►
From YouTube: IETF98-INTAREA-20170328-1300
Description
INTAREA meeting session at IETF98
2017/03/28 1300
A
C
C
C
E
C
We
have
your
name,
please
down
this
rocky.
Thank
you
much
much
appreciate
it.
So
the
agenda
set
quite
packed
we're
going
to
go
through
the
bashing.
Then
we
are
going
to
talk
about
the
liaison
that
we
got
from
my
triple
leader
to
that
one.
On
the
IP
over
partially
partition
links
is
Eric
in
the
room.
C
F
G
So
Francis
I
think
that's
a
very
fair
point
and
the
idea
is
like
to
give
him
a
shot
of
explaining
what
this
is
about,
because
this
has
been
like
kind
of
lingering
around
and
like
multiple
working
groups
and
so
multiple
chairs
and
interior
bottom
slot
request
for
this,
and
so
the
idea
is
to
see
if
there's
anything
there
or
not
like
from
the
draft,
it
doesn't
look
like
that
is,
but
I
think
we
need
to
hear
them
out.
Okay,.
D
H
D
Generic
UDP
encapsulation
is
now
a
working
group
item,
but
note
that
this
has
been
inherited
from
envio
three.
So
that's
why
we
didn't
go
in
the
normal
way
to
ask
for
a
working
group
adoption
and
we
started
a
working
group
last
call
on
them:
IP,
/,
intentionally,
potentially
partition
with
links
and
in
January,
not
on
January
nineteenth,
and
this
is
still
open
actually,
and
we
were
going
to
have
a
quick
discussion
about
it
after
the
presentation.
C
Okay
to
the
next
item,
exciting
the
liaison
right
so
next
item
in
the
agenda.
This
is
the
liaison
that
we
got
from
I
trip,
only
80
two
dot,
one
working
group:
it's
about
relationship
between
a
triple
hero
to
that
one
q
and
the
draft
that
we
have
from
from
Eric
that
we
have
reviewed
so
Glenn.
Would
you
please
give
us
an
overview
of
the
VA's
and
what
is
up?
Yes,.
I
Okay,
the
afternoon
folks,
my
name
is
Glenn
Parsons
I'm
with
Erickson,
but
in
this
capacity
I'm
representing
the
I
Triple
E
802
that
one
working
group
which
I'm
the
chair
of
that
group-
and
this
is
the
liaison
that
that
we
sent
to
I
Jeff
and
the
interior
working
group
as
a
result
of
us,
becoming
aware
of
the
request
to
make
this
document
draft
nordmark,
interior
I
ppl
a
working
group
document,
and
in
our
review
of
this,
this
particular
draft
talks
about
I
triple
e
8
0
to
l,
1,
q,
bridges
and
the
the
way
that
it
talks
about
it
is,
I
guess,
somewhat
confusing
to
us,
because
I
guess
there's
two
ways
that
you
could
write
the
standard.
I
You
could
write
it
as
a
profile.
That's
more,
you
know
focused
towards
how
an
operator
how
a
customer
would
would
deploy
it
or
you
could
I
mean
I,
guess,
write
it
as
a
standard
which
would
be
you
know,
mandatory
language
on
how
an
implementation
of
a
80
to
that
one
queue
bridge
would
be,
would
be
built
and
and
the
the
way
that
it's
written
right
now.
It
can
be
easily
confused
that
this
is
trying
to
be
a
standard
that
as
normative
constraints
on
implementing
dot1q
bridges,
and
we
don't
think
that's
appropriate
for
IETF
to
do.
I
I
Without
you
know,
conformant
language
on
802,
dot1q
bridges
are
built
in
such
a
way.
So
so
that's
the
the
high-level
ask
of
Ann
Arbor
reaction
to
this
draft.
If
you
continue
to
scroll
down,
we
have
additional
information
on
how
implement
8
how
this
could
be
described.
So
we
have
managed
objects
described
in
our
claws,
12
of
802
dot1q,
and
we
have
some
details
here
on
how
the
private
viewing,
how
we
think
the
private
VLAN
concept
could
be
implemented
with
direct
references
to
the
802
dot1q
standard,
and
we
offer
that
for
your
consideration.
I
J
I
think
so,
having
just
refresh
my
memory
of
this
particular
drafted,
it
seems
to
me
that
there
it
there
really
shouldn't
be
any
conflict
in
the
sense
that
the
it
should
be
clarified
that
this
recommendations
in
the
draft
are
simply
about
how
I
p
FP
before
and
ipv6
should
be
made
to
work
on.
You
know
on
this
type
of
network,
and
that
seems
to
me
that
something
that
the
ITF
should
define.
Yes,.
J
K
J
I
So
that's
the
problem
that
we
had
with
the
draft
in
our
reading
of
it
with
with
our
particular
rose-colored
glasses.
We
viewed
it
as
as
constraints
on
layer,
2,
80,
2
del
one
cube
bridges
Barrowman.
So
with
an
I
guess,
an
IP
with
an
ITF
rose-colored
glasses.
You
would
have
the
other.
So
what
we're
looking
for
is
his
language
to
make
it
clear
that
it
what
you
intended
so.
E
Yes,
I
Eric
on
my
toes
so
I
think
it's
very
good
that
you
guys
have
spent
time
on
figuring
out
what
the
actual
description
on
the
mapping
is
and-
and
there
are
two
must
or
must
not
in
the
section
that
talks
about
bridges
right
soon,
but
those
were
the
ones.
That
is
the
main
issue
here
and
then,
but
I
think
it
would
be
very
useful
if
he
can
take
the
text.
I
E
I
I
L
The
lido
there
we
go
if
it's
useful
to
you
and
it
may
be.
We
have
been
sort
of
as
an
appendix
to
the
liaison
the
details
of
how
you
configure
a
bridge
to
do
exactly
what
you
want
it
to
do,
and
we
thought
that
may
be
helpful
either.
You
know
in
your
appendix
or
whatever,
if
you
find
that
helpful
feel
free
to
use
it
steep,
attic
and
I
put
that
together.
I
C
D
C
A
lot
Glenn
than
everyone
else
so
definitely
there's
interest.
Then
there's
a
relevance
of
this
draft,
so
I
guess
we
are
going
to
first
of
all
adopt
it
and
then
we
encourage,
of
course,
the
author
to
take
into
consideration
of
this
fruitful
feedback,
but
we're
going
to
be
progressing
the
document
taking
this
feedback
into
account.
K
Okay,
we'll
start
by
talking
about
your
old
friend
eng
thing
is
a
lightweight
tool
that
we
use
to
probe
the
liveness
of
a
remote
interface.
It's
nice
and
lightweight,
there's
no
credentials,
no
logging
in
anything
like
that,
relies
on
ICMP,
echo
and
echo
reply,
and
it
doesn't
actually
exercise
the
interface
whose
status
you're
probing
the
echo
can
come
in
a
different
interface.
The
reply
can
go
out
a
different
interface
and
it
requires
reach
ability
between
the
probing
node
and
the
probe
to
interface.
K
That's
important
requires
reach
ability
between
the
probing
node
and
the
probed
interface
and
in
the
words
of
notable
sad.
So
let's
go
on
to
the
next
slide.
Let's
talk
a
little
bit
about
your
new
friend,
extended
ping.
Extended
pain
doesn't
require
reach
ability
between
the
probing
node
and
the
probed
interface.
K
However,
it
does
require
reach
ability
between
the
probing
node
and
any
interface
that
is
local
to
the
probed
interface.
