►
From YouTube: IETF101-DMM-20180320-1550
Description
DMM meeting session at IETF101
2018/03/20 1550
https://datatracker.ietf.org/meeting/101/proceedings/
A
All
right,
well
comforts
I'll,
come
to
this
with
the
DMM
working
group,
so
this
is
a
shrink
underbelly.
So
my
coach
here
Dapeng
is,
is
on
holiday,
so
today
it
will
be
just
me
and
Suresh
is
also
here.
So
all
right,
so
we
have
a
full
agenda
today.
Just
before
that
some
official
business,
please
read
the
note
1.
So
this
meeting
is
governed
under
some
setting
ITF
rules
and
regulations,
and
these
are
documented
in
various
RFC's.
A
Please
read
those
specifications
documents
with
respect
to
IPR
policies
with
respect
to
code
and
conduct
and
other
aspects,
and
with
that
so
I
think
blue
sheets
I
think
you
know
please
sign
the
blue
sheets
and
we
need
1
minute.
Take
her
place
and
I.
Think
like
you
are
going
to
take
some
minutes
all
right,
so
we
need
one
more
note
taker
anybody
else.
Please
volunteer
when
it
taker
yeah
yeah
colors,
when
it's
I
will
give
you
a
piece
of
paper
and
a
pen.
A
All
right
we
can
get
started.
So
this
is
the
agenda,
so
we
have
a
full
agenda
today.
I'll
appreciate
your
support
with
respect
to
you
know,
moving
things
quickly
right,
so
we
have
total
I
getting
12
presentations.
So
this
is
the
agenda
and
a
quick
review
of
the
working
with
document
status.
You
know
I
think
we
have
couple
of
you
know
working
with
documents.
One
is
on
demand.
Mobility,
I
think
there
was
a
you
know.
Last
recent
update
posted
only
this
morning,
I
believe.
A
But
as
far
as
we
know,
all
the
authors
have
resolved
all
the
issues
and
currently
the
document
is
in
working
with
last
call.
I
think
this
is
the
third
working
last
call
previously,
because
you
know
few
issues
are
identified
and
we
had
to
send
it
back
to
the
authors
and
author
worked
hard
and
to
resolve
all
the
issues.
So
it's
on
our
last
call.
Please
provide
your
feedback,
there's
an
opportunity
to
fix
any
you
know
make
any
final.
You
know
changes
and
I
think
there's
also
chair
review
pending.
That
is
the
first
document.
A
The
second
document
is
the
DMM
deployment
models.
I
think
it's
it's
in
a
good
shape,
but
I
think
this
is
a
dependency
I
think
the
problem,
why
the
reason
why
we
are
not
moving
this
forward
is,
you
know
there
is
a
dependency
on
the
HPC
document.
We
don't
want
to
move
this
faster
and
late
or
equity
and
later
you
know
you
know,
if
there's
some
changes
in
either
HPC
document
or
other
documents,
we
you
know,
we
don't
want
to
go
back
and
fix
this.
A
So
that's
why
we
are
just
moving
this
one,
a
little
slower
track,
but
overall
this
is
a
good
shot,
good
shape
and
next
is
see.
Pdp
I
think
you
know
this
is
one
document
which
is
like
in
a
tremendous
amount
of
efforts
are
going
into
this.
Like
you
know,
I
think
we
are,
you
know
very
past
or
you
know
original
timeline
for
this,
but
you
know
given
the
amount
of
efforts-
and
you
know,
I
think
it's
it's.
A
A
lot
of
work
is
going
in
I,
think
that
is
one
thing,
but
still
I
think
you
know
we
are
almost
there
and
I
think
I
believe
you
know
Charlie
at
Osan.
We
made,
like
you,
know,
tremendous
Danny.
They
made
some,
you
know,
put
in
a
lot
of
efforts
and
it's
coming
out
good
and
I
believe
in
a
strength
also
also
an
implementational
but
I
believe
I
will
talk
about
this,
but
overall,
this
in
a
great
shape
but
I
think
I.
A
Think
still
one
problem
is
like
you
know,
then
there
hasn't
been
much
feedback
from
the
working
group.
I.
Think
after
charlie
joined
in,
like
you
know,
think
he's
he
made.
You
know
some
very
good
comments
and
a
lot
of
things
is
a
lot
of
changes
are
going
in,
but
still
you
need
broader
in
a
participation
in
otherwise
we
won't
be
moving
this
to
the
IAC
at
any
points.
One
so
appreciate
some
support
on
this
and
then
the
deal
distributed.
Mobility,
anchoring
I
think
Anthony.
This
is
one
thing
you
know
it
was
already
passed.
A
A
We
are
going
to
follow
it
this
and
then
at
that
point,
authors
have
engaged
in
a
cordless
and
the
document
is
now
around
18
pages
and
it
sounds
you
know
it's
in
a
great
shape,
but
this
is
almost
like
a
reset.
Even
the
document
is
like
God
a
you
know.
You
know
good
good,
well
Quentin,
but
you
know,
given
that
we
pretty
much
rewrote
the
document
we
need
to.
You
know
you
want
to
make
sure
this
was.
This
is
reviewed
by
the
entire
group
so
overall,
but
focus
on
this
I
think
you
know.
A
Good
thing
is
you
know
this
progress?
You
know
document
from
85
pages
to
18
pages.
Is
you
know,
and
and
with
the
readability
I
think
they
made
tremendous
amount
of
words
Thank
You
Anthony?
Thank
you
Cutlass
and
the
other
one
is
a
services
for
mobile
user
plea.
I.
Think
there
was
one
revision
from
savasana
of--this.
A
There's
a
presentation
slot
I'll,
let
doesn't
there
are
a
few
discussions,
but
I
think
elect
are
sort
of
zon
to
you
know,
provide
feedback
on
what
the
changes
are,
but
this
is
really
deal
in
it
early
stages.
So
we
need
to
have
more
discussions,
but
you
know-
and
you
know
I
think
before
we
move,
you
know
last
call
or
anything
now
with
respect
to
maintenance
document,
so
I
think
there
is
one
I
usually
discuss
on
forty
to
eighty
three
but
looks
like
it
now
resolved.
I.
Think
Charlie,
congratulations,
Suresh!.
C
Situation
so
yeah
it
got
resolved
so
Charlie
and
I
had
a
call
last
week
and
we
found
way
forward
like
that,
would
resolve
the
disgust.
So
it's
like
it
was
there
for
I.
Think
like
390
days,
yeah
and
I,
discuss
and
I
think
it's
it's
done
now.
So
it's
I
just
Cindy
up
rollin
or
as
yesterday,
but
it's
gone
so
it
should
be
in
a
queue
and
it's
all
to
go.
So.
A
You
Suresh
I
think
Thank
You
charlie,
while
resolving
this
I
think
you
know
this
is
now
the
publication
queue
excellent
and
the
other
one
mag
multihoming
this
published
as
RFC
1878.
This
is
very
good,
and
now
there
were
two
moving
on
to
the
license
statements.
They
were
two
LS
statements,
a
very
safe
from
through
GBP
from
City
for
I.
Believe
one
was
on
the
are
a
you
know,
semantics
with
respect
to
coloring
I
think
you
know
I
work.
A
C
Substitutions
I
haven't
received
this
liaison
statement,
so
the
first
one
I
have
yeah
and
the
second
one
I
haven't
so
likes
desolation
statement,
newer
liaison
statement
for
the
first
case
for
the
year.
So
is
we
discuss
it
on
the
3gpp
IDF
call
last
week
and
the
second
one
I
have
no
idea.
So
it's
not
like,
so
we
need
to
figure
it
out.
Yeah
yeah.
D
C
A
Maybe
it's
not
okay,
it
has
to
come
in
to
the
ietf
system.
That's
why
I'm
not
you
know,
there's
no,
it's
not
created
yet,
but
from
a
statistician,
it's
already
created.
Maybe
you
know
I
should
have,
but
but
yeah,
but
we
know
it's
coming.
Okay,
all
right.
So
moving
on
I
think
that
this
is
the
object
is
to
working
with
documents,
and
now
we
go
on
I
think
you
know,
I
need
some
help
with
the
timer's
please
for
each.
You
know
this
thing
you
fix.
A
A
A
F
Clients,
that's
embedded
with
FBC,
were
then
moving
with
the
agents
and
basically
it's
controlled
or
the
agents
essentially
provisioning.
The
data
plane
odor
DPN.
So
this
allows
us
this
or
and
we've
gotten
the
model.
Several
things
like
multi-tenancy
we've
got
policy
provisioning.
We've
got
quite
a
few
features
in
here.
It's
a
rather
large
document
about
50%
of
it.
It's
sitting
in
there
again
a.
F
F
To
go
through
everything
they're,
just
quite
a
bit:
I'll
only
go
through
the
big
highlights
next
line,
so
the
biggest
change
of
the
system
is
templates
so
before
we
had
a
very
structured
model
and
what
we've
moved
to
is
a
templating
system
with
a
templating
language.
So
what
we're
trying
to
do
is
essentially
simplify
the
development
of
any
further
extensions
of
FBC
in
previous
versions.
F
F
We
kind
of
arrived
at
this
template
notation,
so
every
template
that
you
see
in
there
is
going
to
have
a
set
of
attributes,
and
we
have
a
couple
of
indicators
are
important
to
us.
So
the
extensibility
boolean
basically
lets
us
know
whether
or
not
this
template
can
be
further
extended
as
we
go
through
the
policy
process
and
that's
really
important
from
a
policy
design.
Point
of
view
right.
F
We're
really
talking
about
the
ability
to
set
up
when
we
leave
the
extensibility
open
or
true
the
ability
for
people
to
go
ahead
and
really
define
new
templates
or
restrict
templates,
but
by
setting
it
false
right,
we're
really
kind
of
locking
down,
even
at
the
beginning
or
some
later
points
as
we
go
through
the
template
design.
What
people
can
do
from
an
operational
perspective,
the
nav
state
and
there's
several
states
to
this.
We'll
talk
about
that
here
in
a
little
bit
allows
us
to
kind
of
progress.
F
What
we're
trying
to
get
to
when
we're
basically
adding
to
a
template
as
we
go
through
the
various
lifecycle
stages
and
policy,
is
we're
trying
to
get
to
the
point
that
a
template
is
configured
and
what
we
mean
by
that
is:
there's
sufficient
information
in
the
template
so
that
we
can
actually
take
action
on
a
DPN
and
what
we
then
want
to
see
on
the
D
again.
Is
it
moved
from
the
configured
state
to
an
active
state?
So
then
we
know
it's
actually
in
play.
F
It's
it's
running
and
we
know
it's
being
applied
to
the
mobility
context
in
question.
We
also
have
versioning
just
so
that
we
can
track.
You
know
it's
just
standard
meeting
stuff
now
as
a
part
of
this
and
the
specification
of
attributes,
we
had
to
actually
bring
up
a
definition
notation,
and
so
you
see
all
the
various
items
here.
