►
From YouTube: IETF104-DMM-20190326-0900
Description
DMM meeting session at IETF104
2019/03/26 0900
https://datatracker.ietf.org/meeting/104/proceedings/
A
A
This
this
meeting
is
being
held
under
some
IETF
rules
and
guidelines.
You
know
you
should
be
aware
of
the
participant
rules.
This
is
with
respect
to
IPR
policy.
With
respect
to
you
know,
participation,
other
rules
could
of
cowan
gauge
mint
rules.
Please
see
the
please
see
the
BCP
documents
and
blue
sheets
just.
B
A
Yeah,
if
I
were
to
mention
Suresh,
is
or
ad
area
director
salacious
unable
to
attend
this
time.
So
we
have
a
representative
here.
Welcome
blue
sheets.
Many
take
us.
We
need
one
minute
taker
I!
Think
generally,
like
you
know,
you
have
list
of
president
us
and
some
of
them
sent
their
slights
late.
I.
Think
at
this
time,
Carlos
is
that
you,
you
read
the
sense
slides
in
time
or
would
an
IETF
one
of
three?
A
A
So
quick
status
on
where
we
are
with
respect
to
the
working
group
documents,
the
on-demand
mobility
draw
document
you
know,
is
completed.
The
working
of
last
call.
You
know
we
have
forward
to
is
G
I,
it's
currently
under
I
years.
He
reviews
and
I
think
they're
a
couple
of
discuss
positions,
last
I,
checked
and
I
think
Danny
and
the
author's
are
addressing
this
comes
I.
Think
generally,
I
think
that
document
is
in
a
good
shape,
I
believe
and
we
have
the
DMM
distributed.
A
Mobility,
anchoring
document
I
think
this
was
in
the
working
group
for
a
long
time,
but
you
know
I
think
thanks
to
Carlos,
he
managed
to
do
a
good
rewrite
and
it
passed
the
working.
You
last
call
we
forward
the
document
to
we
had
you
know
about
to
forward
the
document
to
or
a
DAC
I
think
she
reviews
or
pending
and
my
co-chair
tapping
to
action
item
to
the
chair
review
and
after
that,
once
that
sanity
check
is
done,
we
will
forward
the
document
to
to
suresh
for
further
processing.
A
And
we
have
the
pv6
document
I
think
this
also
completed
working
a
blast
call,
but
this
mean
chair
review
is
pending
and
what
I
think
recently
there
has
been
an
IPR
declaration.
I
think
you
know
this
was
like
I
believe
it
was
posted
last
week.
We
are
trying
to
understand
the
implications
of
that,
because
this
IPR
has
been
around
for
a
long
time,
but
why
we
are
trying
to
understand
why
this
was
declared
so
late
after
the
you
know.
A
C
This
is
akbar
from
interdigital,
so
just
because
we
made
the
IPR
declaration,
but
actually
three,
if
you
look,
our
original
IPR
declaration
was
in
2012
on
the
individual
draft.
But
so
this
was
an
update.
So
we
actually
have
made
a
very
early
IPR
declaration
and
the
reason
for
the
update
was
that
the
patent
got
issued.
C
So
no
update
was
to
give
the
actual
patent
numbers
so
I'm,
not
sure
where
the
confusion
is
because
if
you
look
even
in
the
you
know
the
tracker,
it's
their
2012
Oh
3:19,
we
meet
the
declaration
on
the
individual
draft,
so
I
don't
think,
there's
anything
late
unless
I'm
missing
something
that
you're
trying
to
get
out.
Okay,.
A
Thanks
for
that
clarification,
I
II
was
not
aware
of
that,
such
that
details
that
there
has
been
a
posting
on
the
individual
ID
that
part.
You
know
this
needs
to
me,
but
in
general
I
think
anytime.
This
idea,
you
know
IPR
declaration.
We
we
activate
some
process
we'll
do
that
as
well.
Here
sure
I
just
want
to
clarify
else.
C
A
Thank
you
notes.
Thank
you,
I
think
the
next
one
is
SR
v6
mobile
user,
plain
I.
Think
I
haven't
seen
many
discussions,
I
think
no
I
understand
you
know.
There's
some
lot
of
good
work
is
going
in,
but
I
haven't
seen
many
discussions,
but
I
understand.
Six-Man
is
going
through
extensive
reviews
of
that
I
think
from
our
point
of
view,
I
think
for
us
to
really
complete
the
work.
A
We
need
to
understand
the
implications
of
six-mile
discussion
on
this
one
right,
I
think
there
were
a
bunch
of
issues
in
that
in
six
men
if
they're
unrelated
to
this
base
pack,
we
are
okay
with
that.
But
if
any
of
that
issues
impacting
this
specification,
then
we
need
to
be
aware
of
that
and
we
want
to
make
sure
those
issues
are
addressed.
I
hope,
that's
reasonable,
step
and
I
think
we
need
to
plan
for
readiness,
I.
Think
last
call
readiness
right,
I
think
we
need
more
participation.
A
D
D
A
A
Next
one
is,
we
have
the
FPC
document.
I
think
you
know,
I
think
this
is
in
a
somewhat
in
a
weird
state.
Actually
this
is
incredible.
Amount
of
work
went
in,
but
suddenly
it's
like
a
light
switch
overnight.
You
know
this
totally.
No
response
right,
not
much
happened,
right,
I,
think,
but
today,
charlie,
is
the
action
item
to
give
up
status
on
you
know.
You
know
what
you
know
currently
the
document
has
expired.
I
think
that's
what
we
can
forward
the
document,
because
it
has
a
boost
from
what
I
hear
understand.
A
I
think,
last
time
you
know
it
has
addressed
all
the
comments.
It's
just
you
know.