So
exactly
to
mean
we
mean
by
that,
let's
say
I'm
on
a
node,
I'm
I'm
looking
at
a
router
that
has
a
loopback,
but
I
can
reach
and
a
bunch
of
other
interfaces
that
I
cannot
reach
with
ping.
I
cannot
probe
the
status
of
those
other
interfaces.
The
interfaces
other
than
the
light
loop
back
with
ex
pang
I
can
probe
the
status
of
any
interface
on
that
box.
K
Even
though
I
can
only
reach
the
loop
back,
so
X
painting
is
is
applicable
in
situations
where
the
probing
node
has
a
route
to
one
interface,
but
not
necessarily
the
probed
interface,
and
in
all
of
these
scenarios,
where
X
ping
is
applicable,
the
probed
interface
is
not
local
to
the
probing
node.
If
it
were
no
local,
the
local
node
would
have
a
direct
route
next
slide.
K
So
what
are
some
scenarios
in
which
X
paying
is
really
helpful?
Well,
let's
take
a
look
at
scenario:
1,
the
node
that
X
ping
is
running
on
supports
v4
a
router
loopback
interface
is
numbered
from
a
global
p
for
global
space,
but
all
of
the
other
interfaces
A
through
Z
are
unnumbered.
You
want
to
probe
the
inner
fit
of
the
status
of
interface.
A
well
can't
do
that
with
thing
you
can
do
it
with
X
pain.
Let's
take
a
look
at
scenario.
Two
is
very
similar.
K
The
probing
node
is
v6,
aware
of
the
router
that
you're
probing
has
a
global
global
v6
address
on
the
loop
back,
but
interfaces
A
through
Z
only
have
linked
locals.
You
want
to
probe
a
through
z.
You
can't
do
it
with
thing.
You
can
do
it
with
XP.
You
think.
Oh
yeah,
it's
true
if
it's
a
ula
I
just
used
link-local
as
an
example
and
now
on
to
the
next
slide.
The
third
example
X
paying
is
dual
stack.
It
supports
before
or
v6
arm.
K
All
router
interfaces
are
numbered
with
I
can't
read
the
slides,
Seaver,
v4
or
v6
doesn't
matter,
but
only
the
loopback
address
is
advertised
into
any
routing
protocols,
so
the
probing
node
doesn't
have
a
route
to
the
probed
interfaces,
but
it
does
have
a
route
to
the
loop
back
in
this
case.
X
thing
works
thing
does
not
so
next
slide.
How
exactly
does
X
thing
work?
Well,
we
have
two
new
ICMP
messages
and
extended
echo
and
an
extended
echo
reply
arm.
K
They
both
distinguish
between
the
destination
address,
the
address
to
which
you
send
the
ICMP
message
and
well
the
destination
interface
in
the
probed
interface.
You
send
the
ICMP
message
to
the
destination
interface
and
the
ICMP
message
includes
an
identifier
of
the
probed
interface
and
all
of
these
ICMP
messages
are
defined
for
ipv4
and
ipv6
next
slide.
So
let's
talk
a
little
bit
about
the
echo
request.
K
What's
different,
is
you
have
an
ICMP
extensive
extension
structure
that
identifies
the
probed
interface
so
you're,
sending
the
ICMP
message
to
the
destination
interface
and
that
extension
structure
tells
you
which
interface
the
probed
interfaces
next
slide?
Okay,
the
extension
structure
contains
one
or
more
interface
identifier,
objects
and
each
I'd
interface
identifier,
object
and
identify
the
probed
interface
by
index
name
or
address
when
the
inner
fire
interface
identifier
identifies
the
probed
interface
by
address.
It
doesn't
need
to
be
the
same
address
family
as
the
destination
address.
K
So
let's
say
for
a
minute
that
you're
sending
an
ICMP
be
for
a
before
message.
The
destination
interface
is
an
ipv4
address.
The
probed
interface
might
be
identified
by
ipv6
address
or
it
might
be
identified
by
its
MAC
address.
It
doesn't
necessarily
have
to
be
identified
by
an
ipv4
address.
This
gives
rise
to
a
whole
bunch
of
new
use
cases.
In
most
cases,
a
single
interface
identifier
is
enough
to
identify
the
probed
interface
in
some
corner
cases
you
need
to.
For
instance,
let's
say
the
probed
interface
is
identified
by
link
local
address.
K
Well,
two
interfaces
on
a
single
box
might
have
the
same
link
local
address,
so
you
might
need
to
interface.
Identifier
objects
next
slide.
The
reply
message
should
give
you
exactly
two
pieces
of
information:
the
operational
status
of
the
probed
interface
and
what
protocols
are
active
on
the
probed
interface
before
v6,
both
leader.
It
will
not
give
you
any
other
information
about
the
probed
interface,
not
going
to
tell
you
it's
an
admin
status
at
them
to
you,
forwarding
statistics
any
of
the
other
stuff
that
people
thought
might
be
cool
to
throw
in
this.
K
We
decided
not
to
for
security
reasons
also
because
we
don't
want
to
go
into
competition
with
SNMP
or
any
other
management
protocol.
This
is
just
a
ping
look
alike.
Next
slide.
Okay,
this
is
a
view
of
we
prototype
this,
and
this
is
what
a
view
of
X
ping
looks
like
you
know.
You
explain
something
you
give
it
the
destination
interface
and
the
probed
interface,
and
it
sends
back
yes,
I
did
get
a
response
from
a
probed
interface.
These
are
the
two
protocols
that
are
active
in
this
case,
we're
probing
by
whoops.
K
Oh
yes,
we
are
we're
probing
by
interface,
name,
we're
probing
we're
sending
a
packet
10,
10,
10,
2,
and
asking
about
the
status
of
GE
0,
0
0,
which
may
or
not
may
or
may
not
be
the
same
interface
as
10
10
10
2
next
slide.
Here
we
probed
by
a
link-local
address
we're
sending
the
packet
to
10
10
10
2,
which
in
this
case
was
a
link-local
and
we're
asking
about
another
interface
whose
link
local
address
was
at
the
ground.
Read
it,
but
whatever
it
is
next
slide.
K
So
there
are
some
security
considerations
here.
If
we
were
only
probing
by
address,
things
would
be
fine,
but
there's
also
this
ability
to
probe
by
name.
So
you
could
use
expain
to
discover
the
names
of
interfaces
on
a
box,
but
once
you
know
the
names
of
interfaces
on
a
box,
you
know
things
like
what
the
vendor
is.
Maybe
what
version
of
cody's
running?
Yes,
it
can't
go
back.
One
slide
shirt.
Oh.
O
K
In
this
case
on
this
one
particular
router,
there
is
only
one-
and
you
got
an
answer
if
there
had
been
to
you-
would
have
gotten
an
error
message
back
saying:
I'm.
Sorry,
there
are
two
to
such
interfaces
on
this
box.
Please
give
me
some
more
information,
so
I
know
which
one
you're
talking
about.
Okay,.
K
K
Anyhow,
going
to
the
next
slide:
okay,
so
if
you're
probing
by
name,
you
can
discover
names
and
then
you
can
discover
other
stuff,
since
if
you
see
the
name
is
FB
0,
0
0,
you
know
it's
a
Juniper.
You
know
that
it's
fast
ethernet,
so
it's
either
10
or
100
megabits.
You
probably
know
the
MTU
and
you
don't
want
to
discover
that
sort
of
stuff.
So
we
need
some
mitigations
next
slide
by
default.
Routers
disable
the
ICMP
extended
echo
message
you
have
to
enable
it
on
purpose.