The
one
thing
I'd
point
out,
though,
is
the
last
three
items
really
are
about
non
mandatory
attributes
and
I.
F
Think
what's
important
here
is
that
extensibility
attribute
when
it
comes
into
play
right,
especially
when
it
says
the
false
with
non
non
mandatory
attributes
right.
You
can
see
a
lot
of
different
use
cases
for
that
in
terms
of
how
we
can
handle
it
so
next
slide,
so
policy
templates,
so
we
did
have
policies
before,
but
now
really
policies
are
just
effectively
templates
follow
a
very
specific
pattern
for
us.
So
when
we
talk
about
a
policy,
we
reference
it
by
its
key.
F
So
if
you
see
something
like
policy
key
or
DPN
policy,
key
or
DPN
key
any
sort
of
the
key
notation,
it
really
is
a
reference,
and
when
we're
making
reference
or
a
particular
entity
sort
of
configuration
on
the
template,
we
really
have
to
have
the
initial
reference
that
we're
talking
about
and
then
there's
attribute
expressions
that
we
just
described
so
a
good
example
of
this
is
in
the
DPN.
We
actually
have
the
ability
to
push
in
DPN
policies
in
the
previous
version.
These
were
referred
to
as
install
policies.
F
You've
heard
them
also
referred
to
as
configured
policies,
and
so
what
you'll
see
here
in
the
DPN
is
essentially
those
policies
are
now
underneath
the
DPN
they're
installed
there
and
you'll
see
then
the
reference
to
the
policy
key
in
question
and
then
any
sort
of
attribute
specification.
There's
several
entity
config
blocks.
This
is
a
overall
design
pattern
and
you
see
the
various
locations
there,
so
this
is
kind
of
a
rinse
repeat
architecture
and
allows
us
to
really
kind
of
simplify
the
overall
design
of
the
document
next
slide.
F
This
is
the
policy
model
I'll
point
out
to
you
that
everything's
still
there,
you
still
have
your
actions
as
you
solve.
Your
descriptors
rules
were
actually
a
substructure
under
policy,
they're
now
pulled
up
and
they
have
the
same
depth
because
of
the
template
architecture.
So
this
allows
us
to
kind
of
mix
and
match
rules
inside
the
policy.
Space
and
policies
still
have
the
same
idea
of
essentially
a
search
space
with
precedents
for
rules.
F
F
The
overall
lifecycle
is
pretty
straightforward,
so
your
first
step
is
really
up
at
the
top.
You
design
the
what
you
need
underneath
the
policy
information
model
it
can
be
as
simple
or
as
prescriptive
as
you
like.
A
good
example
of
very
prescriptive
is
things
like
a
policy
that
literally
just
says
block
everything
incoming
port
22,
something
very
generic
would
be
a
drop
action
based
off
of
an
RFC
five
triple7
classifier.
So
here's
where
the
entity
status
comes
into
play.
F
If
there's
not
a
lot
of
configuration,
we
refer
to
the
statuses
initial
partially
configured
means
more
attributes
have
been
added.
Configured
means
we
actually
have
gone
to
the
point.
We
can
actually
activate
this
somewhere
in
a
context,
but
typically
we
won't
expect
anything
beyond
probably
partially
configured
at
this
point
in
the
model.
The
next
step,
then,
is
really
in
the
middle
and
that's
to
really
apply
it
to
a
DP
end.
So
this
is
where
we
see
something:
that's
really
defined
at
the
tenant
level,
now
being
installed
into
a
data
play
node.
F
That
doesn't
actually
mean
it's
being
used
yet
right,
so
we're
just
trying
to
make
sure
it's
sitting
there.
It's
ready
and
that's
available
and
that's
at
DPM
policy
configuration
which
we
referred
to
as
install
policy
or
configure
policy
and
pass
provisions,
and
then,
finally,
you
actually
see
the
application
in
this
or
the
assignment
inside
the
mobility
context,
and
there
we'd
like
to
see
the
actual
configuration
progressed
closer
to
configured
and
turned
on
the
model
and
hopefully
right
if
it's
actually
being
applied
right,
we'll
see
that
entity
status
changed
to
active.
F
So
when
you
look
in,
you
should
be
able
to
see
which
policies
were
not
only
installed,
they
were
potentially
maybe
pre,
provisioned
or
pre-configured
on
the
mobility
contact
in
question
and
which
ones
are
actually
active.
So
if
you're
familiar
with
things
like
LTE
and
volte,
this
is
a
lot
of
what
we
do
right.
F
Is
we
set
a
lot
of
stuff
too,
like
the
configured
state,
its
rapid,
ready
to
go
and
then
would
go
active
once
we
actually
decide
what
you
know,
which
one
of
the
options
we're
applying
for
the
negotiated
call
next
slide,
topology
and
service
a
lot
of
renaming.
So
this
is
the
old
model.
We
do
have
a
proposal
out
this
week
to
get
rid
of
service
endpoints,
but
I'll
kind
of
go
through
this.
Just
briefly,
we
essentially
have
VPN
groups
and
peer
groups
renamed
and
using
the
service
group
nomenclature.
F
Service
endpoint
is
currently
a
kind
of
an
older
mechanism
from
the
previous
design.
It's
currently
under
review
with
the
co-authors
to
be
eliminated,
and
then
we
still
have
domain
and
we've
also
had
the
DPN
structure,
which
we
previously
talked
about
so
and
just
a
reminder
right
installed,
policies
gone
so
it's
now
sitting
under
DPN,
so
next
one.
So
this
was
the
relationships,
and
this
is
also
why
we
were
a
little.
You
know
this
is
really
the
last
thing
we
went
through
for
improvement.
F
Mobility
context,
although
it
goes
into
the
new
template
model,
a
lot
of
the
stuff
that
we
know
is
still
there.
So
we
do
have
the
extensibility
attribute
here
it
is
a
template.
So
the
interesting
thing
is,
you
can
lock
context
now
and
make
it
immutable,
so
that
has
some
some
very
interesting
applications
in
the
network
as
well.
We've
still
got
the
basic
parent-child
types
of
designs,
the
the
mobile
node
domain
and
the
DPN
assignments
for
version
nine.
Now
following
this
one
nomenclature
name,
is
the
service
data
flow.
F
We
now
allow
that
essentially
embedded
rule
and
all
of
that
specific
terminology
is
gone.
We
do
have
one
reserved
key
in
the
structure
so
far
for
unassigned
dpns.
There
are
some
architectures
that
do
try
to
track
rules
or
policies
that
have
not
been
assigned
yet
to
a
specific
DPN.
So
we
just
set
aside
that
name
rather
than
having
the
embedded
rule
as
a
separate
structure.
Next
slide,
our
PCs
did
go
under
an
overhaul,
so
we
migrated
to
a
yang
patch
style,
but
we're
not
quite
compliant
right.
Our
PC
is
a
really.
F
We
have
to
remember
two
things.
One
FDC
is
an
information
model,
so
we're
not
here
to
talk
a
lot
about
yang
or
other
structures.
The
other
thing
was,
if
you
recall
right,
we
have
the
ability,
upon
the
first
application
of
an
RPC
in
the
old
style,
conf
or
comp
bundles,
to
immediately
reply
back
with
what
the
client
needs
to
continue
call
processing
and
then
do
a
result
later
and
type
in
in
a
subsequent
response,
so
that
we
can
basically
sync
up
the
control
plane
with
anything.
F
The
data
plane
is
doing
yang
patch
didn't
really
facilitate
that
sort
of
concept.
So
what
we
did
was
we
basically
went
ahead
and
migrated
to
that
style.
So
your
configure
message
does
look
very
much
like
a
yang
patch.
You
now
could
do
multiple
edits,
so
there
was
no
need
for
the
kondal
and
we
went
ahead
and
migrated
all
errors
to
a
common
error
format
once
again
kind
of
following
yang
patch
and
some
of
the
guidance
there
next
time.
The
output
is
where
we
had
to
diverge
a
little
bit.
F
So
what
we
have
in
the
response,
your
typical
yang
patch,
will
just
basically
respond
back
after
that:
okay
or
an
error.
What
we
had
to
do
was
basically
have
the
concept
of
subsequent
editing,
which
is
the
idea
that,
if
you
send
and
edit
down
to
the
agent
the
agent
may
have
to
make
subsequent
edits
on
its
own
in
order
to
make
something
work.
The
easiest
example
of
this
is:
if
the
agent
is
allocating
the
tunnel
endpoint
IDs,
then
clearly
the
agents
going
to
need
to
make
an
edit.
F
So
now
you
can
actually
almost
like
a
log
figure
out
what
the
agents
doing
on
a
step-by-step
basis,
and
it
makes
it
much
easier
if
the
client
cares
to
kind
of
sort
out
which
edits
are
important
to
it
or
not,
so
that
speeds
up
kind
of
that
return,
processing
much
faster
in
the
response.
The
other
thing
to
know
is,
unlike
the
the
previous
stuff
we
do
have
more
noun
or
excuse
me
more
verbs
under
update
our
old
model
and
tell
me
once
a
Yank
batch
was
just
basically
create
update,
deletes
under
yang
patch
type
designs.
F
We
now
have
things
like
merge.
We
have
our
traditional
update.
We
have
move
which
I
think,
as
we
add
more
examples,
I'm
looking
forward
to
things
like
just
moving
the
DPN
entire
structure,
so
we
could
do
things
like
anchor
mobility
as
a
very
concise
command
in
some
of
the
subsequent
stuff
and
that
kind
of
leads
into
the
next
slide.
F
We
definitely
have
to
add
an
example
on
topology
selection,
we'll
be
adding
more
examples
as
well
in
terms
of
the
other
features,
so
we
can
highlight
key
features
and
we've
got
lots
of
little
edits,
but
quite
frankly,
guys
at
this
point
we
just
need
feedback
at
the
end
of
the
day
here.
So
we're
running
out
of
excuses
and
everything
else
to
kind
of
go
much
further
with
this,
so
I'll
leave
it
at
that
open
up
for
questions
so.
F
F
Almost
two
years
now
of
having
running
code
we've
been
demoing.
If
you
see
the
MCOR
project
depending
on
when
it's
launched,
so
the
the
overarching
projects
called
c-3po,
the
the
controller
underneath,
maybe
all
knows,
if
it
is,
it
was
developed
either
by
that's
one
implementation.
There's
another
onus,
FBC
implementation,
also
out
there
that
has
been
developed
by
actual
onf
and
then
you've
got
the
OTL
implementation
developed
like
spreads
and
now
GS
labs.
F
F
F
C
F
F
C
So
so,
basically,
the
idea
is
like
how
so
in
very
short
thing.
I'd,
like
you
know,
how
is
configuration
and
operational
state
maintained
in
the
document,
so
I,
the
latest
version
and
like
there
is
like
stuff
to
fix
and
and
I,
don't
think
like
whoever
benoit
science
will
fix
it
like
it
needs
to
be
from
for
Mars.
The
question.