External
reviews
are
pending
right,
but
from
the
author's
point
of
you,
they
believe
it's
done,
but
but
the
document
is
expired.
I
think
that
we
need
to
see
a
live
document.
I
think
we'll
have
a
separate
discussion
on
that
and
the
files
you
use
a
plain
analysis.
Shinsuke
I
saw
so
you
know
last
time
it
got
adopted
as
a
working
document.
I
think
this
is
0-1
version
posted
in
March,
right,
I,
think
yeah.
A
A
F
Yeah,
please
yeah
thanks.
So
with
this
contribution
we
actually
want
to
complement
the
work
which
has
been
done
so
far
quite
some
time
on
the
mobile
data
plane,
but
actually
focusing
on
an
interface
which
is
not
well
represented
in
3gpp
was
in
charge
of
specifying
the
5g
architecture.
So
far
you
see
the
reference
points
between
a
mobile
device
and
this
correspondent
servers
in
between
you
have
the
data
plane,
which
is
under
control
of
5g
control
plane.
F
F
So
this
is
just
a
repetition
of
the
slide
about
two
exemplary
use
cases
which
we
presented
already
last
time
so
just
to
show
that
in
5g
it's
possible
to
have
multiple
video
session
anchors
mobility
anchors
and
if
you
have
multiple
sources
of
correspondent
service,
you
have
somehow
control
that
the
downlink
traffic
reached
the
most
suitable
anchor.
So
how
to
do
that
traffic
steering
and
treatment
off
of
data
plane
on
this
n6
interface?
That's
actually
what
one
address
another
use
case
is
more
from
the
automotive
space.
F
Then,
if
you
talk
about
low
latency
communication,
so
many
people
think
about
moving
some
application
services
to
the
edge.
And
now,
if
you
have
moving
vehicles
to
maintain
that
low
latency
communication,
you
main
need
to
relocate
from
one
edge
cloud
to
the
other
one
at
the
same
time,
using
the
same
IP
address
and
beaker
may
be
connected
to
a
service
in
the
central
cloud,
so
how
to
enforce
traffic
treatment
policies
on
this
insects.
Interface
same
applies
here
with
one
single
solution,
so
this
go
off
the
draft
that's
tricky.
F
I
can
adjust
yeah
thanks
for
tuning
yeah,
so
the
scope
of
the
draft
actually
proposes
leveraging
a
3gpp
component,
which
is
C
application
function.
It's
a
pretty
general
function,
communicating
through
the
5g
control
plane
using
the
3gpp
principles,
which
is
service
based
architecture
and
rest
kind
of
communication,
but
the
semantics
it's
a
pretty
high
level
in
terms
of
specification.
So
that's
what
we
want
to
address
mainly
and
then
second,
to
be
flexible
in
how
the
application
function
is
deployed
in
terms
of
transport
network
controllers.
That
cannot
force
on
the
southbound
interface
that
can
be
FPC.
F
So
the
draft
wants
to
address
the
use
cases
and
exemplary
operation.
In
second,
that's
main
focus.
We
want
to
focus
on
the
semantics
and
the
data
models
on
the
two
interfaces,
which
is
between
the
application
functions
and
the
5g
control
plane
and
second,
between
the
application
function,
wherever
it's
deployed
and
the
southbound
communication
to
the
distributed
data
thin
nodes.
F
So
the
status
of
this
draft
is
that
we
actually
try
to
connect
to
particular
some
some
people
before
the
last
ITF,
and
we
received
some
interest
and
valuable
comments
that
we
could
take
into
account
and
our
very
first
draft
would
be
published
in
September
last
year
the
draft
has
been
discussed
at
ITF
103,
where
we
had
more
comments
and
feedback
than
we
expected.
So
we
saw
that
there
is
quite
some
interest
in
this
work
and
also
after
IGF.
F
We
received
more
comments,
in
particular
for
the
application
with
the
Etsy
MEK
architecture,
its
integration
with
factory
system
and
in
particular,
also
for
your
automotive
use
case.
So
there
is
some
visibility
of
such
work
and
if
we
look
at
other
SDOs,
let's
see
Mac
as
well
as
3gpp
are
currently
looking
at
how
edge
clouds
can
be
integrated.
Well
will
5g
controller.
F
F
F
So
the
two
coupling
approaches,
I,
don't
make
it
three,
please
thanks
again,
so
I'm
not
on
the
application
service
side
and
the
data
network.
Everything
is
the
same.
So
the
difference
between
this
cafe
type
coupling
is
mainly
on
the
side
of
the
mobility
anchor,
which
is
under
control
of
5g,
controlled
name.
So
the
upper
picture
we
call
loose
coupling
between
because
the
UPF
side
of
the
policy
enforcement
for
traffic
treatment
is
basically
as
much
as
possible
independent
of
the
5g
control
plane.
F
F
F
F
So
last
but
not
least,
so
we
don't
want
to
explain
the
operational
and
semantics
aspect.
So
if
you're
interested
and
I
don't
want
to
contribute,
we
appreciate
so
all
of
that
is
currently
written
in
the
first
revision
of
the
draft.
So
just
for
here
we
want
to
show
one
example
how
that
could
be
deployed,
for
example,
as
so
addressing
one
of
the
current
studies
in
3gpp
and
also
an
Etsy
mech,
how
the
Mac
architecture
can
be
integrated
well
with
the
5g
system.
So
what
we
propose
here,
so
the
green
parts
are
actually
etch
cloud
components.
F
So
you
have
a
Mac
platform.
You
have
multiple
Mac
hosts,
McKay
and
Mac
B
and
per
the
Etsy
Mac
architecture.
There
is
a
Mac
platform,
a
nerd
which
is
under
control,
inter
alia,
of
the
traffic
treatment
policies
on
the
MEK
host.