K
If
you
enable
it,
you
enable
queries
by
type,
maybe
only
support,
queries
by
address.
Maybe
you
only
support
them
by
name.
Maybe
you
only
support
them
by
ifm
decks,
but
you
decide
which
ones
you
support
and
that's
a
manual
operation.
It's
not
by
default.
All
are
not
manual,
but
it's
a
configured
operation.
Then,
let's
say
you've
decided.
Yes,
I
will
accept
queries
by
address.
Only
the
next
thing
you
can
configure
is
I
will
only
accept
queries
by
address
and
then
only
if
the
source
address
of
the
query
is
from
these
particular
prefixes.
K
So
you
won't
accept
it
from
everyone
on
the
planet.
Anyhow,
next,
the
status
of
we've,
a
cup
we've
addressed
many
comments.
Thanks
to
jonathan
ludi,
we
have
a
complete
prototype
thanks
to
Reggie
Thomas,
we
have
two
new
co-authors,
Chen,
Ling,
cova
and
chris
leonard,
who
I
think
are
both
here.
Freeze
your
hands
stay
high,
and
that
brings
us
to
our
next
steps.
Well,
this
is
kind
of
a
dull
draft.
What
would
like
to
do
is
have
a
little
vigorous,
compassion.
Debate
marked
by
you
know
a
little
polarization
and
name-calling
and
gratuitous
scandal.
K
Keep
it
interesting,
no
just
kidding
guys
and
then
would
like
a
call
for
adoption.
I
have
by
hopefully
a
non
scandalous
question.
If
you
go
back
a
couple
slides
to
know
we're.
O
K
Sorry,
Eric
Osborne
level:
three.
Does
the
draft
talk
about
what
active
actually
means,
because
I
could
see
different
response
codes
from
different
vendors
where
sometimes
active
means
configured,
and
sometimes
it's
up,
and
sometimes
it's
hell
operationally
up
visit
to
find
what
that
means.
We
should
probably
be
a
little
more
careful
about
that.
We
should
probably
steal
the
definition
from
maybe
noob
to
oh
yeah.
E
K
G
Suresh
krisshnan,
I
have
one
concern
right,
so
we
already
have
like
heard
ligado
reports
of
icmpv6
messages
getting
filtered
everywhere,
like
you
in
like
stuff
like
packet,
two
bigs
getting
filtered
right.
What
are
the
odds
of
this
like
actually
making
it
on
the
internet
with
the
new
message
code,
like
a
message,
type
right,
Leon
I'm,
just
trying
to
figure
out?
How
do
you
see
the
adoption
of
this
happening
and,
like
the
second
follow
point
for
that?
G
K
Glad
to
run
it
by
v6,
ops
and
maybe
ops
area
working
group
as
for
being
filtered
the
target
of
the
target
of
this,
the
guy
who
should
be
doing
the
filtering
is
the
destination
transit
devices
shouldn't
be
filtering
this
at
all,
because
they
have
nothing
to
lose
by
passing
it
on
and
since
by
default,
it's
turned
off
the
transit
really
shouldn't.
Now,
yes,
it
might,
and
if
that
happens
well,
we
have
a
problem.
Q
K
Had
some
back
and
forth
about
this
in
one
version
of
the
draft
multicast
addresses
were
not
allowed
period,
then
we
decided,
let's
allow
them
exactly
the
way
current
ping
does
and
we
will
get
exactly
the
same
behavior,
I'm
not
sure
if
that's
a
good
thing
and
I'm
open
to
a
little
impassioned
debate
and
polarization
around
that
yeah,
because
you
can
reply
for
each
and
every
interface
leveling.
Member
of
this
group
right,
let
me
pull
you.
Let
me
pull
you
personally
for
an
opinion
which
side
do
you
fall
on?
Q
E
K
K
E
C
C
R
K
C
C
A
C
C
N
The
goo
draft-
this
is
a
generic
udupi
encapsulation.
This
was
made
a
working
group
item,
I
believe
in
soul.
Since
then,
we
had
one
pretty
good
review
from
Marie
cuchara
way.
This
result
in
the
several
editorial
changes.
Few
clarifications:
we
substantially
cleaned
up
the
normative
language,
a
must
in
the
sheds
and
such
I
am
asking
here
what
the
next
steps
are.
Specifically,
what
do
we
need
to
get
this
into
working
drift
last
call
next
slide,
and
so
the
related
draft
is
cool
extensions.
N
This
defines
kind
of
the
basic
set
of
first
initial
options
for
goop
security
check,
some
goo,
and
things
like
that.
So
we
haven't
updated.
This
version
recently
add
this
I
think
we
mentioned
about
getting
this
to
be
a
working
group,
vitamin
int
area,
but
we
would
like
to
get
that
into
the
interior,
also
next
life
and
lastly,
I
la
this
identifier,
locator
addressing
I,
think
I
presented
this
a
couple
times
in
this
group
and
various
other
groups.
So
since
so
we
added
Peter
Lenkov.
N
As
an
author
from
facebook,
the
there
is
discussion
about
applicability,
ideas,
5g
IP
work.
There
is
one
issue
that
was
raised
to
that
I
think
we
want
to
address.
Basically,
we
have
a
checksum
neutral
mapping
as
part
of
ila
in
order
to
retain
checksum
correctness
after
translation,
and
we
also
have
some
other
bits
defined
in
the
identifier
and
the
question
was:
do
we
need
to
do
that
at
all?
N
C
J
A
lovely
one,
one
thing
about
ila:
that
sort
of
just
really
just
realized
recently
is
that
I
think
in
at
least
in
its
basic
format,
only
essentially
deals
in
128.
It,
basically
only
can
it
provides
mobility
of
individual
IP
addresses,
and
so
you
know,
while
we're
on
the
subject
of
applicability,
we
should
say
that
that's
not
recommended
for
assigning
IP
addresses
to
general
purpose
hosts.
So
if
we
take
on
this
document,
we
should
have
applicability
statement
to
that
effect.
N
So
we
did
define
the
primary
use
case
of
the
vile,
a
that
that
is
correct,
wasn't
intended
to
be
for
the
general
purpose
case.
We
are
deploying
it
and
data
centers
now,
but
it
is
specific.
What
we're
doing
were
actually
did
not
even
addressing
hosts
it's
actually
addressing
device
or
devices
or
tasks.
So
there
is
a
little
bit
of
granularity
there.
So
we
can
clarify
that
yeah,
okay,.
G
Suresh
krisshnan
the
80
hat
on
one
of
the
things
that
I
think
we
talked
about
a
bit
earlier
and
with
Ali
as
well.
Is
that
like
how
much
follow
a
book
is
there
for
this
right?
So
is
there
like
more
work
to
be
done
pilot
and
the
answer
is
usually
yes
right,
because
this
works
fine
in
the
data
center,
like
where
things
aren't
there,
you
control,
but
doing
more
stuff
requires,
like
other
stuff
to
be
done,
and
the
interior
the
way
it's
shattered.
G
G
So
that's
like
it
just
a
logistical
thing:
okay,
I
read
the
graft
and
I
don't
have
like
many
issues
with
it,
but
that's
like
this
certainly
missing
functionality
that
to
make
it
work
on
the
internet
right-
and
I
don't
have
an
answer
yet,
but
I
don't
think
you're
ready
to
make
the
call
for
this
document,
but
for
gu
x
yeah
go
ahead
with
a
call
whatever,
but
fri
la.