F
C
I'm
just
trying
to
avoid
this
additional
ground
trip
right,
so
if
he
can
actually
look
at
anything
for
you,
no,
it's
publish
it's
done
so
there's
no
more
like
fiddling
around
with
it.
So
take
a
look
at
it
see
if
you
can
take
like
an
initial
cut
like
you
know
what
you
need
to
change.
So
basically,
the
idea
is
like
you,
don't
keep
duplicate
stuff
in
configuration,
operational
state
trees.
A
I
Well,
Anthony
couldn't
make
it
for
today's
minutes,
one
percent.
We
know
that
behalf
of
authors,
please
next
night,
so
well
I.
This
presentation
would
be
brief
because,
as
we
mentioned,
we
need
a
lot
of
changes
and
it's
like
a
very
different
ball,
very
different.
It's
a
different
document
in
the
sense
that
drastic
changes
were
made.
So
we
basically
this
meeting
what
I
want
to
request
this
feedback
from
those
that
have
already
read
the
document
and
asked
for
additional
reviews.
So
I
will
briefly
review
the
status
overview
of
the
document
and
then
the
next
step.
I
So
status
as
we
mention
version,
seven
had
like
46
radials
and
we
drastically
shortened
the
document
because
we
need
that
was
the
one
of
the
main
messages
from
the
working
group.
That
document
is
too
long
is
too
complex.
We
need
to
make
it
shorter
and
make
it
more
understandable,
so
now
version
8
has
to
be
team.
Of
course,
due
to
or
as
part
of
the
process
of
because
of
the
process,
there
are
still
some
inconsistencies
that
we
need
to
to
fix.
I
So
it's
not
meant
by
any
means
this
version
8
to
be
ready
yet,
but
we
wanted
to
have
a
version
available
for
the
working
look
to.
Basically,
let
us
know
if
this
is
on
the
track
in
the
right
track
or
not
the
main
goal.
When
we
did
this
shortening
was
to
reduce
complexity,
so
we
have
simplified.
The
terminology,
which
was
one
of
the
I,
may
not
main
potential
corners
of
the
document.
A
terminology
was
a
bit
complex
and
also
the
drawings
have
been
simplified
and
many
of
them
has
been
removed.
I
We
have
mostly
focused
on
the
data
plane
or
the
control
plane
of
courses.
Is
there
because
the
important
piece
we
have
completely
removed
parts
about
our
article
networks
and
another
mobility?
We
believe
that
those
things
could
be
addressed
and
in
an
additional
document,
separate
documents
if
the
working
group
believes
is,
is
deception,
but
that
was
not
the
main
goal
of
the
original
document.
The
scope
of
the
document
is
basically
limited
to
three
things.
One
is
the
case
of
nomadic
user,
so,
okay,
there
is
no
PDA.
A
none
corner
is
seen
of
addresses.
I
Then
there
is
the
case
of
keeping
the
anchor
on
moving
the
traffic
to
the
actual
location
of
the
of
the
device
and
the
third
case,
which
is
basically
the
uncover
location,
should
actually
be
in
the
anchor.
So
those
three
cases
are
the
ones
that
we
want
to
document
and
in
the
document,
and
we
also
focus
or
highlight
that
the
document
is
informational,
so
we
also
try
to
modify
the
language
and
the
focus
to
keep
ease
as
information.
So
there
is
no
solution
normative
takes
on
this
account.
I
So
this
is
the
overview
of
the
current
structure
of
the
document.
So
it's
been
simplified
and
I
think
we
will
probably
we
will
be
merging
section
of
foreign
section
of
five
US
based
on
some
people
that
we
already
got
from
from
work.
There
are
some
duplication
there.
So,
even
though
there
are
some
additional
information
that
we
aim
to
up
also
based
on
that
feedback,
we
will
be
merging
the
we
don't
expect
the
document
to
get
longer
than
its
is
it
as
of
today.
I
So
next
steps-
and
this
is
the
main
message-
I
wanna-
convey
or
I
wanna
ask
you-
is
we
have
done
a
lot
of
changes
on
version
8,
so
we
first
need
to
get
some
feedback
like
okay.
Is
this
on
the
right
track?
Knowing
that
is
not
meant
to
be
finished
is
not
meant
to
be
the
final
document,
not
at
this
point,
and
if
so,
we
would
like
to
have
additional
reviews
mark
or
very
provide
some
of
my
comments
that
we
will
address.
I
But
of
course
we
will
would
like
to
get
more
more
comments
from
the
working
group.
Here
we
go
those
reviews.
The
commitment
from
the
authors
is
to
really
push
this
and
to
get
a
document
ready
before
mantra,
because
this
has
been
sitting
there
for
a
long
time
already.
So
this
is,
you
have
any
any
feedback
on
on
the
current
status
and
if
in
any
way,
if
we
can
get
some
volunteers,
that
would
be
very
appreciative.
I
A
I
H
I
I
J
To
answer
a
couple
of
questions,
so
this
question
is
on
the
right
track.
Yes,
I
think
it
is
as
I
provided
feedback
to
you
already
regarding
three
signature
question
in
terms
of
readability:
they're
spaced
from
provement
I
think
the
very
that
the
biggest
improvement
is
really
the
cut
off
document
which
may
motivate
people
to
start
reading.
H
J
A
C
I,
don't
mind
if
you
do
this
thing
on
github
like
put
the
document
on
github
and
like
he
does
like
PRS
for,
like
you
know
things
you
want
to
change
and
then
summarize
say:
the
group
like
I
just
want
the
group
to
see
like
the
final
changes
like
that
I
get
that
are
getting
done
and
why
but
used
any
intermediate
to
you
want
I
seem
like
a
lot
of
groups
using
github
for
doing
this,
and
that's
perfectly
fine
I
support
doing
that.
If
that's
what
you
want
to
do,
okay,
okay,
yeah.
A
So
all,
but
this
is
no
right
track,
but
I
think
you
know
this
is
a
reset.
Let's
make
sure
that
you
know,
or
you
know
we
don't
you
know
who
this
to
quickly
without
you
know,
making
sure
no
super
obvious
right,
I
think
but
again,
I
think
thank
you
and
working.
You
please
and
provide
feedback
all
right,
I
think
next
is
use
a
plane.
Protocol,
steady
chatter,
son.
D
D
So
the
pattern
here
is
so
far:
there's
no
intensive
study
or
use
doctrine
protocol,
while
the
in
5z
there's
a
drastic
change
in
the
contraband
side
moving
to
the
SBA.
So
during
that
we
can
see
the
yellow
to
the
ipv6
adoption
in
the
production
network,
so
that
may
introduce
a
lot
of
space
to
improve
the
user
playing
operation
in
the
ipv6
network.
So
that's
what
the
IAB
recommend
the
other
sto
to
revisit
the.
D
Specification
that
can
be
done,
but
more
efficient
in
ipv6
on
the
network,
so
city
for
decided
to
visit
I
used
up
the
importable
to
investigate
potential
limit
of
the
existing
user
plan
protocol
and
also
the
seek
the
potential
benefit
of
alternative
is
a
plan
protocol,
including
ITF
solution.
So
that's
the
the
brief
background.
D
So
how
we
proceed
the
study
is
here.
The
objective
is
the
phase
approach
so
phase
one
is
just
be
a
DH,
because
I
see
you
like
ITF
two
to
collect
the
candidate
user
plane
protocol
and
also
give
it
some
information
regarding
existing
user
plan
photo
of
rapid
GTP
you,
which
is
defined
by
TS
twenty
nine
point.
Two,
eight
one
latest
button
at
the
bottom
fifteen
point
one
one
zero
and
also
they
did
soon
following
the.
D
D
That
study
to
make
the
decision
so
how
to
approach
that
phase
two
is
the
one
with
the
approach
is
investigation
so
that
investigate
solution
alternative
to
gtp
you
including
active
solution,
so
it
may
crude
as
a
working
group
in
CTP
like
round
three
City,
three
four:
is
it
easy
to?
If
need
it,
and
also
a
little
bit
weight,
so
the
after
that
comprise
them
to
be.
D
D
A
K
D
So
how
do
you
measure
that's
so
d
TPU
it
tightly
coupled
to
the
GTP
control
frame?
So
when
we
see
the
opportunity
to
swap
the
user
friend
protocol,
we
may
also
see
the
impact
to
the
control
plane
protocol
like
HTTP
see,
but
nowadays
we
also
already
have
the
cups
to
make
sure
the
user
plain
contra
brain
the
caffeine,
so
that's
also
be
impacted
by
a
user
frame
for
call
swapping.
So
so
we
also
ask
to
the
investment
investigate
the
study,
how
much
the
impact,
to
the
add
those
existing
control
protocol.
So
does
that
mean.
A
M
316
has
a
study
item
in
essay
2,
which
is
looking
at
UPF
to
be
connected
to
the
services
based
architecture,
which
means
a
UPF
could
have
the
services
based
interface
in
the
architecture.
So
you
may
not
necessarily
be
restricted
by
GT
PC,
it's
actually
an
for
in
race,
15,
next-generation
core
network,
so
it's
and
fro,
which
is
PFC
P
based,
which
is
specified
by
a
3gpp
and
in
release
16.
There
is
an
opportunity
to
look
at
em
api
from
the
user
to
the
services
based
on
maybe.
D
Yes,
but
the
other
thing
happen
on
the
assets
is
high,
but
this
is
the
city
for
so
the
we
are
awaiting
from
a
CSI.
What's
the
what
could
be
required
for
this
yeah.
M
M
N
M
That
just
got
approved
two
weeks
ago,
so
the
strategy
item
we
can
either
we
can
copy
them
or
because
we
have
participants
in
both
sides.
We
can
tell
our
delegates
to
they
can
LS
sent
or
because
we
are
dredging
the
two
sides
we
can
send
when
we
are
responding
back
to
the
CT
for
LS.
We
can
inform
I,
say
two
also
because
the
CT
for
study
says
if
there
are
impacts
to
the
rest
of
them,
like
control,
plane,
impacts
or
they
could
go
into
the
ran
three
and
the
sa-2
as
well.
M
C
C
What
group,
like
you
know,
ran
3
CT
3
C,
before
like
the
name
of
the
TS
or
TR
whatever
it
is,
and
then
the
idea
of
trust
that
corresponds
to
it
and
I
think
we
need
to
start
tracking
this
more
closely
because
otherwise
it's
going
to
get
out
of
hand,
especially
if,
like
s
a2
and
CT
for
don't
agree
what
this
is,
then
we
need
to
have
some
kind
of
tracking,
so
I'll
take
an
idle
chatter.
Some
and
some
time
maybe
see
you.
C
We
sit
down,
write
something
up
and
send
it
to
Gardenhire
and
Gonzalo,
so
they
can
track
this.
We
are
a
call
last
week
for
the
coordination,
3gpp
IETF,
the
we
have
a
coordination
meeting
and
we
had
no
idea
about
any
of
this.