So
we
thought
it's
wise
to
place
the
application
function
on
the
Mac
platform
manager,
which
is
then
under
control
of
enforcing
policies
that
apply
to
traffic
between
v2x
application
as
an
example
and
the
actual
vehicle.
So
as
one
example,
we
could
indirect
traffic
for
whatever
reason,
failure,
handling
or
load
distribution
between
UPF.
F
F
Also,
if
we
want
to
have,
for
example,
due
to
mobility
of
the
vehicle,
some
relocation
of
the
edge
cloud,
the
same
could
apply
whether
you
use
a
single
map
that
form
manager
or
multiple
ones,
each
responsible
for
each
MEK
site.
The
same
cook
could
apply
so
the
mobile
car
at
least
have
to
take
care
of
setting
up
a
new
UPF
serving
as
mobility
anchor
in
the
new
max
I'd.
And
then
all
this
relegation
of
traffic
you
could
handle
on
the
green
part
and
forcing
policies
on
the
m6
interface.
F
G
Stream
just
to
comment
that
done
on
this
stuff.
On
the
on
the
MEK
for
the
via
Kula,
we
may
I
mean
the
fighter
for
myriad,
which
is
a
European
play
that
I
guess
you
are
familiar
with.
We
are
doing
things
with
make
a
molecular
stuff
and
I
guess
we.
We
may
be
interested
in
evaluating
this
from
a
practical
viewpoint,
so
maybe
we
can
discuss
offline.
This.
F
F
H
F
F
So
in
particular,
if
we
talk
about
moving
from
one
administrative
domain
to
the
other
one
for
sure
there
will
be
two
Mac
platform
managers
which
have
to
talk
to
each
other
if
the
two
different
edge
cloud
sites
belong
to
the
same
administrative
domain,
that
may
be
dependent
on
one
single
platform
controller,
so
we
just
show
having
two
platform
controllers,
but
that's
not
not
not
binding
to
their
to
the
use
case.
So.
F
H
F
F
H
F
J
It's
so
it's
not
so
much
about
the
architecture
scalability,
but
if
this
is
about
low
latency,
then
all
of
these
paths
are
relevant
right
so
from
the
end
device
to
the
mobile
ID
host.
What
does
the
path
look
like,
so
obviously
the
fewer
hops
we
have
better.
How
many
handovers
do
we
have
every
time
we
cross
a
different
cell
site,
for
instance,
are
we
connecting
to
a
new
MVC,
so
the
architecture
from
that
point
of
view
may
make
sense,
but
obviously
what
is
the?
F
I
think
also
that's
an
interesting
question
where
we
may
not
have
one
single
answer,
so
it
depends
a
lot
on
how
edge
clouds
are
being
deployed,
and
that
differs
from
operator
to
operator.
So
some
operators
may
see
many
decentralized
edge
clouds
where
they
want
to
deploy
a
Mac.
Some
others
say
a
few
on
a
certain
geographical
region
are
fine,
so
that
all
has
impact
on
how
often
you
really
need
to
relocate
between
one
edge
to
the
other
one
in
terms
of
flow
latency.
B
J
Of
this
well,
all
of
this
means
it
has
to
be
super
simple,
so
this
collapses
into
very
simple
paths
like
from
my
device
to
the
host
I
have
like
one
or
two
or
three
hops.
It
has
really
really
compressed
so
I
think
that
that's
actually
a
good
design
attribute
if
you
consider
that
low
latency
that
will
enforce
the
conclusion
that
your
network
has
to
be
very
simple
and
very
flat
in.
B
J
F
So
if
I
mention
low,
latency
communication
I'm,
not
necessarily
implying
the
one
millisecond
kind
of
reference,
so
today,
anyway,
most
of
the
latency
comes
from
radio.
So
that
may
change
in
the
future
and
there
are
more
reasons
than
low
latency
to
put
particular
service
to
the
edge
which
is
reliability
and
offload
from
from
the
core
network.
So.
K
Enough
link
from
Nokia
just
to
answer
Charles
and
the
previous
question
of
the
scope
of
the
NIC
server.
Its
lateness
is
the
most
defining
parameter,
the
internet
latency.
So
what
we
are
looking
at
in
5g
environment,
it
is
some
something
below
100
milliseconds
to
10,
milliseconds
and
and
of
course,
that's
a
question
of
deployment
and
money.
Then
how
many
servers
and
where
do
you
locate?
You
see
it
can
be
part
of
the
run
system
or
it
can
be
next
to
your
base
station.
It
depends
how
deep
pockets
you
have.
L
Thank
you
very
much
for
your
work.
I
think
it's
very
important
that
you
consider
also
the
mistakes,
and
there
is
a
we
in
that
case
we
could
achieve
the
end-to-end
solution
and
the
you
also
manse
mentioned
that
I
saw
v6
to
support
and
six
so
could
you
fill
it
in
a
paper
a
bit
more
on
this
you're
thinking
on
using
a
highway,
six
could
support
and
six.
F
So
our
intention
is
to
keep
the
draft
more
or
less
independent,
off
or,
let's
say,
broadly
applicable
and
not
dependent
on
a
particular
protocol
that
you
use
for
traffic.
Steering
that's
why
we
said
it
just
depends
on
the
policies
that
are
enforced
between
AF
or
the
controller
and
the
data
plane
notes.
So
the
endpoints
that
enforce
traffic
treatment
policies
that
can
be
something
like
segment
routing
the
can
can
be
a
tunnel
that
can
be
ila
or
some
kind
of
other
or
locator
rewrite.
F
A
Yeah,
maybe
have
one
follow-up
session,
but
with
respect
to
standardization,
is
it
I?
Think
three
jamika
two
points
3gpp
doesn't
recognize
any
node
on
n6
today
right
does
this
spec
now
requires
that
the
Memphis
and
my
functions
without
the
cool
functions
require
a
new
interface
and
it
should
recognize
a
new
node.