I
think
we
probably
need
to
I'd
rate
a
big
and,
like
there's,
been
like
no
discussion
on
this
for
like
five
months,
like
I.
N
My
question
right
now:
it
should
have
scripture.
We
just
adopted
more
I,
I,
didn't
hear
you
there.
There
has
been
there's
been
discussion
and
a
lot
of
other
contexts.
I
would
point
out
that
at
both
you
and
I
la
our
data
path-
and
that
implies,
we
need
the
control
plane.
So,
for
instance,
we
just
have
the
ideas,
lunch
off
and
I
think
that
would
be
great
for
the
mapping
subsystem,
but
thats
also
probably
be
in
the
routing
area.
If
I'm
not
mistaken.
So.
G
So
the
reason
I
was
okay
with
taking
on
GU
was
like
it's
kind
of
like
GRE
like
and
like.
We
did
GRE
stuck
here
right
in
interior,
because
there's
not
a
home
for
it
right
and
I
want
to
make
the
determination
independently
for
ila
than
for
goo
goo
right
like
even
though
they're
like
all
kind
of
related,
like
you
say,
but
I
just
want
to
make
the
determination
independent,
so
I'm,
fine
with
like
this,
not
a
home
for
Iowa
at
this
point
right.
G
So
I
would
like
you
to
still
continue
discussing
you
start
here,
but
I
just
want
to
have
a
better
idea
of
like
how
much
follow
book
is
there
to
figure
out
if
it's
going
to
be
here
or
should
be
a
ball,
and
it
is
going
to
be
a
ball
like
it's
going
to
be
here
or
and
routing.
So
that's
the
part
I
want
to
figure
out
before
taking
this
up
as
well.
N
But
we
did
talk
about
that,
I'm,
not
sure
that
it
would
need
it
son
working
group.
Maybe
that's
what
your
question
is:
I,
don't
I,
don't
know
how
much
more
there
is
so
like.
I
said
I'd
like
to
restrict
it
to
just
be
the
data
plane
protocol
and
then
yes,
we
need.
We
need
some
control
plane,
but,
as
Lorenzo
pointed
out
right
now,
I
la's
primary
first
application
is
going
to
be
in
the
data
center,
which
looks
a
whole
lot
like
an
encapsulation
in
some
sense.
N
G
C
C
Okay,
so
taking
out
their
chairs
hat,
presenting
here
on
behalf
of
my
co-authors-
hey
Charlie,
Perkins
myself
for
Carlos
Moya,
Dorothy
and
more.
So
this
is
a
draft
that
we
worked
a
few
ITF
meetings
ago.
Then
we
discuss
about
multicast
over.
I
triple
e
802
wireless,
the
different
issues
and
we
got
feedback
from
from
different
sides.
So
what
we
have
done
now
is
taking
a
whole
redo
remake
of
the
of
the
draft.
C
One
of
this
enterprise
level
access
point,
so
we
want
to
point
that
out
and
then
try
to
avoid
making
generic
statements
about
the
behavior
of
802
11
in
general,
and
we
also
want
to
start
working
on
on
recommendations.
So
we
are
calling
for
feedback
at
this
point
in
time
we
got
already
some
from
from
Michaela
burn
some
and
that's
appreciate
it
so
we'll
take
that
into
account
for
our
next
revision,
but
definitely
what
we
want
now
is
a
feedback.
M
Okay,
I'm
kind
of
another's
from
University
College,
remarried
and
I
in
present
in
this
work,
which
is
basically
just
to
make
a
community
aware
of
implementation
and
a
test
bed
for
validating
and
measuring
multicast
over
Wi-Fi
extensions.
So
this
basically
related
to
what
fun
colors
are
just
presented.
Could
you
click
a
little
novelty.
C
M
Yes,
a
very
quick
reminder:
the
legacy
one,
which
is
the
way
11
dot11
multicast
it's
transmitted,
is
basically
you
transmit
the
frame
using
a
low,
multicam
and
CS,
and
and
you
don't
have
any
any
proof
or
any
any
acknowledgement
that
the
multicast
ring
has
been
received
by
by
all
the
receivers.
And
then
there
are
three
different
improvements.
M
M
Sorry
several
times
in
trying
to
improve
the
the
rate
of
the
success
in
the
receivers
get
in
the
packet
and
then
the
last
approach
is
what
is
called
bloke
acknowledgment
will,
basically
you
send
a
block
of
packets
and
do
then
ask
for
the
receivers
to
say
what
I
did
receive
or
not,
and
then
you
will
transmit
those
that
has
not
been
received
to
the
procedures
that
didn't
get
the
packets.
So
these
are
the
forest
tensions.
M
M
We
had
a
test
wet
in
order
to
first
implement
those
extensions
in
Linux
and
then
to
validate
and
to
assess
the
performance
that
we
can
obtain.
So
we
have
this
test
bed
with
around
30
notes.
All
the
test
has
been
performed
using
channel
14
to
avoid
some
interference,
although
we
also
did
some
tests
in
channel
11
for
the
blog
and
management
approach
we
explore.
Different
values
of
m
m
is
basically
the
size
of
the
of
the
the
different
the
blocks
of
packets
at
you
that
you
sent
and
then
for
the
solicitation
retries.
M
We
say
we
try
with
different
values
of
number
of
retransmissions,
so
studio
is
just
one
single
transmission.
The
legacy
mode
want
you
to
meet
one
for
your
attorneys
for
packets,
four
times
the
packet
and
then
for
the
MSC
in
mcs
choice
for
the
legacy
with
it
set
that
to
24
megabit
per
second
and
for
the
different
extensions
we
use
54
married
per
second,
then
we
did
a
set
of
experiments.
M
The
first
one
is
basically,
we
have
a
multicast
CBR
traffic
over
ipv4
we
have
10
receivers
of
that
multicast
traffic
and
additionally,
in
the
network
we
have
lower
10
station,
stands
between
a
dense,
regular
unicast
data
and
then
under
this
scenario,
we
basically
look
into
two
performance
figures.
One
is
the
big
arrival
rate,
so
the
multicast
really
rate
and
another
one
is
the.
In
addition,
the
aggregated
data
traffic
that
is,
that
Taurus
able
to
to
transmit,
in
addition
to
the
multicast
track
here
we
have
different
results.
M
The
one
about
anxiety
said:
unsolicited
retries
with
our
series
should
be
basically
the
same
thing
that
the
yellow
one,
but
the
difference
because
of
the
MCS,
the
difference
MCS
that
is
being
used.
We
have
a
result
for
channel
11.
We
have
also
results
for
the
the
data
traffic
and
we
have
also
a
different
set
of
spare
amends
using
real
video
with
a
full
HD
video
with
a
very
brave
traffic,
and
there
we
also
evaluated
under
different
scenarios.
M
M
M
There
are
different
things
there,
but
in
addition
to
this,
we
can
use
the
test
fit,
for
example,
to
show
the
issues
that
did
in
the
draft
in
the
multicast
wrap
we
can
try
to
emulate
the
same
scenarios-
ipv6
signaling,
for
example,
and
to
say
to
see
okay
with
the
legacy.
This
is
the
the
performance,
those
are
the
issues.
Then
we
try
to
put
one
of
these
tensions
or
others
I
guess
these
guys
are
open
for
implementing
other
stuff.
M
We
consider
useful,
so
we
we
are
open
to
to
basically
encounter
is
the
test
that
isolated
or
it
can
be
mixed
with
the
real
traffic
or
other
type
of
traffic
through
the
traffic
that
we
can
put.