So
we
need
to
make
sure
that
we
kind
of
keep
track
of
all
these
things,
because
it's
not
just
this
working
group
on
sure
it's
going
to
spill
over
into
the
other
thing
that
came
up
and
as
an
LS
was
like
six
month
later
at,
for
example,
right.
C
N
M
N
A
A
I
think
a
Liana
clarify
that
this
is
not
on
a
spy
or
you
know.
Is
it
I'm,
forgetting
the
4G
interfaces?
Fire
is
sake,
but
mainly-
and
you
know
in
my
interface
yeah
so
yeah.
O
A
few
comments
actually
here.
First
of
all,
the
study
group
here
what
you're
trying
to
do
is
just
to
analyze
different
protocols.
The
selection
has
three
actually
indicated:
it
belongs
to
the
3gpp
guys
and
the
secret
court,
so
those
are
the
guys
who
actually
will
destroy.
Ultimately,
what
happens.
Our
objective
is
actually
is
to
provide
a
thorough
analysis
of
different
protocols
and
contending
protocols
that
provides
the
provider
based
or
3gpp
to
go
and
do
their
selection.
Okay,
first
one
regarding
control
plane.
O
The
idea
is
not
to
have
the
control
thing
untouched.
Through
the
analysis
we
will
find
out
which
of
these
protocols
have
what
type
of
impact
on
different
interfaces
and
different
protocols
across
the
3gpp.
Once
we
actually
have
all
the
data
collected,
then
we
can
actually
compare
them
and
see
which
one
actually
provides
minimum
or
maximum
impact
or
like
changes
into
3d
PP
interfaces.
In
fact,
yeah.
L
Jose
cardenas
go
away
just
a
question
on
the
last
zone
coming
from
3gpp,
it's
mentioned:
there's
a
set
criteria
of
protocol
selection
based
only
16,
but
how
then
ITF
is
evolved
in
this
process,
so
I
assume
that
this
is
something
that
the
study
item
in
3gpp
will
do,
but
how
ITF
will
actually
contribute
to
the
criteria?
You
know
that
will
be
used
to
select
the
solution.
D
C
Okay,
so
description
just
cause,
you
gone
from
this
like
so
Charlie.
If
you
want
something
to
be
sent
to
3gpp
about
the
existence
of
this
which
I'm
pretty
sure
they're
aware
of,
but
if
you
want
officially
send
a
note,
we
can
send
a
note.
We
can
send
an
LS
to
put
together
something
and
say:
okay,
let's
do
it
and
then
I
can
ship
it
off
to
sa.
To
saying
like
hey,
we
have
this
document
like
you
can
use
for
a
TCP
separation
actually.
P
The
line
is
closed,
but
just
a
clarification
on
the
previous
comments.
I
think
this
is
fairly
crucial
and
I
believe
this.
The
statement
was:
is
the
IETF
to
provide
selection
criteria
or
is
the
IETF
to
provide
information,
and
the
3dep
applies
their
selection
criteria?
I
will
assume
the
latter,
but
I
heard
the
former,
so
I
just
wanted
to
make
them
leave
it
on
that
I.
D
A
Q
So
I'm
going
to
talk
about
Lisp
in
the
mobile
network,
we
have
kind
of
two
forms
of
mobility
in
Lisp
in
the
handset
itself,
which
we're
not
talking
about
here.
This
is
purely
putting
Lisp
in
the
carrier
network.
All
so
just
some
high-level
goals,
they're
very
high
level.
We
really
want
to
simplify
the
mobile
network.
We
have
new
latency
requirements,
some
people
say
one
millisecond
for
VRA
are
applications.
We
want
to
address
newer
and
more
demanding
applications
like
the
size
of
the
number
of
things
we
have
to
keep
track
in
the
network.
Q
If
we
use
a
polled
based
mapping
database
control
penny,
we
could
probably
scale
and
secure
a
more
centralized
way.
This
should
make
management
easier,
so
not
to
touch
so
many
notes
and
we
can
reduce
effects
through
incremental
deployability.
It's
pretty
much
been
proven
so
over
the
years
they
are
using
overlays
that
you
could
deploy
things
incrementally
and
not
have
to
change
touch.
The
coin
Network
put
things
on
the
edges
where
you
need
to.
Q
So
that's
the
sort
of
approach
you
want
to
take
here
as
well,
using
a
dynamic,
encapsulating
overlay,
there's
a
lot
of
judgment
about
it.
Gtp
is
static
or
dynamic,
or
provision
dynamic
or
whatever
it
really
doesn't
matter,
but
it's
a
way
of
putting
headers
on
packets
dynamically
when
they
come
in
and
not
having
to
set
up
tunnels.
That
may
or
may
not
look
like
virtual
point-to-point
lanes.
Q
So
we
think
we
can
do
address
management
much
simpler,
much
simpler
by
having
the
devices
that
are
sending
that
originating
and
consuming
packets
in
a
mapping
system
and
not
the
routing
system,
and
we
believe
we
can
do
fast
mobility,
handoffs
and
this
we
have
various
ways
of
doing
handoffs.
On
many,
we
have
a
couple
ways
of
doing
signal
based
handoffs
and
signal
free
based,
handoffs,
meaning
that
when
you
move
you
don't
have
to
talk
on
the
control
plane,
it's
definitely
a
make
before
break.
We
could
talk
about
that
at
some
other
time.
Q
We
really
want
to
focus
when
we
look
in
list.
When
we
talk
about
mobility,
we
say
Eid
mobility,
which
means
we
want
to
move
a
system
from
any
layer
to
link
to
anywhere.
So
we
want
mobility
across
carriers
as
well
as
carriers
and
non
carrier
networks.
So
we
want
to
be
able
to
roam
between
a
carrier,
LTE
3gpp,
based
network
to
Wi-Fi
and
throughout
the
presentation.
Anything
you
see
in
green
and
referring
to
e
IDs.
These
are
can
be
opaque,
addresses
they're,
not
in
the
routing
system.
They're
put
in
the
mapping.
Q
Q
The
GNote
bees
and
the
UPF
SAR
list
X
TRS,
with
our
locus
for
those
of
you
who
don't
know
list
a
lisp
x.
Tr
is
a
box
that
encapsulates
or
D
capsule
eights
packets.
So
when
an
encapsulated
capsule
eights
packets,
what's
in
the
outer
header,
are
the
our
load.
So
the
underlay
can
route
the
packet
to
the
D
capsule
later
the
underlay
is
the
existing
EPC
in
4G
or
the
NGC
and
5g.
To
me,
these
are
just
new
terms,
but
it's
the
same
IP
network
that
we're
sending
packets
over
it
is
the
underlay.
Q
Q
This
mapping
system
can
run
anywhere
in
the
end
NGC,
and
at
this
point
in
time
the
recommendation
is
the
carrier
would
manage
this
mapping
system.
That's
modulo
where
the
roaming
scope
is
for
the
Eid
capsulation
occurs
over
the
NGC
and
not
the
ran.
You
will
not
have
encapsulation
overhead
on
the
bandwidth
and
resource
sensitive
ran.
Capsulation
format
is
gtp
or
less
with
real-time
setup.
I'll
explain
later
why
it
could
also
be
gtp.
Q
Yes,
I'll
get
to
it
soon.
If
a
gtp
any
Lisp
vx
land,
it's
just
the
encapsulation,
get
over
it
supposed
to
be
a
joke.
Nobody
so
pros
free
capsulation
does
not
change
the
users
packet,
header,
okay,
it's
a
it's
a
standard
violation
to
change
anything
in
a
header
that
you're
not
supposed
to
the
only
thing
you're
supposed
to
change
in
an
IP
header
is
the
TTL
and
the
checks
on
ipv6
doesn't
have
a
checksum.
So
you
don't
so
you
get
to
save
cycles,
not
doing
that.
Q
We
want
to
identify
a
is
always
maintained
while
staying
private
okay.
By
doing
that,
we
keep
this
identity,
that's
the
pros
for
encapsulation
middleboxes
camp.
It
can
maintain
flow
State
due
to
no
header,
translation,
overlay
and
underlay
address.
Families
can
be
different
with
encapsulation
debugging
and
monitoring
always
tells
you
where
a
packet
is
from
from
where
I
should
say
from
whom
from
where
who,
where
that's
a
very
useful
thing
where,
when
you
do
address
translation,
you'd
lose
bits
you
destroyed
this
okay.
Q
The
cons
for
encapsulation
is
that
you
have
packet
overhead,
that
is
the
cost
of
encapsulation.
It's
very
clear.
We
use
tunnels
all
the
time,
I
believe
we're
at
the
point
in
the
internet
generation,
where
we
have
enough
bandwidth
and
resources
that
we
can.
You
can
deal
with
this
overhead
because
the
features
that
it
gives
are
pretty
powerful.
Q
Okay,
just
a
quick
look
at
the
list
of
encapsulation
format.
The
host
builds
this
packet
here.
Okay
and
the
host
dictates
the
length
based
on
its
application.
The
inner
header
contains
the
IDS
routing
system
does
not
see
any
IDs.
What's
added
is
always
a
fixed
header,
so
the
Lisp
encapsulation,
no
matter
what
features
you
run
is
always
sixteen
bytes,
an
8
byte
Lisp
adder
is
prepended
and
then
the
UDP
header
is
prepended
there.
That's
the
nation
port
in
this
UDP
header
is
a
well-known
port
and
the
source
port
changes
based
on
a
five
tuple
cache.
Q
Q
Q
R
Q
Q
If
you
look
at
the
white
arrows
we're
showing
the
communication
path
here
so
in
this
example,
we're
showing
that
the
left
Yui
wants
to
send
packets
to
the
right
Yui
and
of
course
we
want
to
do
it
on
the
shortest
path.
That's
the
whole
purpose
of
this
is
to
have
one
millisecond
low,
latency,
and
so,
where
you
see
the
green
arrows,
those
are
regular
packets
that
are
being
sent
that
are
uninhabited,
so
the
packet
that
our
packets
that
are
being
originated
by
the
host
the
genome,
be
then
we'll
do
a
lookup.
Q
Locator,
the
IP
address
that's
routable
in
the
ng
c
and
basically
encapsulates
it.
So
you
see
the
red
line
there
and
then
the
decapsulation
happens
and
sent
over
to
ram
okay.
Next,
this
is
showing
an
example
where
ue
wants
to
talk
to
a
server,
and
it's
a
server
that
doesn't
support
lists.
That's
an
important
consideration.
Q
Change
at
all
with
the
G
note
B
do
would
do
is
do
a
lookup
for
this.
It's
not
mapping
system
will
return
back
that
there.
This
is
kind
of
a
default
EW.
So
we
did
this
return
back
as
the
outlook
of
that
you
UPF
that
you
see
there
and
then
the
encapsulation
goes
to
the
UPF.
You
PFD
capsulate
sends
the
packet
natively
into
the
Internet
which
can
get
brought
it
because
it's
an
on
the
destination.
Next.
Q
Q
And
the
Manning
system
could
have
that
outlook:
storage,
storage
for
that
Eid
next,
okay,
this
is
going
to
show
you
the
mobility
case.
