the
re
three
requirements
haven't
been
satisfied,
if
there's
no
time
on
the
agenda
or
if
the
draft
has
previously
failed
to
call
for
adoption
now,
this
last
bullet
is
important.
H
If
you're
asking
for
a
call
for
adoption,
you
probably
want
to
have
some
review
on
the
mailing
list.
Early
clean
up
some
of
the
issues
that
came
up
because
then
you,
you
know
you
must
get
your
request.
Shares
may
not
deny
a
request
for
any
other
reason.
H
That's
to
say
I
can't
say
you
know
you've
requested
agenda
time.
You've
met
all
the
pre-requirements.
I
don't
like
the
draft,
so
I'm
not
going
to
give
you
time
on
the
agenda.
No,
I
can't
do
that
as
a
chair
next
slide,.
H
Calls
for
adoption
again
three
requirements
are
the
same:
a
new
or
updated
in
charter
draft.
Some
mailing
list
discussion
chairs
may
deny
a
request
if
the
pre-requirements
haven't
been
satisfied
or
if
the
draft
has
previously
failed
to
call
for
adoption
chairs
must
not
deny
a
request
for
any
other
reason.
H
I'm
sorry,
no
yeah.
They
begin
to
track
issues
in
a
transparent
way.
So
we
know
what's
outstanding
next
slide.
H
Now
the
chairs
judge
consensus
to
adopt
if
they
they
adopt.
If
the
working
group
agrees,
they
want
to
address
the
problem
if
they
agree,
the
approach
is
reasonable
and
if
all
blocking
issues
in
the
tracker
have
been
addressed
to
be
adopted,
you
only
need
to
be
good
enough
to
be
a
starting
point.
You
don't
need
to
be
good
enough
to
publish
and
in
the
discussion
on
a
call
for
adoption
would
appreciate
it.
If
you
posted
technical
issues,
only
plus
one
support
or
opposition
doesn't
count.
H
Those
kind
of
messages
you
know
won't
make
the
way
into
the
tracker
and
won't
have
much
of
that
chairs
must
recuse
themselves
from
consensus,
calls
on
drafts
that
they
have
co-authored.
So
if
I
co-author
a
draft
in
this
working
group,
I
must
recuse
myself
from
the
consensus.
Call
next
slide.
H
H
That
was
the
last
line.
Okay,
so
this
should
be
just
you
know,
rules
for
fair
play.
I
don't
think
it'll
be
very
controversial,
but
I
see
folks
at
the
microphone.
R
R
So
I
think
this
is
a
good
procedure
for
managing
the
job
submitting
to
the
working
group.
I
agree
with
that,
but
I
think
that
in
our
charter,
there's
also
a
statement
say
that
you
know
the
six
of
long
tools,
basic
input
from
the
operators
and
users,
and
if
we
insist
that
you
know
well,
if
you
want
to
provide
input,
you
must
have
a
duet.
R
R
If
you
stop
these
skills,
you'll
spend
half
the
time
the
you
know
patching
ipv4
on
ipv6
ipv6,
but
we
are
not
really
acknowledging
that
you
know
the
the
sharing
they
see.
So
I
wonder
if
it
makes
sense,
for
example,
to
to
add
one
more
session
in
each
ietf,
because
there
are
a
lot
of
interest
on
ipv6.
There
are
a
lot
of
challenges,
and
this
way
we
can
bring
more
input.
F
H
B
R
Yes,
thank
you
because
I
think
that
some
of
these
challenges,
for
example
various
ipv6
consoles
or
you
know,
ipv64
or
maybe
none
of
them
also
talking
about
it.
But
my
feeling
is
like
v6
up
will
probably
you
know,
have
more
talent.
I
think
like
this,
maybe
the
better
places
than
than
other
places,
and
I
think
like,
for
example,
if,
if
people
talk
about
their
challenges
here,
some
of
us
may
document
their
challenges
and
eventually
us
publish
the
job
rather
than
those
guys
because
of
the
operator
guys
they
may
not
have
the
time.
R
H
You,
okay,
let's
exchange
some
emails
on
the
list
about
the
logistics
of
this.
You
know
whether
it
would
be
an
interim
meeting
what
not
would
be
invented
invited
what
other
organizations,
but
this
is
a
good
idea.
H
A
Thank
you,
warren
kumari,
google.
So
I
don't
know
I
mean
I
I
like
this
general
idea.
I
don't
know
if
we
definitely
need
a
sec
separate
session
each
time.
I
think
what
it
should
be
is
what
many
working
groups
do,
where
they
use
the
first
part
of
the
agenda
for
adopted,
work,
etc,
and
then,
if
there
is
any
time
left
over,
we
use
it
for
new
work,
I
mean
for
some
meetings.
A
We
might
have
two
and
a
half
hour
sessions
and
only
15
minutes
or
20
minutes
of
working
group
business
and
it
seems
silly
to
have
a
separate
session
just
for
new
work.
If
we're
not
using
all
the
time
but
yeah
I
mean
to
me,
it
seems
like
a
reasonable
idea
to
have
some
set
of
the
meeting
time.
Put
aside,
for
you
know
new
science
project
ideas
or
presentations
on
other
stuff
and
v6
ops
was
doing
something
kind
of
similar
where
it
would
do
invited
presentations
from
like
operators
on
what
they
found
from
deployment
or
similar.
A
F
F
H
I
So
there
so
there's
one
other
thing
which
is
sort
of
a
well-known
secret,
but
not
a
secret
ietf,
which
is
the
iepg
meeting,
which
was
yesterday,
where
the
you
know
where
a
number
of
operator
network
operators
actually
do
show
up
and
talk
about
issues
as
well.
And
so
it's
sort
of
this
interesting
thing
because
actually
a
lot
of
people
even
in
the
ietf
leadership.
And
you
know
the
iab
and
isg-
don't
really
know
what
it
is.
I
And
so
it's
it's
something
where
I
think
there's
an
opportunity
to
also
leverage
that,
because,
for
example,
at
least
one
of
the
presentations
that
was
here
today
was
also
there
yesterday.
But
there
was
opportunity
to
have
a
bit
more
of
a
conversation
dialogue
based
on
the
the
agenda
time
that
was
available.
So
I
think
that
that's
another
opportunity
and
venue
outside
of
these
direct
working
group
activities
as
well
to
go
and
socialize
things.
H
D
Well,
so,
can
you
hear
me.
H
M
D
So
with
that
we're
at
well
within
two
minutes
at
the
top
of
the
hour,
I
guess
this
is
going
to
go
to
the
list
in
the
next
week
or
two,
and
so
thank
you
very
much
for
your
time.
Everybody.