This
I
mean
you
can
put
any
any
kind
of
traffic
there
I
mean
we
can
use
synthetic
traffic.
We
can
try
to
put
real
traffic
regulations.
M
C
J
Good
yeah,
I
I
may
have
missed
this
and
you
may
have
said
it
already,
but
you
you
did
you
or
did
you
not
test
like
AC
performance
numbers
with
data
heavy
multicast,
but
are
you
have
you
been
testing
with
the
small
packets
that
are
ipv6
control,
packets?
You
know
NS
and
ra.
I.
M
J
I
think
that
the
scaling
characteristics
are
very
different
here.
You
can
like
you,
can
do
a
lot
of
optimization
by
like
gagra
gating,
the
packets
into
one
big
blob
and
maybe
using
fewer
TX
ups
or
whatever,
but
the
but
the
sort
of
control
traffic
is
probably
more
spiky
and
there
the
packets
are
really
really
small,
so
it
might
become
inefficient
because
you're
you're,
if
you
can
aggregate
them
into
bigger
packets
and
it's
expensive
iron
yeah.
M
M
Q
Okay,
so
remember
to
present
you
a
work
we
have
done
with
bazi
Tommy,
père
David
and
myself,
Eric
binky
and
the
last
one
is
during
my
last
name
so
quickly.
What
are
we
trying
to
solve?
What
you
are
trying
to
solve
is
basically
going
back
to
the
basic.
There
are
multiple
v6
races,
/
oath,
you
can
have
the
link
locale,
the
global
unity
Pro
global.
We
all
know
this
right.
Q
So
what
happen
in
a
network
like
this
one,
typical
multihoming
for
small
and
medium
business
will
get
to
egress
to
the
internet
and
you
receive
one
prefix
red
one,
prefix
blue.
So
on
the
top,
the
pc
select
the
source
address
blue
exit
by
the
blue
router.
Everything
is
fine.
You
get
the
story.
If
you
take
the
same
address
blue
and
exit
by
the
red
router.
Hopefully
the
red
is
P
is
doing
vcp
38
and
you
drop
the
packets.
Let's
thanks
god
for
the
internet.
That
is
pretty
bad
for
the
user.
Q
So
how
can
we
solve
this?
First?
They
are
some
enforcement
telling
a
if
you
use
the
prefix
red.
Please
use
the
next
up
as
being
right.
That's
lots
of
the
problem.
If
you
have
multiple
layers
of
routers,
because
you
can
only
select
the
first
one,
then
after
there
is
only
one
before
12
normally
so
you
see
never
the
problem,
but
it's
not
enough
assume.
One
of
those
link
is
linking
like
a
VPN
to
your
company.
Now
you
have
two
dns
server,
one
from
your
company
and
one
from
the
internet
and
I'm
working
for
cisco
systems.
Q
Q
And
the
other
issue
of
potentialities
is
that
if
you
have
different
kind
of
network
services
sing,
for
instance,
quality
of
service,
now
whether
they
are
done
by
traffic
engineering
or
you
are
doing
by
classification
and
and
this
we've
got
point,
for
instance,
there
is
different
level
of
services.
Now
you
can
say
based
on
the
source
address
you
have
to
prefix,
red
and
blue
the
red.
You
have
no
quality
of
service,
but
you
say
the
blue.
We
act.
We
classify
your
packets,
hop-by-hop
kind
of
lad
if
serve
God
pong
diffserv,
but
at
least
on
this
one.
Q
You
cannot
change
the
address
this
with
good
point.
We've
seen
this
morning
at
Mataji
can
disappear.
At
least
your
PVC
address.
The
source
does
not
change
on
the
path,
so
there
are
ways
of
using
the
prefixes
to
select
the
service
and
that's
basically,
what
oops.
This
kind
of
graph
wants
to
say
a
mr.
application.
Yes,
you
do
have
if
the
router
advertise
different
services
based
on
the
source
address,
you
may
select,
different
kind
of
quality
of
service
of
quality
of
experience
sounds
logical,
but
a
we
done
this
before.
Q
Q
All
of
us
whether
its
iOS
or
Android,
I've
got
phones
where
you
can
select.
This
application
may
use
or
may
not
use
the
C
learning,
because
it's
expensive
most
of
the
time
there's
a
reason
why
and
set
OS
vendors
deeply
do
it.
So
we
have
already
the
knowledge
that
application
by
application
you
can
use
till
I,
yes
or
no,
so
they
could
do
and
extend
it.
A
please
use
this
source
address
or
dead
source
address,
basically
to
select
the
network
service
so
now
I'll
throw
down
traversing
the
main
dvds.
Q
That's
not
a
new
concept.
It
came
from
the
milk
working
group
as
well.
You
get
multiple
interface
on
each
interface,
multiple
routers
and
Onision
routers
you
can
advertise.
Multiple
critics
is
so
very
multiple
perfect
is
you
can
get?
Dvds
are
the
set
of
a
PVD
ID
that
you
can
use
to
glue
all
the
set
of
information
think
about
at
the
first
level
as
an
index
into
the
database.
Q
It
links
the
prefix,
of
course.
The
next
stop
the
dns
information
that
I
call
you
a.
Can
you
look
for
inside
the
anus
or
the
outside
unit,
but
also,
for
instance-
and
there
are
other-
will
conclude
that
may
be
interested?
Is
there
a
captive
portal
here?
So
I
don't
start
the
application
in
the
exactly
to
go.
There
do
I
have
access
to
the
internet
globally
or
is
it
just
a
world
garden?
Just
look
about
the
movies
in
the
plane
on
the
train.
It's
also
important
to
know
right
and
you
can
also
get.
Q
Is
it
free
of
charge
on
or
three
of
charge
all
do?
We
communicate
the
set
of
information
and
we
are
in
2017,
so
it's
a
pv6
on
a
solution
right,
so
we
are
sending
what
you
call
to
bootstrap
PVD
into
the
router
advertisement.
It
is
important.
There
is
no
round
trip
time.
You
don't
need
to
do
more
thing
of
probing
of
fetching
information
in
the
bootstrap
PVD
you
get
enough.
We
assume
that
the
PVD
applies
to
all
prefix
information
option
into
the
array.
Q
If
you
want
to
advertise
different
PVD,
you
will
need
different
traffic
ceased
to
be
sent
in
different
Eric's
from
different
link.
Local
address
eric
is
it
ok,
I
can
wait
or
you
can
wait?
Okay,
so
let
me
finish
it
and
then
we
can
also
put
into
the
array
or
two
could
advance
information.
Condensed.
Of
course
right.
You
are
limited
to
the
1k
little
bit
more
of
the
array,
and
we
put
another
sentence
like
this
again:
that's
a
proposal
by
the
way
that's
0
dash
00,
so
I
guess
Eric.
O
J
Q
G
Kristen
so
having
seen
this
trainwreck
many
times
right
and
I'm,
just
wondering
how
okay,
how
are
things
going
to
go
differently?
The
sound
sorry
for
being
skeptical
here
right,
but
very
thing
the
differentiation
is
like
because,
like
you
only
hear
it's
like
smiling
as
well,
we
gone
through
this
like
stuff
ourselves
right
and
so
so
either
it's
a
question
of
like
how
much
you
wanna
cover,
for
example.
G
Already,
like
you
know,
audion
SS
is
there
right,
so
there's
gonna
be
other
options
coming
in
the
future
and
they
might
need
to
be
contained
or
not
so,
and
that
is
the
biggest
issue.
That's
sang
like
waters
in
there
because
like
if
you
have
a
tightly
scope,
small
set
of
things
it
becomes
doable.