What
we're
showing
here
is
that
the
URI
is
talking
to
a
server
next
G
note:
B.
Okay,
so
as
we
showed
before,
encapsulation
server
go
ahead,
kept
it.
So
when
the
Uni
moves,
what
happens?
Is
this
G?
No
VCS
that
2001
:
:.
Q
This
is
what
we
added
in
the
oh
3
spec,
where
what?
If
we
need
a
branch
point
in
between-
and
we
have
to
support
these
n
3
and
n
6
interfaces
in
3gpp
unless
you
can
have
a
ID
and
rather
than
having
a
single
our
hope,
you're
gonna
have
an
explicit
look
and
then
explicit.
Locator
path
is
the
list
of
our
loans,
which
means
that
the
first
our
locus
Gino
B
and
since
he's
in
this
list,
he
knows
that
he
s
in
caps
like
next
to
this
midpoint,
and
then
we
keep
them.
H
Q
Q
S
Here
very
important
point
here:
I'm
in
ten
and
nine.
What
list
gives
you
is
your
underlay
can
be
anything
it
doesn't
mandate
that
it
has
to
be
ipv6.
We
can
use
MPLS
if
you
want.
If
you
have
existing,
that
took
is
MPLS.
You
can
do
PE
until
node
B
to
be
PU,
PF,
r,
PP.
U
PFD,
UPF
ta
can
be
done
on
these
two
interfaces.
Lisp
is
agnostic
to
that.
Okay,.
Q
G
Q
Many
do
you
know
time:
ok,
one
more
minute:
okay,
one
more
minute:
yep,
okay,
okay,
so
handoff
performance,
like
I,
said
earlier.
We
have
a
signaling
approach
using
this
pub/sub,
where,
whenever
the
our
log
set
changes,
the
map
server
knows
to
notify
the
ITRs
or
the
map
caches.
That
are
that
care
about
it.
They
send
subscription
requests
at
the
same
time
they
send
map
requests
to
the
mapping
system.
That's
the
signal
way
of
doing
it.
Arguably
that
can
work,
but
it
is
slower
because
you're
you're
trying
to
get
the
information
when
you
need
it.
Q
If
you
use
predictive
arlok's,
the
information
is
not
needed.
You
always
are
sending
packets
to
the
decap
slater's,
so
predictive
arlok's
uses
the
bandwidth
of
the
network
and
sends
packets
to
where
you
think
you're
going
to
go.
So
if
you
have
some
directionality
of
movement,
you
can
actually
predict
where
you're
going
to
encapsulate
so
the
remote
guys
encapsulate
to
multiple
places.
Next,
so
what
we
like
to
do
here
is
so
you
mentioned
that
the
IETF
will
select
protocols
and
offer
them
the
GPB.
P&Amp;G
PPP
will
decide.
I.
Q
Think
that's
a
good
idea,
but
I
want
to
go
a
step
further.
Well.
What
we
want
to
suggest
in
this
proposal
is
that
the
IETF
and
the
3gpp
work
together
to
solidify
the
standards.
We
not
only
need
Network
designers,
but
we
need
network
operators
from
each
group
of
those
schools
of
discipline
to
work
together
and
it
comes
out
as
a
combined
standard,
not
this
us
versus
they
think
we
have
to
stop
that
sort
of
thing.
You
have
to
do
the
right
thing.
Yes,.
O
Q
Okay,
I'm
somebody
has
to
just
update
on
the
list
of
standards,
so
we've
had
RFC
standard
since
experimental
RFC
since
2013
we're
making
a
selection
of
some
of
the
documents
are
turning
him
into
this
documents,
making
them
standards
track-
and
these
are
almost
close
to
working
group
last
call
Luigi
and
Joel-
are
here,
I,
don't
know
if
you
they're
just
the
beginning
thumbs
up.
So
that's
good
next
I'm
almost
done,
and
then
these
are
all
the
experimental.
Our
season.
T
Quick
come
any
quick
question:
oh
hi
come
over
so
I'm
a
little
confused
on
what
you're
proposing
for
the
data
plane,
because
you
mentioned
that
either
gtp
or
list
would
work,
but
the
way
I
read
it.
I
didn't
see
that
either
of
those
had
anything
and
particularly
it
seemed
like
IP
and
IP
would
work.
So
can
you
comment
on
it
so
is?
Is
lyft
data
plane
part
of
this,
because
it
has
something
that
the
others
don't
have
yeah
great.
Q
Question
so
I'll
answer
in
reverse
order.
We
don't
want
to
do
IP
and
IP,
because
we
need
the
UDP
entropy,
because
there's
router
vendors
that
do
lags
and
they
hash
on
the
UDP
header.
So
it
can't
be
IP
IP.
This
is
something
you
learn
through
GRE
and
when
built
list
in
2007,
all
the
ISPs
told
us
do
UDP,
plus
we
needed
to
get
through
Nats.
Okay.
Those
last
answers
the
first
question,
and
now
his
next
question
was
GDP
versus
list.
Q
Q
One
is
list
script
off
because
we
have
to
carry
key
IDs,
so
the
d
capsule
later
knows
how
to
decrypt
in
which
key
to
use-
and
we
have
this
thing
called
echo
announcing-
where
we
may
want
to
test
the
reach
ability
of
the
forward
data
plane
in
the
data
path.
So
it
only
am
type
function.
If,
if
the
3gpp
people,
who
are
the
operators,
believe
we
don't
need
that
that
any
encapsulation
could
be
used,
even
ila
could
be
used.
C
So,
generally
speaking,
I
agree
with
your
sentiment,
but
there's
no
way
it's
going
to
happen.
Gonna
have
many
because
there's
like
you
know,
IP
are
considerations
like
3gpp
folks.
Definitely
idea
works
differently.
Three
DPP's
membership
base,
like
you
know,
I,
cannot
do
a
3gpp
meeting.
Personally,
my
company
is
not
a
member
so
like
that
kind
of
like
different
contribution.
Now,
so
that's
not
gonna
happen.
So
first
thing
we're
not
going
to
be
able
to
work
together.
Is
that
what
you
mean
no.
Q
Q
C
The
way
it's
done
is
like
this
draft
has
written
today
cannot
fit
a
3gpp
stack
because,
like
if
you
look
at
a
stage
three
spec,
it's
got
way
more
details
than
what's
in
here.
Okay,
so
what's
in
this
specification,
no
more
than
this
right
because
there's
actually
specifying
which,
in
the
wire
one
of
the
options,
have
been
picked
and
stuff
like
that,
so
let's
not
go
there
like
I.
C
Think
it's
like
good
this,
but
I
would
like
somebody
from
3gpp
to
pick
this
up
and
run
with
it
in
3gpp
and
then
like
come
back
to
us,
and
we
can
have
this
collaboration
all
we
want
right,
but
somebody
it's
like.
We
cannot
sit
in
this
room
and
make
3gpp
do
some
stuff.
That's
something
I'm,
not
comfortable!
I
read
this
document.
Yes,
it's
not
yeah.
There
are
bits
on
the
word
sky
that
I
understand.
Okay,
but.
Q
U
Yeah
yeah,
so
this
I
think
is
interesting
work
and
thank
you
for
them.
I
just
came
to
say
something
about
this
sort
of
cooperation
among
multiple
STS
and
and
while
that
you
know
3gpp
ITF,
it's
an
IT
at
the
same
time.
That
sounds
a
lot
of
fun
but
actually
had
a
serious
point,
which
is
that
there's
a
good
way
of
doing
this
and
then
there's
maybe
the
wrong
way
of
doing
this
and
and
I'll
start
from
the
bone
and
I
I.
U
Don't
think
we
can
write
like
one
spec
and
that
these
organizations
also
have
different
roles
like
the
IETF.
Typically
believers,
you
know
great
practical
components,
you
define
Lisp
and
then
it's
so
good
that
it
sells
itself
in
and
it
gets
used
in
multiple
different
places
and
I.
Think
that
that's
the
model
that
we
want
to
retain
here.
Also
from
like
a
3t
PD
perspective,
you
know
they're
more
of
a
systems
organization.
Then
then
a
protocol
details
specification
organization,
so
they
want
to
probably
want
to
retain
their
ability
to
specify
the
albumen
I'm
gonna.
D
So
we
have
received
a
bunch
of
the
comment
after
the
IDF
hundred
in
Singapore,
so
we
have
many
positive
comment
from
almost
only
stand
off
list.
So
as
a
result,
we
adopted
that
document
as
a
working
group
document
so,
but
that
have
a
some
complain
about
the
mainly
on
readability.
So
we
need.
We
need
to
improve
that
takest,
especially
on
pocketable
description.
So
the
other
thing
it's
a
concern
so
much.
So
it's
always
in
such
my
IP,
this
game
caps
with
us
our
age
to
be
one
of
them
and
those
only
that's
increasing
hit.
D
The
seed
would
be
imparted
to
the
overhead
side
and
next
is
applicability
for
5zr
picture.
Many
people
ask
whether
the
5g
is
the
target
or
not,
and
also
the
other
generation
architecture
is
targeted
to
one
on.
So
we
need
to
make
clear
product
and
the
others.
The
feedback
is
how
to
cover
the
the
existing
user
frame
of
functionality
of
such
a
gtp.
You
message,
which
bring
the
end
marker
to
make
sure
the
handle
performance
much
better
and
also
the
sum
of
the
extension
hitter
defining
gtp.
D
D
D
But
it's
not
limited
to
do
that,
so
we
can
defer
any
kind
of
reactive
child
we
need
to
mention
I
mean
if
we
need
to
cover
the
the
function
with
the
Excel
basics,
so
we
change
the
terminology
from
the
basic
mode
to
traditional
mode.
It's
much
clearer
for
the
purpose
of
the
deployment,
a
deployment
scenario
and
also
we
change
the
terminology,
the
aggregate
mode
to
enhance
mode
which
much
larger
and
indeed
the
functionality
for
the
mobility
of
mobility
management
and
on
the
user
plain.
D
So
next
thing
is
use
ipv6
encapsulation
by
default-
maybe
it's
the
SRH,
so
it's
thereby
supporting
other
type
of
PDU,
much
easy,
ipv4,
Ethernet
or
any
kind
of
things
and
I
think
the
readability
of
the
drawers
with
dramatically
improved
ID.
So
we
added
a
very
precise
packet
flow
diagram
and
please
take
care.
It
and
I
think
it
would
be
much
better
to
understand
how
a
service
clocks
for
mobile
use
of
grain,
and
also
we
cover
the
interworking
scenario
with
the
existing
is
the
friend
gtp.
D
So
next
step
is
to
deal
with
the
the
feedback
from
the
community.
We
will
extend
the
a
function
coverage
that
make
sure
the
capability
to
carry
the
gtp.
You
message
and
the
extension
header
as
it
is,
or
some
sort
of
the
trans
translation
much
suitable
for
the
SLV
six
month
and
we
will
integrate
other
type
of
the
UPF
function,
a
UPF
such
as
the
traffic
measurement
interception
and
map
performing
and
charging
etc.