In
addition,.
F
F
So
here
so
that's
the
map.
So
we
see
quite
some
gaze
on
the
authorities
with
what
3gpp
currently
discusses
and
they
recently
started
the
study
item
on
H
cloud
integration
with
5g
system.
So
we
may
have
some
valuable
input
here,
also
an
Etsy
mag
standardization.
They
look
at
how
at
CMAC
inter
works
and
can
be
well
integrated
with
the
5g
system.
So
we
see
some
turnigy
synergies
here.
How
much
impact
we
have
in
leaving
some
specification
out
and
given
to
other
stos
like
fruity
poppy
and
Etsy.
That's
something
we
need
for
now.
Okay,.
M
David,
like
apologies
for
a
light
question.
You
mentioned
the
gaze
on
work
with
3gpp,
which
I
think
is
very
important.
You
also
talked
about
moving
the
device
to
the
edge.
Would
that
impact
n3
in
any
way,
because
n3
is
already
set
in
release,
15
and
16
as
GTP,
so
we're
back
into
some
of
the
discussions
we've
had
in
XZ
Mac
about
bump
in
the
wire
for
gtp,
or
would
you
always
try
and
negate
that
so,
whatever.
F
Is
to
be
done
on
and
3
it's
under
control
of
3
to
BP
and
the
5g
control
plane.
So
I,
don't
think
we
have
interference
here,
because
the
policies
that
we
consider
for
intersects
traffic
steering
is
in
between
the
UPF,
which
serves
as
either
replica
sapphire
or
P
accession
anchor
and
the
data
network.
So
n3
is
between
the
radio
base
station
and
an
update,
classifier
UPF
sergeant.
I
think
this
is
more
complimentary
than
interfering.
N
Yeah
Cisco
Systems
very
good
document.
Do
you
have
any
correlation
with
the
networks
rising.
F
Of
course
needs
to
complement
each
other,
and
also
here
I,
don't
think
if
we
deploy
a
particular
automotive
slice
for
that
kind
of
service.
There
is
interference,
though
we
don't
address,
particularly
particularly
the
slicing
aspect
in
the
draft.
So
slicing
mainly
comes
okay
and
slicing.
You
may
have
an
Orchestrator
that's
in
charge
of
life
cycle
management
for
particular
slice,
but
also
you
have
to
have
components
and
force
traffic
treatment
policies
on
the
slice
data
plane,
so
I
think
all
of
that.
What
the
draft
talks
about
can
be
complimentary
to
a
slicing
approach.
F
N
F
B
F
Okay,
so
author's
intention
anyway,
is
to
revise
the
document
structure
because
in
the
serie
serial
version
we
have
a
long
introduction
telling
about
the
background
and
deployment
of
capability.
So
I
think
we
need
to
focus
more
on
the
actual
scope
of
the
draft.
After
understanding,
then
we
need
to
converge
on
a
suitable
notation
for
both
the
operational
aspects.
I'm
also
here
to
have
a
description
which
fits
into
a
3gpp
service
based
architecture
between
AF
and
5g
hydroplane,
and
then
also
for
the
semantics
and
model
description
that
apply
to
both
of
these
interfaces.
F
H
E
F
A
A
A
Next
one
is
Charlie
Perkins
on
FPC
status,
update.
Yes,.
H
Can
you
hear
me
yep
okay,
so
this
is
the
only
slide.
We
have
this
draft
on
control,
plane,
forwarding,
plane
separation,
it's
been
it's.
A
lot
of
work
has
gone
into
this
draft
and
it's
actually
in
relatively
good
state,
but
it
has
hasn't
been
commented
on
or
revised
since
that
summer,
so
it
was
requested
that
we
try
to
find
a
way
forward
with
this
and
I.
H
H
So
the
draft
is
basically
got
a
lot
of
functional
module
definition
and
information
model
in
the
main
part
of
the
draft
and
then
there's
a
relatively
large
appendix
for
a
yang
data
model.
That
goes
along
with
the
information
model
and
that's
a
lot
of
text,
and
it
actually
involves
using
some
existing
yang
model
and
defining
some
new
new
things.
And-
and
this
is
not
really
something
that
I
know
a
lot
about,
so
see
to
me
that
possibly
the
draft
would
progress
faster
and
get
more
comments
if
we
were
to
split
it
into
two
drafts.
H
Well,
two
points
of
view.
One
is
that
if
it's
in
saves
in
the
same
draft
and
any
changes
to
the
information
model
are
more
likely
to
be
immediately
reflected
into
the
yang
data
model
and
are
even
more
to
the
functional
interfaces
between
the
functional
modules
or
if
you're
splitted
in
the
two
draft,
and
we
can
focus
on
getting
I,
guess
the
more
important
content
out
and
then
get
the
yang
shortly
afterwards.
D
Hey
Charlie,
yeah
I
think
you
have
a
good
point.
Charlie
I
think
two
drafts
might
be
easier
to
get
reviewed
with,
but
the
issue
I
have
is,
and
the
person
who's
gonna
review.
The
yang
graph
needs
to
understand
the
other
drafts
all
right.
So
it's
like
a
slight
wane,
but
not
that
much
of
a
win.
So
one
of
the
things
I
talked
to
Sri
and
dapping
about
is
to
get
clarity
from
somebody
with
the
GPB
knowledge.
D
H
A
A
I
I'm
presenting
an
update,
it's
a
this
time,
British
a
lightweight
update
due
to
did
some
limitation
of
the
time
so
I
even
I.
We
had
a
good
feedback
from
the
reviewer.