If
you
want
like
a
broader
scope
of
things
to
go
in
there,
it
becomes
very
difficult,
especially
if
you
don't
know,
what's
gonna
go
in
there
right
so
yeah
mark.
Did
you
want
you
something
on
that
pleasure?
Okay,.
Q
So
on
this
one,
the
mechanism-
here
it's
pretty
simple
compared
to
other
one-
that
we
are
much
more
complex
and
what
we
sent
is
the
bare
minimum
right,
PVD
ID
next
slide
and
swing.
What
we
can
do
more,
but
all
the
same
time
they
could
their
the
string
is
extensible
messaging,
so
do
not
need
to
define
like
tlv
again
in
another
draft
for
the
tea
another
tlv
and
so
on
and
so
on,
and
whether
you
support
it
or
not.
It's
up
to
you
right.
Q
G
So
that's
how
like
we
start
off
as
well
right
and
one
of
the
things
that
came
up
was
like
people
said.
We
need
some
trust
in
the
person
telling
us
this
thing
would
make.
Then
you
had
to
add
like
a
signature
option
on
top
of
it
and
so
on
and
people
say
like
Oh
takes
things
are
like,
like
you
know
so,
19th
century.
G
Q
The
right
balance
now
regarding
the
security
next
slide,
I
get
more
on
this.
Just
on
this,
of
course,
there
is
not
enough
security
that
we
don't
know
why
we
get
the
second
stage.
Pvd
I
believe
this.
You
can
stand
the
interface.
You
know
you've
kept
it
for
that.
I,
don't
select
this
one
I
select
this
one
and
I
go
so
you
like,
for
instance,.
T
T
I
think
two
of
the
reasons
and
now
three
you're
standing
at
the
microphone
over
there
is
that
you
know
well
now,
there's
two
of
them
and
that's
the
the
the
host
vendors
are
very
seriously
looking
at
this,
and
this
is
something
that
I
think
is
a
not
that
they
were
ignoring
it
before,
but
I
think
it's
pretty
serious
this
time,
and
also,
at
speaking
on
behalf
of
my
company,
we're
starting
to
see
enterprise
networks
that
are
seriously
motivating
us
to
deploy
ipv6
only
we're
setting
that
I
think
ipv6
only
as
a
sort
of
a
predicate
for
this
to
even
really
be
thought
about
and
done
properly
and
take
off,
and
we
saw
data
from
this
morning
in
map
RG
that
three
percent
of
the
of
the
hosts
that
Tommy,
C's
or
v6
only
I
think
I
just
stole
his
thunder
I'm.
B
B
I
agree
with
all
that
I
think
some
of
what's
different
is
yeah.
We
do
have
more
few
six.
We
have
more
other
things,
I
think
what
it
comes
down
to
is
also
there's
more
of
this
multiple
PVD
stuff,
that's
really
going
on
and
that
we
want
to
be
able
to
actually
distinguish
not
just
between
oh
yeah
cell
and
Wi-Fi,
that,
like
we
figure
that
out
a
couple
years
ago,
but
they're
more
cases
which
we
have
these
two
different
uplinks
off
the
same
router.
B
B
That
is
something
we're
trying
to
open
up
the
main
points
that
you
have
multiple
pv
DS
that
you
know
about
explicitly,
and
so,
even
if
you
just
look
at
it
as
I
have
different
prefixes
and
I
know
that
these
different
prefixes
go
to
different
uplinks
and
that's
all
I
know,
there's
no
extra
metadata.
That's
still
really
useful
to
know
that
these
are
two
different
things.
P
Yeah
Tim
China,
I
agree
with
everything
marks
I
think,
maybe
is
that
the
time
is
now
is
now
much
more
favorable
for
either.
There
was
certainly
a
lot
of
feature
creep
in
the
earlier
myth
work
on
this,
but
there
is
also
some
good
ideas.
Things
are
distantly
determining
PVD,
so
you
don't
necessarily
have
to
have
it.
In
your
face
from
an
RA,
you
can
deduce
the
existence
of
a
PVD
from
a
set
of
information.
J
I
think
I
mean
to
sure
Ash's
point
I
think
that
optional
information
has
got
to
go,
but
but
but
there
I'm,
like
we
thought
about
this
as
well.
I
mean
I
I'm
not
one
of
the
others,
but
we
had
a
discussion
about
this
and
it
to
me
it's
extremely
important
that
the
information
that's
given
to
you
at
the
earliest
possible
time
is
contained.
It
is,
can
be
applied
to
a
given
PVD.
J
You
need
to
know
that,
and
I
think
we
said
we
should
limit
it
to
the
smallest
amount
of
information
possible
and
for
the
question
about
like
whether
tomorrow's
pony
should
or
should
not
be
part
of
this
particular
PVD.
I
think
that's
something
where
we
can
look
at
the
existing
option.
That
say
these
are
hereby
PVD
specific
and
these
are
not,
and
we
can
try
to
say
future
options
should,
in
their
documents,
say
whether
they
are
or
are
not
part
of
the
PVD.
That's
in
the
RA.
J
J
Q
E
Yeah
Eragon
Mike,
so
one
thing
that
isn't
so
I
know
that
people
have
done
stuff
with
different
classes
of
interfaces
on
phones
right.
But
one
thing
that
I
wonder
about
is:
what's
the
intended
scope
of
this
to
impact
all
of
the
other
applications
in
the
universe?
Because
if
you
have
this
star
comfort
to
be
useful,
it
means
that
everything
running
on
your
laptop
will
need
to
know
about
these
things.
To
be
able
to
give
you
the
selection
menu.
E
Are
we
have
on
the
phone
sure
right
and
does
it
mean
that
every
of
the
billion
IOT
devices
will
need
to
know
the
same
thing?
It's
like
they
don't
have
user
name,
but
where
does
this
lead
us
on
the
application
side,
not
the
network
stuff
right
because
that's
worthy
and
on
the
phones?
That
work
has
already
been
done,
but
are
we
going
down
a
path
where,
even
if
you're,
connecting
into
wired
Ethernet,
all
of
your
software
needs
to
know
about
the
stuff
for
it
to
work?
No.
J
D
S
I
was
going
to
enforce
Eric's
point
the
reason
the
you
can
cite
hear.
The
phone
example
is
that
there
is
a
very
simple
user
model
for
the
phone
and
who
so
it's
tied
to
something
Columbus
and
they
were
and
pay
their
is
not
the
same
kind
of
user
model
for
a
generic
network,
in
particular
what
you
said
about
selection,
it's
not
on
a
single
Network.
S
If
my
lap
dog
moves
here
or
they're
out
there,
I
mean
the
network,
two
will
be
multi
home
too,
so
am
I
supposed
to
put
not
only
one
line
that
says
all
these
of
that,
but
on
this
location,
do
this
on
that
location?
Do
that
I
mean
it
becomes
completely
event
unmanageable?
Okay,
so
I,
don't
think
having
the
ID
that
the
user
would
actually
poke
into
the
data
and
make
decision.
That's
completely
unrealistic
and
imperfectly.
Q
Right,
if
you
wait
for
23
slide,
there
is
a
project
called
neat
where
couple
of
people
here,
including
gory
at
least
maybe
others
are
working
on
to
provide
something
which
is
not
ready
directly
to
this
PVD,
but
leverage
them
with
all
about
it.
So
quickly
to
finish
the
two
three
slides
and
I
guess
would
be
enough,
so
that
was
yeah
the
bootstrap
DVD.