So
that
would
be
our
next
step.
D
So
then
we
update
the
draft
and
then
when
we
did
see
a
certain
label,
the
core
team
will
ask
the
Road
ask
oh
I,
mean
I'm,
not
sure
the
way
to
be
happen,
but
it
must
be
close.
So
the
part
of
the
draw
this
right
is
the
difference
as
much
as
trendy
for
this
CD
PP
format.
That's
protocol
stock
picture
make
Korea
the
Whizzer,
which
is
the
the
SL
basics
portion
in
the
user
plain
side.
D
D
So
we
have
a
pretty
good
at
least
right
for
each
of
the
deployment
scenario
mode.
This
is
just
traditional
mode.
I
have
to
save
it
the
time
and
let's
just
go
through
the,
what
we
have
in
the
slide
deck.
So
we
have
a
pocket
role
in
a
pink
in
and
also
down,
link
as
well,
and
we
introduced
a
new
and
the
function
of
the
is
our
basics
to
support
a
traditional
mode
and
the
map.
With
quite
simple,
please
see
the
document
so
in
a
small,
different
diagram
here
and
some
interesting
points
listed
on
this
side.
D
The
packet
flow
up
in
here
and
downlink
and
also
be
mentioned
at
the
ipv6
gtp
case,
so
CPP's
type
protocol
stuck
here
to
GPU
on
top
of
the
ipv6
on
the
entry
interface,
so
DeForest
I-I'm
here
and
some
objective.
Is
they
ship
to
deal
with
and
so
a
pink
diagram
diagram
here
packet
flow
is
being
a
bit
complicated,
but
I
think
you
can
find
the
house.
The
dis
interworking
scenario
works.
D
Honestly.
We
need
to
reduce
the
new
and
the
function
and
I'm
gtp
six
decapsulation
in
direction
of
the
uplink
and
downlink
traffic,
a
diagram
here
packet
flow.
You
can
see
that
and
the
interface
as
it
is
right
now
introduce
a
new
end
function.
On
the
downlink
side,
it's
named
and
M
G
tp6
encapsulation
from
daikon
to
the
ether
basic
to
gtp
direction.
D
So
we
this
is
the
almost
same
with
the
first
version
that
cover
the
ipv4
TTP
case,
so
cspp
style
protocol
sub
picture.
Here.
If
I
ask
apology
interactive
as
as
much
the
same
with
the
gtp
physics
case,
some
of
our
objective
in
this
tip,
uplink
Patrol,
you
can
see
darling
as
well.
We
introduce
the
function
of
them.
Did
an
MDT
p4e
as
much
as
at
same
as
the
previous
version,
and
also
TM
team
up
in
the
direction
of
the
encapsulate
and
gtp
to
SL
basics
case
we!
A
Thank
You
sakura-san
before
one
her
one
question
actually
looks
like
in
the
recent
version,
you
included
a
removed
non
encapsulation
insertion
of
a
sorry
shader
without
a
v6
setter,
so
any
possibility
of
operating
the
plane
mode.
In
other
words,
you
know
essentially,
without
you
thought
tunnel
requirement,
in
other
words,
insertion
of
a
cirrage
without
the
tunnel
requirement
any
possibility
you
will
be
able
to
get
the
you
know
find
some
consensus
on
that.
Each.
D
Yeah
in
the
case
of
the
ipv6
video
case,
it'll
be
possible
to
to
deploy,
but
it's
depend
on
the
standardization
one,
implementation
and
I,
but
at
least
an
implementation
side.
We
already
have
in
such
a
mode
or
the
Avakian
code.
Okay,.
C
Answer
Krishnan
speaking
as
a
leaf
about
this
and
six
month,
the
segment
Orion
had
a
service.
Expesive
ID
today
does
not
allow
insertion
in
six
month
and
pending
I
think
like
darkness
behind
me,
so
he
prolly
speak
about
this
as
well.
So
there
is
another
draft
which
explains
how
to
do
this
safely.
So,
unless
like
how
to
do
the
safe
is
written,
it
won't
be
there.
Okay,
but.
A
C
That's
okay,
so
the
safe
header
insertion
draft
is
supposed
to
specify
the
cases
where
this
is
safe.
So
the
idea
was,
if
you're
in
a
closed
domain
and
you
can
specify
the
applicability
and
the
idea
is
to
make
sure
that
a
host
that
doesn't
know
anything
about
the
insertion.
It's
not
going
to
get
an
ICMP
error.
It
cannot
pass
or
understand.
I
do
anything
about
okay.
So
as
long
as
that
condition
is
specified
by
this
other
graph,
which
has
not
been
updated
in
a
while
and
explained,
then
we
can
consider
it.
C
A
C
N
We
have
introduced
the
you
forgot.
You
cares
if
you
case
is
activates.
In
such
a
case,
a
purifier
should
have
been
included
in
metidq
header,
but
on
para
for
your
paper.
You
cannot
tell
your
people
body,
you
know
PP,
you
to
know
the
total
amount
of
clear
I'm,
also
in
the
second
moments
in
the
faculty.
N
There's
a
toast
to
that
also
may
be
a
SMF
and
their
ups
there's
a
kitty
bu,
because
there's
a
two
type
of
when
they
you
in
the
other
mode,
the
to
type
or
know
that
they
haven't
that
an
imitator
once
a
UPF
and
knowledge
SMS
equanimity
is
packet
in
some
method,
then
the
SMS
sure
data
forwarding
the
the
data
probability
when
the
E
and
I
connect
mode.
In
some
case
there's
also
a
GPU
is
used.
N
T
V
Dudes
from
Cisco
Systems,
just
if
I
can
I
just
wanted
to
mention
the
the
encapsulation
thing
and
since
it's
come
up
a
couple
times,
so
the
I
believe
and
correct
me
if
I'm
wrong
here
store,
oh
that
the
year
you
have
an
IP
and
IP
encapsulation,
the
IP
encapsulation
is
ipv6
encapsulation,
which
also
contains
a
floor
able
which
allows
for
load
balancing
to
be
done
on
that
flow
label.
So
in
this
case
IP
and
IP
encapsulation.
Do
you
know
what
mentioned
we
don't
have
that
original
problem?
Okay,
if.
A
H
H
A
D
R
R
R
A
O
Movement
delay,
minimization,
is
voicing
one
of
the
most
important
one
is
smooth
transition,
smooth
migration
path
and,
last
but
not
least,
is
advanced
mobility
and
explore
types
of
urges.
So
these
are
the
challenging.
Obviously,
if
that
RG
is
trying
to
dress
in
in
in
there
in
there
in
the
new
study
are
examined
that
has
been
put
forward
next,
please
so
how
I
saw
every
six
fits
into
this?
You
just
provide.
O
In
the
protocol
that
we
have
been
studying
for
for
this
task,
so
the
the
current
state
of
the
business
is
basically,
you
have
gtp
in
you
PFC
and
then
the
first
thing
that
we
do
is
to
read
the
end
for
as
untouched
as
possible
and
drop
us
our
v6
as
a
drop-in
replacement
for
gtp
carry
the
gtp
information
within
this.
Our
v6
are
there
as
tid
or
options
or
tlvs
or
such
and
such
and
poor,
and
then
this
basically
allow
us
to.
O
O
Service
shading
and
entropy
that
comes
naturally
with
services,
so
these
are
again
in
some
cases
we
have
to
actually
make
changes.
The
question
is
that
how
are
we
going
to
obtain
these
TR
say
it
sits?
There
are
different
values
and
different
options
that
we
have
in
terms
of
obtaining
those
things.
Smf
can
be
involved
to
not
be
involved
with
third
party
to
be
involved,
so
there
are
also
different
options
that
we
can
actually
use
to
to
obtain.
O
These
sets
for
for
traffic
engineering
entropy
and
an
service
chain,
so
I
think
I've
covered
basically
slicing
in
a
very
short
form.
If
you
have
to
the
slices,
they
can
actually
share
resources
in
the
network
or
they
can
actually
you
to
complete
these
20s
or
v6
lends
itself
to
this
diagram.
It's
designed,
but
really
very
easily
sliced
one
on
his
voice
too.
In
this
case
they
are
different
paths
and
then,
but
but
it's
like
still
requires
to
have
backup
or
protection
paths.
You
can
actually
start
using
the
resources
from
this
slice.
O
O
P
Going
to
hold
off
till
the
end,
but
if
you
want
to
take
it
now,
no
offers
no
yeah
using
a
Sid
to
say
this
is
subject
to
legal
intercept
is
an
idea
that
is
really
bad
and
would
not
play
you're
advertising
to
the
whole
network
for
any
man
in
the
middle
attack
that
the
that
particular
flow
is
under
surveillance.
But.
O
O
O
Without
making
too
many
changes
into
the
the
control
plan,
now
they
want
to
actually
go
toward
Jasmine
Beach
and
it's
rb6.
These
are
bad
thing,
but
fixable
what
convergence?
What
we
can
do
here
now
we
can
actually
start
looking
into
to
the
mode
of
operation
where
we
use
some
sort
of
ideologue
architecture
in
conjunction
with
us
or
v6
to
to
provide
advanced
mobility
and
conversation.
So
what
we
do?
O
The
first
option
basically
uses
the
UPS
as
as
the
our
ease,
but
Yogi's
have
no
clue
that
they
are
running
s
for
me,
six,
they
use
gtp
to
Accord
a
connectivity.
The
gtp
is
intercepted
by
the
look
locators
and
the
locators
are
basically
or
anyone
that
are.
We
were
running
SR
v6.
They
use
the
only
law.
Mapping
system
that
mapping
system
can
be
anything
can
be
a
separate
control
plane.
It
could
be
a
common
database.
O
W
W
A
G
Value
and
I
was
comment
making
what
to
make.
This
comment.
I
think
that
from
some
of
the
presentations
you
know
before
from
Deena
when
I'm
sure
other
people
behind
is
that
basically,
none
of
us
are
changing
their
gtb/c
everybody's,
changing
Georgia,
Peach
JTP,
you
so
basically
I
find
there
are
diagrams
important,
but
I
want
to
underline
that
every
single
solution
that
we
are
going
to
is
actually
you
can
use
the
same
diagram
with
the
different
name.
Thank
you
all
right,
concluding
the
ones.
O
Here
for
well
I
want
basically
the
working
group
to
start
thinking
about
what
kalyana
me
myself
and
Dino
and
others.
We
are
working
on
a
draft
and
we
are
trying
to
position
that
as
a
draft
that
compares
different
architecture,
different
protocols
for
3d
TV
and
we
can
try
and
be
a
real
choice,
actually
get
momentum
here
to
position
this
as
a
partial
response
to
back
to
3gpp
and
would
like
to
charities.
If
you
consider
that,
in
their
response
back
Thank.
A
M
M
For
a
lot
of
people,
we
had
a
few
calls
on
the
DMM
working
room.