Thank
you
very
much
so
what
we
had
update
it,
pseudocode
correction,
which
founded
by
kentaro
and
others
and
other
it's
some
clarification
text
for
the
define,
a
service
function
and
the
traditional
mode
and
ipv6
is
a
traffic,
and
the
argument
name
ask
more
session.
Other
IDEs
simplify
the
text
regarding
networks
rising
so
other
point,
its
name
in
compliant.
I
We
had
some
comment
in
previous
meeting
that
naming
is
not
appropriate
for
that.
But
in
terms
of
the
argument
we
have
no
idea
received
after
the
last
meeting
and
other
naming
is
a
t.
Dot
m
dot
team
up
Kentaro
had
suggested
us
to
use
a
t,
dot
end
of
GTP
fold,
em
daddy,
sorry,
other
kind
of
the
follow
the
other
the
function
name
instead
other
musa
feedback
from
the
viewer,
but
whether
to
support
dropping
scenario
which
is
demonstrated
by
a
Schmidt
in
last
IETF
Hakkasan.
The
drop
is
not
it
like.
I
The
packet
received
from
TTP
side
and
then
transmitted
to
the
SR
basics.
It
then
transit
again
to
did
before
it
enable
that
is
a
plane
pass
to
be
able
to
handle
is
our
basic
function
in
the
middle
with
a
mobility
passed,
so
that
requires
some
mapping
rule
change
country.
We
have
source
IP
address
of
the
gtp
you
into
the
seat
at
the
destination
ipv6
address,
but
it
it
need
to
be
in
the
source
side
and
with
the
UDP
for
support
to
reproduce
the
UDP
for
sport
in
to
keep
the
transparency.
I
So
please,
next
site,
so
I
have
a
Hawks
on
project
during
that
weekend,
so
we
had
the
taxon
project
to
talk
to
target,
to
implement,
to
function
from,
for
the
SR
basic
to
GTP
and
gdp
to
sr
basics.
Kota
is
now
available
on
github,
both
for
the
PvP
and
p4,
and
so
during
this
hackathon.
We
study
how
to
deal
with
ma,
yet
the
suggestion
to
support
the
open
studio
and
also
to
follow
the
argument
format
to
be
integrated
into
that
translation
scenario.
So
we
have
well
studied
what
we
missing
in
the
current
spec.
I
I
So
I
think
I
think
we
can
update
this
draft
based
on
this
outcome
and
take
away
in
the
next
meeting
next
division
with
the
draft.
So
next
slide.
Please
III
couldn't
mention
the
detail
of
the
new
mapping
rule,
but
you
can
you
can
read
and
chase
the
lab
at
all,
so
next
step
I
need
to.
We
need
to
handle
the
dbo
comment
code
correctly
and
deflects.
The
our
analysis
based
on
or
the
bayview
comment,
I
think
or
muscle
comment
come
from
the
unclear
techies.
I
So
then
we
can
I
think
we
can
deal
with
that
comment
with
the
improving
clarity
and
lead
ability
to
read
to
deal
with
the
add
comment
and
of
course
also
a
feedback
from
working
group
is
really
welcome
and
second
thing
is
update.
Mapping
rule
between
gtp
on
a
service,
as
I
mentioned
in
the
previous
file.
I
A
O
E
Eric
lang,
he
says
she
believed
this
accessibility.
This
you
showed
a
diagram
of
mapping
ipv4
gtp
but
addresses
into
SR
v6.
Is
there
a
mapping
for
v6o
basics,
yeah
bucyk,
for
what
DGP
gtp
encapsulation
did.
I
I
A
P
H
Numbers
of
aircraft
operating
worldwide
today
is
currently
in
the
order
of
tens
of
thousands,
perhaps
growing
to
hundreds
of
thousands
in
coming
years.
However,
we're
starting
to
see
unmanned
air
systems
and
personal
air
vehicles
as
anticipated
growth
in
the
near
future,
so
will
soon
need
to
consider
large
orders
of
magnitudes
in
terms
of
the
numbers
of
mobiles.
H
H
So
I
want
to
talk
about,
centralized
versus
distributed
mobility
management
in
centralized
mobility
management.
We
would
have
one
mobility
anchor
point
for
the
entire
worldwide
irrigation
environment,
so
you
can
see
where
I'm
going
with
this
you've
got
thousands
and
thousands
of
planes
all
grouped
around
the
same
ability
anchor
point
advantages
of
CMM
are
that
you
get
immediate
mobility
and
quality
service
signaling,
since
all
the
aircraft
are
serviced
by
the
same
mobility.
H
Anchor
point
disadvantages
are
that
your
scaling
limitation
is
not
only
in
terms
of
the
number
of
aircraft,
but
also
in
the
amount
of
mobility,
signaling
and
also
localized
mobility.
Events
cause
global
instability,
so
we
really
need
to
move
away
from
the
CMM
model
and
that's
where
we
start
to
talk
about
distributed
mobility
management
in
distributed
mobility
management.
We
might
have
many
regional
mobility
anchor
points,
perhaps
maybe
hundreds
or
thousands
of
them
to
distribute
the
scale
in
load
without
impacting
the
global
routing
system.
H
So
here
you
see
the
arrangement
that
we
advocate
in
the
draft
where
we
have
a
global
routing
overlay
system
that
runs
BGP
as
a
core
autonomous
system
with
the
mobility
anchor
points
in
spokes.
So
it's
a
hub
and
spokes
arrangement
and
this
the
the
mobility
anchor
points,
are
stub
autonomous
system
border
routers
in
BGP.
H
Now
the
advantages
of
this
is
that
it
distributes
the
load
amongst
many
Maps
scalable
terms
of
numbers
of
aircraft,
a
single
BGP
routing
core,
we
believe,
can
service
up
to
1
million
/
56
prefixes.