Then
we
get
the
second
stage
we've
put
in
the
ID
couple
of
proposal,
because
within
the
others
we
don't
agree
out
to
get.
Q
The
second
stage
could
be
HTTPS
and
test
part
of
it
where
we
get
some
security
because
with
what
we
get
by
HTTPS
or
body
in
a
sec,
we
get
the
prefix
again.
So
we
can
kind
of
close
the
loop.
It's
not
fully
secure,
but
it's
a
good
level
of
security.
Somehow
we
need
to
read
in
the
Refresh
to
accept
language
and
blah
blah
blah
blah.
We
can
skip
this
the
information
we
already
talked
about
it
about
the
captive
portal,
where
you
get
access
to
the
Internet.
Q
If
you
are
not
what
are
the
nap
timer
you're
using
that
64,
for
instance,
without
the
nap
timer
for
UDP?
So
that's
important
for
timing
again,
you
don't
want
to
take
this
information
back
on
your
choice
and
if
you
want
it's
okay,
we
could
even
enter
switch
capital
of
?
here
right
link,
this
to
add
pv
for
information.
Q
If
you
receive
one
PVD
from
one
mac
address-
and
you
only
receive
one
PVD
from
this
MAC
address,
then
you
can
make
the
link
the
dhcp
v4
information
send
by
this
MAC
address,
okay
or
into
the
big
JSON
on
the
big
takes
a
record.
You
can
say
a
the
dhcp
server
v4
is
this
one?
So
you
can
make
the
link
between
augmented
information
the
second
stage
PVD
with
the
rough
bootstrap
DCP
v4p.
Q
That's
one
way
of
doing
it
and
on
some
lines
like
3gpp,
we
have
no
choice
right,
that's
what
you
want
interface,
so
one
router
and
beyond
that.
So
there
is
running
code.
It
is
being
some
tests
done
by
to
me
like
a
ton
a
couple
of
times
ago,
when
you
get
two
routers
and
we
can
select
a
even
if
they
were
both
of
a
wife.
I
was
back
on
the
cellular
uplink.
So
when
the
cable
modem,
which
was
cheap,
was
done,
some
application
can
disconnect.
Q
So
this
over
Wi-Fi
rights,
which
is
not
oh,
no
matter,
this
Wi-Fi
connected
back
and
over
storyline.
In
this
case,
somebody
changes
where
the
Firefox
who
understand
this
particular
of
dvds
example
to
come
back
to
you,
Eric
and
Christian
on
this
and
the
need
project.
You
can
get
the
pointers,
there's
much
more
because
you're
not
need
selecting
the
source
address,
but
also
the
decision,
transport
protocol,
v4,
v6
or
as
quietly
I
completed
here
anyway,
not
to
run
on
my
time
next
step.
Q
G
Sarah
share
I
I
think
you
should
continue
discussion
on
the
list.
I,
don't
think
it's
ready
for
even
deciding
whether
it's
very
hard
option
or
not
so
I
think
continued
discussion
on
the
list
and
see
how
things
go,
because
I
don't
want
integrate
to
become
myth.
Okay,
so
if
there's
like
more
work,
that's
coming
up
like
I
would
really
like
to
keep
that
separate
again
because
I
don't
want
to,
like
you
know,
like
cherry,
to,
come
and
claim
area
on
this
scary.
F
G
E
Eric
I
would
suggest
you
go
to
the
apps
area
early
as
opposed
to
wait
because
I
suppose
I'm
finding
out
three
years
down
the
road
you
did
what
and
you
expect
the
applications
to
deal
with.
Go
there
now
all
right
and
see.
Do
you
think
that
this
is
reasonable
for
the
apps
there
already?
Some
of
them
are
already
doing
it,
so
maybe
they
think
this
is
helpful
or
maybe
they
think
it's
a
bad
idea,
but
finding
out
earlier
than
later
is
a
good
idea.
This
is
pretty
good
point.
Thank.
C
U
U
There's
a
comment
about
multicast
over
80,
2011
and
dave
says
my
comment
on
multicast
over
80.
To
was
that
I
think
that
draft
should
also
point
out
operational
issues
with
weird
ap
implementations
that
block
multicast
to
UDP
ports
other
than
a
specific
set,
for
example,
the
mdns
port
and
thus
interface
with
application
protocols.
A
protocol
designers
should
be
aware
of
that,
and
he
also
says
that's
80
2011
specific,
but
this
graft
seems
as
good
a
place
as
any
to
call
it
out
good.
T
Everyone,
hopefully,
my
co-author
Joe
touch,
is
in
the
Java
room,
etc,
etc.
I'm,
of
course
presenting
this
on
his
behalf
as
well
as
mine.
This
is
a
document
that's
been
around
for
quite
a
while,
has
had
a
lot
of
sparked
a
lot
of
discussions
and
debates
and
digging
through
various
piles
and
piles
and
piles
of
rfcs
and
documents
for
terms,
and
things
like
that-
and
this
is
a
status
update.
It
not
a
fun,
exciting
presentation
that
I
gave
on
this
couple
of
years
ago.
Why
isn't
this
working
ding
ding?
T
I
actually
want
to
go
to
slide
that
slide.
Update
summary,
oh,
this
is
really
awkward.
You
can
look
at
my
back,
so
there
was
us
there
was
a
version
for
that
was
supposed
to
just
before
the
deadline
and
then
a
version
five
yesterday
or
day
before,
with
a
bit
more
nits
and
such
fixed
in
it.
There
was
a
lot
of.
T
In
some
other
past
document,
clarifications
reorganizations
fleshed
out
some
of
the
you
know
existing
protocols
issues
you
know
where
we
can
identify
that
some
existing
tunneling
protocol
doesn't
match
this
tunneling
model
that
we're
trying
to
define
a
bit
on
load,
balancing
and
a
summary
of
recommendations
which
is
in
section
5,
and
that
relates
to
the
current
status,
some
of
which
I've
already
given
you
a
review
of,
but
the
real
question
that
Joe
and
I
would
like
the
answer
to
is
this
third
bullet
point,
which
is
what's
the
path
forward?
T
This
is
already
obviously
working
group
document.
If
the
working
group
wants
this
to
become
a
bcp,
we
will
do
the
extra
work
to
flush
out
section
5,
which
gives
recommendations
if
the
working
group
does
not
want
this
to
be
a
bcp
or
see
some
problem
in
it
getting
published
as
such.
That's
you
know
can
convince
us
that
it
would
be
a
waste
of
time.
We
will
delete
that
section
or
most
of
it
and
focus
on
the
informational
aspects
which
is
sort
of
you
know
pulling
together.
T
This
model
description
based
on
the
past
tunneling
that
we've
done
in
in
the
ITF
I,
don't
think
I
have
time
and
I
don't
have
Joe
here
to
go
into
details
about
this
version.
T
I
very
much
would
like
people
to
review
it
and
make
those
comments
on
list,
and
we
Joe
and
I
committed
to
you
know
within
a
couple
of
weeks
after
some
decision
on
bcp
versus
informational
will
give
you
the
next
revision
and
be
active
as
long
as
the
list
is
active
in
order
to
get
this
moved
forward,
hopefully
by
working
group
in
Prague
or
or
or
the
next
IETF,
but
we
kind
of
need
to
know
the
answer
that
fundamental
third
bullet
and
would
like
input
on
that
if
you're
willing
to
give
it
here
today.
Thank.