Some
participants
came
in
from
the
3gpp
side
and
they
helped
us
look
at
the
architecture
and
give
us
some
input
into
what
would
be
useful
to
have
in
such
a
people
that
we
can
use
as
guidelines,
and
there
were
references
to
references
from
the
city
for
LS,
and
then
there
were
two
references
that
are
from
the
essay
to
architecture,
and
we
have
come
to
know
that
HC
is
also
working
on
an
ng
P
next
generation
protocol.
M
Looking
at
the
same
kind
of
work
and
trying
to
do
the
same
kind
of
evaluation,
it
looks
like
there
may
be
benefit
of
that.
They
will
also
benefit
from
this
evaluation
being
conducted
right
now.
There
are
several
protocol
candidates
in
IE
PF,
so
we
started
with
a
few
of
them
like
a
service,
it's
less
violate.
M
However,
as
we
have
been
progressing
on
the
work,
we
find
that
the
control
three
protocols
and
the
use
of
nine
protocols
can
be
mixed
and
matched,
and
some
of
that
is
coming
up
right
now
and
as
you've
seen
in
the
list,
Buddha
talked
and
as
well
as
in
the
SR
v6.
And
there
is
this
document,
we
call
it
a
document.
Maybe
we
should
see
how
to
move
this
I
think
this
came
up
on
the
email
list.
M
Should
this
be
an
individual
document,
or
should
this
be
considered
as
a
working
group
document
and
the
scope
of
the
work,
and
then
we
guided
such
that
it
is
beneficial
to
the
IETF
3gpp
and
the
HC
community's
little
background
on
the
architecture.
I
got
a
little
bit
of
feedback
from
the
IADC
IETF
side
of
folks,
saying
that
they
hear
all
these
terminologies
and
not
clear
exactly
how
they
all
fit
into
the
architecture.
We
may
not
have
a
fully
blown
tutorial
right
now,
but
maybe
we
love
it
out
a
little
bit
later
in
another
session.
M
So
what
is
shown
here
is
from
the
3gpp
essay
2
specification,
23.5
o1.
They
have
two
kinds
of
representation
for
the
architecture.
There
are
network
functions
specified
and
there
are
interfaces
specified.
The
left
side
of
the
diagram
shows
what
is
called
a
reference
point
representation,
and
this
one
is
a
similar
to
what
you
are
all
familiar
in
the
4G
architecture
in
4G.
You
are
familiar
with
the
Mme
SK
to
Fe
gateway,
PCRF
and
HSS.
In
the
5g
architecture,
there
has
been
a
slightly
different
spirit
of
the
functionalities.
M
Amf
loosely
is
called
access
and
mobility
management
function,
but
it
has
some
functions
of
the
mme
SMF.
Is
the
services
management
function
session
management
function,
however,
in
file
G,
the
NASS
is
terminated
on
SMF
and
not
on
the
mme,
like
in
the
4G
and
then
pcrf
is
now
called
policy
control
function.
Charging
has
something
different:
the
authentication,
a
USF
is
authentication
server
function
and
you
DM.
In
the
4G
architecture,
there
was
a
triple-layer
and
HL
h.
Ss
function
that
a
triplet
function
that
was
included
in
the
HSS
part.
M
There
is
similar
functionality
here
there
is
a
new
function
called
MSS
F,
it's
the
network,
sly
selection
function,
and
this
is
one
of
the
features
and
by
G
the
UPF,
what
they
did
in
the
4G
architecture.
They
are
scalar,
P
gateway
has
been
combined
and
then
they
split
it
into
the
control
plane,
user
pain
part.
So
the
s
MF
is
the
control
plane.
Part
UPF.
Is
the
user
plane
part,
so
that
is
kind
of
what
you
see.
M
There
are
a
lot
of
point-to-point
interfaces
and
the
concern
was
that
having
interoperability,
testing
and
running
and
operating
these
networks
is
not
going
to
be
simple,
so
the
services
based
interfaces
have
been
defined.
Also
in
the
same
release.
The
write
diagram
shows
that
the
functions
are
same.
There
are
some
additional
functions
and
each
of
the
control
plane
functions
has
a
services
based
interfaces
and
each
function
would
say
what
services
it
offers
with
the
rest
of
the
functions
and
they
are
registered
in
what
is
called
an
N
RF
network
service
repository
function.
M
There
is
also
a
data
storage
function,
actually,
the
it
is
also
applicable
for
the
difference.
Point
function.
If
some
of
these
functions
are
stateless
can
share,
there's
a
storage
function,
they
can
do
so
and
then
there
is
an
N
EF,
which
is
a
network
exposure
function.
The
the
services
that
are
offered
in
here,
or
the
data
collected
in
here
can
be
used
by
the
other
parts
of
the
network
or
for
exposure
to
the
external
parties.
So
that
is
a
quick
rundown
of
what
that
the
architecture
is.
M
So
what
we
will
try
to
do
in
this
work
right
in
this
document
is
we
are
looking
at
the
services
based
interfaces,
so
the
diagrams
that
you
have
seen
in
the
Lisbon
there's
our
v6
thing
kind
of
fit
into
the
architecture
that
is
shown
on
the
right
side.
There
is
an
SM
F,
and
you
see
the
UPF
2
ups
with
n
line
between
them,
and
that
is
kind
of
what
you
saw
in
the
documents
in
the
presentations
before
in
the
document.
You
see
how
the
rest
of
the
other
protocols
fit
into
the
architecture.
M
3Gpp
also
has
roaming
architectures
in
the
relevance
for
this
is
n9
sometimes
goes
across
operator
boundaries
that
these
terminologies
are
very
specific
to
3gpp
mobility
and
roaming,
and
some
of
the
IETF
folks
wanted
to
know
what
the
definitions
are.
We
found
that
there
is
a
3gpp
document,
21.9
or
file
that
has
the
definitions
of
those
and
the
requirements
are
given.
We
put
up
the
specification
names
and
there
are
also
specific
section
references
to
say:
we're
roaming
requirements
are
listed
in
the
mobility
requirements
are
listed
in
the
roaming
itself.
There
are
two
scenarios.
M
One
is
called
the
local
breakup
scenario
where,
when
the
Yui
is
roaming
in
a
visited
network,
the
service
could
be
offered
by
the
visited
network
itself.
However,
the
authentication
is
done
with
the
home
network
and
the
policy
could
sometimes
come
in
from
the
home
network.
The
home
router
case
is
where
the
services
are
offered
by
the
home
network
and
in
the
bottom
diagram
you
see
a
nine
crossing
the
operator
boundaries.
M
Related
to
the
n9,
we
have
some
samples
in
there.
For
example,
this
top
diagonally
show
has
multiple
PD
when
iue
connects
to
two
different
data
networks
for
with
two
different
PD
you
sessions.
So
in
that
diagram
you
see
the
G
not
be
connecting
to
a
UPF
entity
at
your
network
with
its
own
SMA.
For
each
of
the
two
cases,
the
network
design
decides
whether
there
is
an
n9
or
not,
but
the
sample
thing.
We
show
the
second
session
connecting
to
a
a
UVs
with
an
align.
M
So,
regarding
the
requirements
for
just
study
here
is
or
this
evaluation
is,
the
UPF
requirements
are
specified
in
there.
Basically,
the
mobile
anchor
for
UPF
is
the
mobile
anchor
point,
and
it
it
has
all
the
packet
processing
routing
all
the
functionality
that
you're
familiar
with
in
the
4G
architectural
remains.
There
are
new
viewers
tables
in
the
specification.
They
also
have
a
reflective
QoS
this
time,
so
the
n9
has
to
map
map
all
of
those
as
well.
M
So
those
requirements
are
in
those
sections
and
we
have
identified
a
few
scenarios
for
evaluation.
It
would
be
good
if
we
get
input
from
the
other
operators
as
well.
What
we
have
fakest,
not
roaming,
scenarios,
where
the
UE
connect
connects
to
the
Internet.
There
are
three
mobility
cases
defined
in
the
architecture.
M
One
is
called
a
service
and
session
continuity
mode,
one
which
means
that
the
IP
address
does
not
change
and
services
a
session
continuity
mode
to
where
the
IP
address
can
be
dropped.
Them
then
reconnected
service
and
session
continuity
mode.
Three
means
that
it
can
share
a
draw
a
friend
connect.
However,
there
is
no
impact
on
the
service.
M
What
this
means
is
you
are
making
the
session
before
you
are
breaking
it.
So
that
is
why
the
service
is
having
continuity
and
the
other
scenarios.
Are
you
e
to
UE,
because
everybody
is
talking
about
the
low
latency
applications,
so
we
need
to
ensure
that
the
unity
UE
packet
closed
happen
and
then
mobility
also
I
supported
you
a
two
two
data
networks.
One
multiple
video
sessions
and
single
video
sessions-
those
are
what
we
picked
up
and
then
roaming
scenarios,
local
breakout
and
home
routers.
M
M
Not
all
slices
may
be
using
the
same
mobility
management
mechanism,
so
the
system
has
to
support
slices
with
each
of
each
slice
could
have
a
different
mobility
management
mechanism,
and
we
would
like
to
see
how
the
protocols
behave
in
such
scenarios,
and
we
have
one
requirement
that
says
if
a
uni
is
connected
to
multiple
slices
using
different
mobility
protocols,
we
hoping
that
we
address
that
as
well.
Then
we
wanted
to
see
what
the
impacts
2
&
2,
&,
3,
&,
4,
&
6,
the
n9
around
the
in
mind.
There
are
these
other
interfaces.
M
N
M
Okay,
maybe
we
can
refine
the
requirements.
We
started
these
as
high
level,
so
once
we
get
some
more
input,
if
there
is
input,
please
identify
so
we
can
get
consensus
that
these
are
the
right
requirements
we
are
looking
at,
and
so
you
have
seen
of
these
protocol
stacks
in
the
individual
presentations
before
just
to
put
them
in
the
context.
M
The
top
left
tile
shows
the
existing
party
protocol
stack,
so
the
m3
interface
and
nn3
net
interface
is
between
the
G
naught
B
and
the
UPF,
and
there
is
a
gtp
or
UDP
override
IP
over
there,
and
then
there
is
an
anion
interface
between
the
UPF
closer
to
the
edge
and
the
UPF
that
is
closer
to
the
internet
or
the
data
network.
That
is
the
video
session
anchor.
M
M
I
think
you've
seen
all
of
these.
Let's
go
to
the
next
steps.
What
we
would
like
to
do
is
expand
on
the
description
of
the
protocols
in
the
document
for
roaming
charging
security
and
scalability
and
other
stuff.
What
we
would
like
to
ask
is:
can
the
DMM
adopt
this
draft
as
a
working
document,
and
then,
when
we
respond
back
to
3gpp,
we
can
send
a
copy
of
this
and
we
were
included
as
I
do
in
here.
M
D
Thank
you
very
much
for
your
such
extension
really
appreciate
the
standpoint
of
a
city
for
I'm
really
appreciated,
and
at
the
I
mentioned
before
the
that's,
the
work,
we'll
really
sure
so,
but
I
really
appreciate.