It's
also
scalable
in
terms
of
mobility
signaling,
because
the
localized
mobility
events
are
kept
local
without
causing
global
instability.
H
H
A
H
H
H
Today's
okay,
so
each
the
entire
system
is
going
to
support
a
very
coarse-grained
mobility
service
prefix
like
a
slash
30
to
each
one
of
these.
Thousands
of
writing.
Cores
is
going
to
maintain
an
independent,
BGP
routing
information
base
with
up
to
1
million
mobile
network
prefixes
Perotti
information
base.
So
we
tried
information
base.
It's
going
to
service
at
different
mobilities
group
prefix
like
a
slash
44,
so
the
mobility
anchor
points
are
going
to
appear
with
each
routing
core
and
apply
route
filters
so
that
mobile
network
prefix
registers
with
a
single
route
information
base.
H
So
if
we
had
a
thousand
routing
information
basis,
each
service
in
a
different
mobility
group
prefix,
the
total
system
should
be
able
to
support
up
to
1
billion
BGP
routes.
1
billion
aircraft,
with
their
their
individual
mobile
network
prefixes
Mobile's,
can
register
with
any
available
mildy
mobility
anchor
point
route.
Optimization
keeps
the
data
traffic
out
of
the
core
and
mobility
anchor
points.
Keep
mobility
signaling
out
of
the
court.
H
So
these
are
the
two
drafts.
A
simple
BGP
based
mobile
mobile
writing
system
for
the
aeronautical
telecommunications
network.
Is
a
working
group
item
of
the
writing
working
group.
It
also
addresses
the
distributed
mobility
management
solution
that
that
this
group
is
discussing,
and
then
we
have
the
scale
with
the
aggregation
for
overlays
using
the
border
gateway
protocol,
and
this
is
a
new
draft
and
it
talks
about
this
massively
distributed
mobility
management
that
I
referred
to
earlier.
H
One
other
thing
I'll
say
is
that
we
have
a
group
in
Austria
from
a
company
called
free
cuentas.
That
is
experimenting
with
and
evaluating
this
proposal
and
they
notice
that
with
BGP,
when
you
withdraw
our
route
from
the
routing
system,
it
can
take
several
hundred
milliseconds
for
that
route
to
go
away,
and
during
that
time
you
could
have
packet
loss.
H
If
the
previous
notes
didn't
remember
the
link
layer
address
of
the
next
hop
from
that
route
and
the
way
we
get
around
this
is
that
we
actually
have
our
some
hysteresis
in
the
system
that
allows
nodes
to
remember
where
that
mobile
node
was
before
the
route
was
they
remember
for
40
seconds,
so
I
guess
ample
time
for
BGP
routing
to
reconverge.
So
it's
a
combination
of
the
BGP
routing
at
the
network
layer
and
also
Forney
at
the
link
layer
that
keeps
from
losing
packets.
When
we
have
that
BGP
route
withdrawn.
A
J
May
may
mean
a
million
or
10
million
aircraft,
but
if
we're
talking
about
all
worldwide
mobile
devices,
it's
several
orders
of
magnitude,
greater,
so
I
wonder
if
you
can
comment
on
first
of
all
is
this:
does
this
scale
till
two
really
big
numbers
like
the
the
billions
and
billions
of
mobile
devices?
And
secondly,
the
other
part
of
this
is
what
is
the?
How
dynamic
is
this?
So
if
I
have
a
lot
of
mobility
and
things
are
constantly
changing,
that's
obviously
going
to
generate
load
on
on
BGP,
which
I
don't
think.
H
Questions
so
this
this
system,
we
more
or
less
sized
for
1
billion,
Mobile's
I
to
go
larger
than
that.
You
start
pressing
on
bgp
peering
limitations.
Bt
routers
can
handle
about
a
thousand
periods
that
they
can't
handle
hundreds
of
thousands
of
periods.
So
that's
where
you
have
a
little
bit
of
limitations
in
scaling,
because
the
numbers
of
BGP
periods,
but
if
you
wanted
to
have
billions
upon
billions,
you
could
have
multiple
instances
of
this,
each
of
which
instances
Services
1
billion
prefixes.
In
terms
of
the
mobility.
H
These
mobility
anchor
points,
concentrate
mobility
and
a
regional
area.
So
if
I'm
in
northwestern
in
the
United
States
I
would
pick
a
mobility
anchor
point,
that's
located
in
Seattle
and
only
if
I
move
to
say
New
York,
but
I
move
away
from
that
anchor
point.
So
you
have
localized.
Mobility
is
contained
within
the
regions
and
only
regional
changes
get
updated
into
the
BGP.
So
that's
how
we
keep
the
mobility
churn
out
of
the
BGP
routing
system.
A
H
Not
the
aggregate
it's
true,
so
the
slash
56
is
that
the
airplanes
get
when
I'm
moving
to
say:
New
York,
I
sign
up
with
the
mobility
Anchor,
Point
and
I
say
here's
the
prefix
I
would
like
to
assert
into
the
network.
Please
use
this
prefix
in
traffic
that
you
know
comes
back
and
forth
to
the
airplane.
If
I
moved
to
Los
Angeles
I
withdraw
the
prefix
from
New
York
and
assert
the
prefix
in
Los
Angeles,
okay,
so.
A
H
Move
or
just
just
option,
yeah
thanks,
charlie
Park
in
some
future
way.
So
I
have
two
questions.
First,
one
I'm
sure
you
looked
at
our
vertical
mobile
IP
and
the
figures
from
looked
a
lot
like
what
you
already
drew
and
even
they
use
some
of
the
same
terminology
as
map
and
so
yeah
yeah.
So
there
should
be
some
overlap
in
the
solution
space
there
I
guess.