C
T
C
G
G
R
I'll
try
to
distribute
them
a
colleague
for
all
the
slides,
so
today
we're
going
to
discuss
the
IV
v10
and
for
the
first
time,
people
think
that
I
bb10
is
a
new
suit
in
a
new
structure
for
the
IP
version.
Actually,
it's
not
it's
a
solution
that
will
solve
the
problem
that
I'm
going
to
discuss
on
the
slides.
R
First
of
all,
all
of
you
know
that
the
address
space
for
ipv4
is
43
4.3
billion
unique
addresses
and
that
the
fast
growth
of
the
internet
started
to
make
this
address
space
to
be
exhausted,
and
actually,
let's
go
to
the
next
slide.
All
the
the
five
regional
internet
registries
after
iono,
almost
all
of
them,
announces
the
depletion
of
the
ipv4
address
space,
which
is
a
big
problem
that
we
are
going
to
discuss
next.
D
C
H
Okay,
okay,
well,
first,
first
question
why
we
are
presenting
this
here.
We
believe
this.
This
belongs
more
to
basics,
ops,
but
basic
subs
done
a
standardized
new
protocols
and
the
HCP.
They
understand
the
rise
new
option,
so
we
need
to
be
in
the
middle
of
nowhere.
So,
okay,
let's
see
I,
cannot
see
these
lights.
H
Basically,
the
point
here
is:
we
have
different
transition
mechanisms
that
use
not
64
Italy
either
is
stateless
or
stateful,
and
in
some
situations
you
may
need
to
have
several
net
64
boxes
and
we
need
a
way
to
tell
the
clients
how
we
need
a
way
to
provide
the
information
to
the
clients
about
good.
Are
they
the
prefixes
but
ipv4
and
ipv6
that
those
different,
not
64
boxes,
are
able
to
handle
for
them?
Basically,
what
what
we
see
is
the
need
for
ipv4
destination
routing,
depending
on
the
Box.
You
are
using
okay.
H
The
proposal
solution
is
basically
the
same
as
any
other
dhcpv6
option.
I
am
NOT
going
to
explain
the
detail.
We
don't
have
too
much
time.
I
think
it's
very
detail
it
in
also
in
the
in
the
draft.
We
have
updated
it
just
a
couple
of
days
ago.
We
have
aided
the
the
difference.
The
main
difference
with
the
previous
version
is
that
now
we
are
also
incorporating
suffix
information
following
the
nat64,
a
standard
and
well
basically,
that's
that's
it.
We
are
seeking
for
a
working
group
adoption
and
to
move
forward.
H
V
It's
gonna
see
Apple,
so
there
is
a
mechanism
for
detecting
or
being
formed
in
at
six
four
boxes.
It's
documented
in
RC
7050
and
it
uses
dns
yeah.
H
H
H
B
That's
really
a
case
in
which
we
really
want
to
know
like
I,
have
two
different
networks,
I'm
managing
and
that
it
becomes
that
bigger
Peabody
problem,
maybe
in
the
PVD
world
we
have
this
and
we
know
what
the
different
things
are,
but
just
adding
into
dhcp.
I
don't
know
how
that's
going
to
be
useful
on
a
given
host.
B
V
W
W
Thank
you
first
time
on
behind
my
co-authors
and
my
name
is
Dave
key
from
Bloomberg
and
my
code,
snippet
are
also
in
the
room.
Are
my
other
co-authors
schumann,
boys,
ansa,
not
post,
partum,
Oh,
Vic
and
neyo
la
hija.
Now
hearing
a
sorry.
W
So
we
want
to
fix
it
the
outline
busy.
We
want
to
talk
about
four
things.
Number
one
is:
what
is
the
problem
statement,
what
you
know
what
we
trying
to
solve
and
what
is
requirement
for
SDM
multicast
framework
and
number
three?
What
is
the
proposed
framework
and
then
we'll
talk
about
the
next
steps?
Right.
W
Excuse
me,
so
so,
first
of
all,
what's
what's
the
problem
statement?
What's
the
problem
in
today's
multicast
solutions
and
number
one
is
the
their
various
multicast
stability,
scalability
and
negative
impact?
Unicast
are
not
going
to
mention
complexity
of
the
multicast
multicast
protocol
procedures
right
and
some
of
them
some
of
the
multicast
protocol
procedures
are,
they
have
a
periodic
stay
refreshes
and
some
then
a
very
chatty
in
terms
of
for
the
receiver
join
us
join
and
leaves
now.
W
In
addition,
the
distributed
multicast
control,
States
Armed,
are
taken
away
from
the
unicast
resources
right,
t,
kempson
and
you
know
bandwidth
and
etc
and
number
two,
the
lack
of
a
uniform
multicast
admission
control
and
really
a
fine-grained
path,
computation
mechanisms
across
implementations
right
so
basically,
there's
no
unified,
really
unified
and
Mission
Control
for
sending
receivers.
There's
no
arm
really
fine
grained
bandwidth
competition.
If
you
want
to
do
a
diversity,
if
you
want
to
do
arm,
you
know
qos
computations
you!
If
you
want
to
think
about
latency
and
Jeter
all
those
things.
W
You
have
we're
out
of
time,
so
if
you
can
also
memorized
in
about
a
minute
I
guess
I'm
sure
I'll
be
quick
number
three
restrictions
and
constraints
that
carries
multicast
across
multiple
administrative
boundaries.
Right
it
just
doesn't
exist.
If
you,
you
know
wanted
me
to
end
to
end
multicast
delivery
from
your
dated
data
center
you
either
and
your
whatever
area
network.
W
Your
campus,
it
just
doesn't
happen,
then,
before
the
inability
of
the
operators
to
flexible
design,
multicast
requirements
and
delivery
and
doesn't
have
to
pace
down
your
Andalite
underlay
restrictions
and
your
operators
requirements
it's
that
level
of
abstractions.
Not
there
are
number
five:
the
lack
of
uniformed
security,
Marcus,
Marcus
security
policies.
Right
I
mean
in
not
to
mention
adidas,
but
with
this
malicious
or
not
on
those
type
of
multicast,
security
mechanisms
are
not
provided
in
a
uniform
fashion
and
number
six
is
really
of
multicast
telemetry
information
arm
in
a
very
fine
grain,
detailed
fashion.
W
So
what's
the
framework,
what
what
we
try
to
solve
I
mean.
What's
genesis
of
this
SDM
framework,
we
want
to
provide
for
a
decoupling
of
a
multicast
control
plane
from
from
the
routing
40
entries
and
unique
ass
right
and
number
two,
a
unified
control
playing
across
various
forwarding
element
in
combinations.
Whether.
G
W
So,
basically,
we're
here
to
solicit
solicit
feedback
for
the
you
know,
for
the
draft
and
obviously
is
or
drafting
this.
It's
really
well
and
we
like
people,
give
a
feedback.
Our
number
two
is,
you
know
we
want
to
obviously
add
more
content
for
tolerance
and
more
on
the
multi,
for
example,
multi
provided
PPI
interface,
etc,
and,
and
lastly,
we
you
know
we
just
want
to
make
sure
that
is.
This
is
a
you
know,
interest
interesting
topic
and
people
are
well
I
like
to
go
connect.
C
For
example,
thank
you
thank
you
very
much
and
thanks
to
all
the
presenters
and
people
are
participating,
apologies
for
the
people
that
did
not
get
chance
to
present,
but
clipping,
keep
in
mind
that
this
is
meant
to
be
a
discussion
on
the
mailing
list.
So
continue
the
discussions
and
many
points
that
may
be
missing
on
the
mailing
list.
Please,
and
thanks.