H
The
main
point
of
difference
is
that
they
never
really
considered
using
BGP
when
they
were
doing,
but
that
brings
a
lot
of,
as
you
pointed
out,
additional
protocol
that
needs
to
be
tuned
for
for
this
purpose.
So
I
was
wondering:
do
you
know?
Also
there
was
a
FM
fast,
Mobile
IP,
which
really
does
a
great
job
of
solving
the
handover
problem,
because
you
know
you
can
have
it's.
It's
really.
The
temporary
tunnels
that
go
from
previous
to
new
access
point
so
on
yeah.
Is
that
applicable
here
as
well?
H
I
had
in
the
hub
and
spokes
charlie
yeah
we're
those
mobility
anchor
points
were
out
in
the
spokes.
Each
one
of
those
mobility
anchor
points
could
be
a
it
could
be
an
coming.
Yes,
what
I?
What
I
so
so
yeah
yeah
hmmph
a
mobility
anchor
point.
What
ICAO
is
looking
at
today
is
three
alternatives
for
the
mobility
solution,
one
being
Pima,
the
other
being
less
than
the
third
bein
arrow.
So
each
one
of
these
mobility
anchor
points
a
pimp
or
domain
now
in
terms
of
the
fast
mobile
IP,
maintaining
routes.
H
While
you've
got
some
instability,
so
you
don't
have
packet
loss,
that's
one
way
of
doing
it.
We
have
another
way
of
doing
it
by
treating
the
network
as
if
it's
layered,
with
a
network
layer
at
the
BGP
level
and
link
layer,
information
at
the
tunneling
level.
So
when
you
have
a
BGP
route
withdrawing
as
long
as
you
remember
that
old
tunneling
State,
you
can
continue
to
forward
packets
during
that
period
of
instability.
So
it's
it's
achievements,
the
same
sort
of
thing
as
fast
mobile
IPs
but
I
think
it's
slightly
different
in
the
way.
B
H
Well,
there's
no
limit
on
the
number
of
IP
layers
you
can
have
anyway.
So
that's
true
that
can
be
Turtles
all
the
way
down
right.
So
now
the
second
question
I
had
so
I,
pretty
much
believe
that
this
has
a
lot
to
do
with.
Not
let
me
Airlines
crash
and
everything
and
but
there's
another
factor
coming
into
play
in
the
airspace.
I
understand
to
be
drones.
B
H
You
know
guidelines
from
is
this
consideration
of
what's
going
to
happen
when
the
drones
and
the
onion
you,
you
know
freighters
and
things
like
that
start
flying
around
alright,
so
it
means
you
could
have
it
so
that
if
you're
any
drone,
this
registered
is
doing
the
right
thing
and
if
you
detect
one,
that's
not
in
this
system,
in
other
words,
ever
route
from
global
overlay
routing
system,
that's
considered
to
be
an
intruder
or
something
like
that
and
something
bad
happens.
Definitely
Network
authentication
and
access
control.
H
Q
A
Q
Everyone
I'm
just
home
from
entity,
so
I
I,
will
report
update
texture
analysis
on
CCP
t5z
system.
So
I
already
presented
this
draft
several
times
in
past
meetings,
so
I
I'll
skip
the
blonde.
Q
To
draft
so,
as
you
know,
this
draft
other
was
adopted
working
group
document
in
January
this
year.
Thank
you
very
much
for
your
support
and
after
the
adaptation
we
received
some
feedback
from
CCPP
and
we
updated
feedback
on
the
draft
and
published
new
draft
at
Elevens
much
this
year.
So
I
will
explain
the
content
of
this
effect
in
this
presentation.
So
following
two
slides
shows
the
list
of
measure
updates
in
this
updates.
Q
For
example,
we
started
clarifying
specification
of
PF
CP
control,
plane,
protocol
related
to
user,
prey,
manipulation
and
added
terminologies
definition
of
FCP,
the
PDR
and
FA
are
in
this
update.
Also,
we
clarified
that
differences
or
definitions
of
ipv6
multihoming
between
three
CPP
and
ITF,
so
in
3gpp
a
branch
of
single
prefix,
it's
not
called
bus
coming,
so
it's
different
from
I
defeat,
ITF
division.
Also,
we
received
requests
to
add
and
Omata
supporters
and
we
deflected
it
on
this
update
and
as
the
update
point,
having
large
impact
is
specification
of
networks
rising.
Q
Specific
specifications
of
network
dressing
is
added
in
the
current
SVP
document,
ts,
23,
5,
5,
0
1.
So
in
this
specification,
3
TPP
has
some
assumption
to
use
Network
instance
to
glue
UPF
and
transport
rice,
and
we
derived
three
ways
to
glue
ups
and
network
slices
with
networking
instance
from
the
specification.
So
I
will
explain
the
details
of
specifications
and
and
our
analysis
in
the
flooring
feels
right.
Q
So
this
right
shows
overview
of
thank
you
PFC
P
and
network
instance.
This
CV
is
a
control
brain
to
hunt,
handle
you
use
a
brain
traffic
and
used
it's
used
between
a
memory
and
POS
gateway,
or
between
SMF
and
UPF
so
and
PFC
P
contains
some
information
element
and
information
element
is
used
to
detect
traffic
and
decide
its
holding
action
loops
in
user
plan
function.
So
network
network
instance
is
one
of
information
element
and
it
is
used
for
texting,
IP
domain
when
network
addresses
or
IP
addresses
of
you
is
duplicated,
like
this
figure.
Q
Q
So
from
this
specification,
we
derived
two
three
ways
to
glue
ups
and
transports
prices.
In
this
example,
we
have
assumption
to
use
MP
areas
as
transports
prices,
option
1
means
that
UPF
and
feel
outer
integrated,
as
all
in
my
boss
and
a
UPF
mega
to
relate
appropriate
MPLS
15
based
owns
it.
What
instance
indicated
from
SM
death
and
option
to
mean
that
UPF
and
customer
is
louder
interrelated,
also
option.
3
means
that
UPF
is
separated
completely
separated
from
MPLS
architecture
and
put
UPF
behind
of
customize
your
router.
A
Q
Q
I
Saturday
master,
Softbank,
yeah
I,
think
the
these
analysts
allowed
to
bring
the
much
more
detail
collecting
information
to
idea
of
community
right
so
not
only
for
sizing,
but
also
the
mobility
management
concept
appear
on
the
fabric
from
phifer
G,
don't
Network
instance.
It's
for
example,
so
I
think
many
people
not
aware
but
network
innocence.
Even
for
that,
yet
that
typically
sophistication
had
not
so
much
career
to
define.
B
I
A
Q
Q
Okay,
sorry,
lastly,
I
will
mention
status
and
next
steps
of
this
draft,
so
we
reflected
off
of
feedback
on
mailing
list
and
it
looks
appropriate
and
also
we
will
add
gratification
or
PFC
P
related
to
use
a
pre
hungry,
because
it
will
help
for
clarifying
evolution
aspects
of
user
protocol.
For
example,
data
structure
of
the
FCP
is
very
useful
for
verifying
user
base
specification
so
also.
Q
I
Subtle
much
muscles
buck,
adding
another
document.
Another
idea
of
EDS
VPP
specification
that
kind
of
the
endless
work
so
I
don't
want
to
agree
that
the
workload
but
it's
a
it's
good
to
study
the
user
per
emulated
specification
to
bring
to
the
idea
of
community
I.
Think
I
really
appreciate
to
join
other
people
then
receive
the
comment
so
I.
I
Sort
of
different
concepts
on
the
idea
of
architecture
specification,
so
I
think
it's
clarifying.
What's
the
UPF
exactly
do.
Clark
is
the
very
different,
so
I
think
the
the
qualifier,
the
the
how
UPF
is
defined
as
a
body
rubble
to
bring
to
clarify
on
this
dotnet
so
on.
So
if
a
certificate
document
is
updated
continuously,
so
chasing
that
update,
it
I
think
much
Padova
right.
So.
D
I
D
I
do
follow
like
a
lot
of
the
3gpp
work.
I
just
want
to
know.
Like
you
know,
what
is
the
end
goal
of
this
right,
like
you
know,
is
that
like
so
when
this
work
is
done
or
you
wanna
RFC
is
an
RFC.
What
you
want,
or
is
there
some
liaison
statement
you
want
us
to
send
towards
3gpp.
What's
the
end
goal
of
this.
D
D
So,
if
you
remember,
like
you
know,
like
I
was
working
on
as
she
was
working
on
it
in
release
a
time
frame
right,
we
took
the
ITF
protocols
to
3gpp
and
started
writing
TSS
right
and-
and
is
that,
like
the
way
you
want
this
to
go
forward
right
like
because
I
existence
of
an
IETF
RFC,
doesn't
it
only
helps
like
do
something
in
3gpp,
but
the
work
doesn't
get
done
by
ITF
Pharisees
in
3gpp
right
so
like
Denis,
sleep
at
all?
Well
that
happens
in
3gpp
too.
So
that's
what
I'm
really
looking
at
like.
I
A
John
John
from
Warwick
I
just
wanted
to
add
that
this
document
is
really
useful.
It
does
identify
some
gaps
between
what
3gpp
defines
in
terms
of
transport
segments
yeah
and
what
is
actually
happening
on.
You
know
multiple
transports
that
cross
across,
so
those
views
on
transport
and
slicing
it
could
lead
to.
You
know
we
need
to
do
some
work
here.
I
think.
O
I
G
G
Then
we
did
some
minor
victorious
and
we
also
updated
some
references
and
do
also
some
final
proofreading,
an
under
vision
of
all
the
all
data
for
together.
So
basically
we
are,
we
are
done
and
we
are
waiting
for
the
safer
shares
review
for
for
the
next
step.
By
the
way
we
received
some
some
minor
things
that
we
need
to
update
from
dapeng
I,
guess
that
we
wait
until
the
the
final
review
and
do
all
the
all
the
updates
in
one
single
shot
right
right.
A
A
And
the
last
one
on
the
agenda
is
PV
six
extensions
for
D
again
by
Carlos,
okay,.
G
So
again,
very
short,
go
to
the
next
one,
so
we
submitted
version
zero
for
address.
You
know.
So
the
comment
have
we
received
early
in
the
working
of
last
call.
So
then
the
main
changes
that
we
did,
which
our
kind
of
minor
was
when
we
change
to
experimental.
Based
on
the
discussion
that
we
had.
We
with
you
guys
on
the
intended
status
of
the
document,
then
we
did
some
minor
editorials
based
on
the
feedback
that
we
received.
G
We
were
also
checking
the
status
of
the
different
flags
in
the
video,
a
ba
heavens
to
basically
adapt
with
its
status
and
again
with
it,
also
the
definer
proofreading
and
revision,
and
we
are
in
the
same
States
with
the
document
like
with
the
other
one.
We
are
waiting
for
the
church
review
to
update
whatever
is
missing.
Okay,.
A
Yeah
I
think,
like
I,
said:
I
think
the
cherry
is
pending.
That's
one
thing:
the
impact
of
IPR
declaration
that
will
be
analyzed
and
any
any
questions,
comments.
I,
guess
that's!
That
was
our
last
presentation.
No
other
comments.
We
can
close
the
meeting
here.
Mia
say:
Mia
has
the
notes
right,
I
think?
Yes,
a
she
you,
you
have
done
reading
much.
Okay!
Thank
you!
So
much
thanks,
everyone,
yeah
and
the
bow
sheets.
Please.