If
you
can
make
the
document
much
easy
to
digest
so
that'll
be
a
very
concise
compiled
some
criteria,
some
other
dependency
on
the
architecture
of
the
generation
which
Suresh
mentioned
before,
like
so
with
this
is
the
user
friendly
solution
technology
and
also
otherwise,
it
kind
of
the
that
require
architecture.
H
D
X
Push
the
columns
and
rails
very
quick,
yes,
sorry
I
just
were
to
speak
strongly
against
the
working
group
adopt
u.s.
or
using
this
as
the
response
to
the
request
from
3gpp.
The
request
was
for
a
list
of
things
that
are
relevant.
I,
think
that
is
something
we
can
produce.
A
full
comparison
in
evaluation
would
be
a
very
bad
idea
at.
A
A
You
know
this
is
the
drug
that
you
know:
I
posted,
actually
I.
Think,
unfortunately,
dapeng
is
not
my
coach,
it
is
not
there,
but
this
is
a
individual.
So
so
this
is
essentially,
if
you
look
at
nothing,
what
irony
and
utter
asanas
presented
over
all
with
respect
to
you
know:
Phi,
G,
user,
plane,
optimizations,
I,
think
the
key
goals,
or
you
know.
How
do
we
realize
an
axis
independent?
You
know
use
a
plane,
but
is
one
aspect?
Second
thing
is
you
know
you
know?
How
do
we
truly?
A
You
know
leverage
the
data
plane
has
some
optimal
routing
in
the
user
frame.
If
you
look
at
the
classic
mobility
or
architectures,
it
was
always
an
anchor
based
approach
where
we
would,
you
know,
steer
the
traffic
from
the
anchor
point
so
with
respect
to
routing
they
were
like
many
in
efficiencies
and
I.
Think
you
know
it's
trying
for
to
really
remove
that.
You
know
inefficiencies
in
the
data
plane
right.
That
is
one
aspect
and
also
given
that
you
know
the
next
set
of
you
know
billions
of
IOT
devices.
A
If
you
are
the
skill,
the
network
to
few
billion
IOT
devices,
you
know
question
is
what
kind
of
you
know
what
kind
of
optimizations
to
be
needing
the
user
plain
I
think
that
is
another,
and
the
one
thing
is
renewal
of
the
anchor
I
think
you
know
today
we
draw
everything
through
centralized
mobility
anchors,
so
that
will
take
a
GG
s
and
corrosion
P
gateway,
or
any
of
that
fundamentally,
you
know
we
always
had
an
anchor
right.
The
question
is:
how
do
we
remove
the
anchor?
A
So
if
you
look
at
the
steady
item
that
Kalyani
talked
about,
it
was
mostly
about
inline,
but
when
you
talk
about
inline
this
anchor
right,
the
anchor
is
not
going
away.
So
here
only
between
two
points
and
then
line
between
pull
ups
here
yeah,
you
know
Ramon
the
tunnel,
but
fundamentally
the
anchor.
Is
there
I
think
so
this
specification
is
this.
Work
is
about
like
how
do
we
remove
the
anchor
from
the
user
plane?
I
think
that
is.
A
The
goal
is
also
to
you
know
that
angle
and
now,
with
respect
to
you
know
whatever
I
do
we
have
to
coexist
with
the
cups
architecture?
You
know
because
most
of
the
you
know,
mobile
out
pictures
already
heading
in
that
direction,
with
respect
to
user
and
control
plane
separation.
So
that
is
another
thing
now.
So
if
you
look
at
it,
you
know
symbolic.
You,
like
you
know
as
a
picture
right.
You
are
your
use
of
linear
separated
in
the
use
of
plane.
You
don't
want
any
any
protocol
stack.
Essentially,
the
control
thing
is
already
separated.
A
I
think
you
know.
So
what
the
only
solution
that
we
pick
right,
then
we
don't
want
to
introduce
a
control
plane
again
in
the
use
of
plane
right
because
the
goal
is
to
be
already
separated
the
control
plane
out
from
the
use
of
thing
right.
So
now
what
we
do
in
the
use
of
pliny
is
whatever,
if
you
can
lower,
is
the
rounding
semantics
in
steer
traffic
I?
Think
that
will
be.
You
know
that
good
so
in
general,
I
think
this
is
like
you
know,
information
of
centralized
ankles.
A
You
know
optimized
optimal,
routing,
no
tunnel
hole
right.
You
know.
Obviously
you
want
to
scale,
but
that
subject
you,
but
in
straight
direction.
I
think
you
know
you
the
state
that
we
introduced
in
that
user
plane.
You
know
as
far
as
possible,
it
has
to
minimal
right,
and,
finally,
you
know
do
a
little
lower
is
the
the
cures
that
is
present
in
the
transport
and
you
have
to
steer
the
trans.
You
know
travel
with
some.
A
You
know
traffic
awareness,
so
so
the
approach
that
you
know
what
you
know
we
are
proposing
ready
to
think
you
know
how
do
we
realize
that
with
s
our
basics
of
or
any
other
use
of
plane,
which
can
you
know,
steer
the
traffic
in
optimal
plane?
I
think
this
is
more
about.
You
know
how
we
you
know
if
this
is
an
approach
where
you
don't,
you
know
where
we
can
get
rid
of
the
anchors.
I
think
that
is,
let's
look
at
quickly
a
call
flow
right.
A
You
know
in
this
approach
when
the
mobile
node
comes
up.
Initially,
it
is
anchored
at
the
local
access
gateway.
It's
almost
like
a
first
approach
right.
This
is
your.
You
know
starting
point
right,
so
the
IP
address
that
is
given
to
the
mobile
or
the
UE
essentially
is
from
the
local
address.
Prefix
that
is
anchored
on
that
on
that
access
gateway.
A
Now,
at
this
point,
if
you
see
the
traffic
flows,
the
flows
are
optimized
right,
essentially,
they're
not
announced,
there's
no
use
of
plain
state
right,
essentially
because
yeah
you're
all
gravity
is
entering
from
the
wrong
after
the
base
station,
your
first
off
everything
enters
and
exits
from
that
right.
So
in
that
sense
you
know
traffic
is
optimal.
The
user
has
multiple
predictors
with
colored
prefixes,
one
prefix
with
address
properties
and
other
prefix
without
at
this
properties.
But
you
know
the
question
is
now
the
mobile
moves
right,
you
know,
what
do
we
do
with
this?
A
A
mobile
you
chose
on
a
g2
right
now
it
moves
to
a
g6.
Now
they
were
acting
active
flows
at
this
point
now.
What
would
happen
to
the
cycling
flows
at
this
point?
This
is
where
you
know,
I
think
you
know
what
we
are
saying
is.
Essentially
you
know
initially
with
the
who's
that
event
would
be.
You
know
you
would
get
that
in
the
control
plate
somewhere
in
the
you
know,
there's
a
mobility
controller,
we'll
get
an
update
from
the
SMF
AMF
or
from
the
rear.
A
Sly
in
the
Radio
Network
right,
based
on
that
event,
you
know
we
know
that
the
mobile
is
now
somewhere
in
the
base
station.
You
know
attached
to
a
g6
now
in
one
approach
at
this
point
if
their
default
policies
that
can
be
push
to
a
g2
and
a
g6
which
says
that
you
know
any
traffic
for
this
previously
assigned
prefixes
can
be
directly
routed
between
82
and
86.
That
is
one
aspect.
Second
thing
you
also
using
the
semantics
of
a
service.
A
We
can
say
that
you
know
any
time
you
see
an
optimal
flows
right
report,
those
flows
to
the
mobility
control
3.
So
this
is
more
about
steering,
but
in
this
case,
essentially
still
this
anchor.
If
you
for
a
transient
pay
off
time,
the
traffic
will
still
go
through.
You
know
essentially
like
this
from
a
j2
to
a
g6,
but
we
don't
want
this,
because
this
is
like
the
pinhole
routing
that
we
we
have
had
this
from
all
along
right
now.
A
Can
we
leverage
your
son,
v6
or
some
other
in
approaches
where
you
can
get
rid
of
the
the
initial
anchor
from
the
path?
So
this
is
where
I
think
you
know
the
default
policy.
The
very
initial
policy
after
the
radio
went
is
you
are
saying
any
traffic
that
you
see
from
a
g2
to
steer
it
to
eg6
right.
The
second
thing
is
any
optimal
flows.
If
you
see
a
report
that
to
the
control
plane.
So
now
the
initial
flows,
which
are
there
at
this
point,
is
seen
to
be
storming.
A
A
If
we
can
push
us
all
v6
or
some
other
user
plane
policy,
it
says
that
you
know
traffic,
you
know,
send
it
directly
from
you
know:
r3
do
a
g6
or
energy
for
286
at
this
point
for
those
flow
there.
No
there.
No,
there,
no
active,
you
know.
Essentially
the
agent
is
no
longer
in
the
fact.
So.
Finally,
at
this
point
you
know
once
you
know
this
first
on
it
at
some
point,
those
flows
will
terminate.
We
remember
the
prefix
and
we
now
give
a
prefix
from
the
localized
anchor.
A
At
that
point,
the
again
the
network
has
converge
back,
so
we
started
with
no
stick
and
for
a
transient
period
of
time,
when
there
was
mobility
evening
we
created
some
stairs
and
I
saw
basic
state
or
some
steering
state
and
later
we
removed
that
state
right
now
again,
the
traffic
you
know,
essentially,
you
remove
the
state
will
not
announce
in
the
user
plane.
Essentially,
you
know
we
are
back
to
the
square
one
without
any
any
special
steering
straight
eternal
state
on
an
anchor
in
the
in
the
conclusion.
A
So
that
is
somebody
I
think
that
this
is
about
elimination
of
fixed
angles,
elimination
of
user,
plane,
tunnels,
right
traffic,
stealing
with
surveillance,
because
we
can
use
some
semantics
in
the
innovation
by
service.
Essentially,
you
can
steer
the
traffic
and
what
we
have
is
an
access
agnostic,
car
plane
right.
So
this
is
the
proposal
publishing.
K
So,
thanks
for
the
presentation
now
you
in
pretty
fast
I
have
to
go
back
and
look
at
the
details,
but
it's
on
Charlie
Perkins
from
Norway,
so
you're
fitting
anchor
is
a
bad
word
and
really,
if
you
have,
if
you
think
about-
and
mobile
device
has
an
address-
let's
say
it
already
has
an
address
and
it
moves.
There's
gotta,
be
two
packets
from
the
internet
have
to
go
somewhere.
So
wherever
that
is,
and
in
that
diagram,
you've
got
interesting.
K
A
Think
you
made
a
good
point
with
respect
to
the
signalling
grid.
It
is
in
this
case.
Eventually,
you
know
there
can
be
any
time.
There's
a
mobility
event.
You
know
at
times,
they're
active
flows,
there's
a
lookup
that
happens,
but
if
you
look
at
any
of
the
mapping
protocol
seamen,
there's
a
cache
took
up
our
cache
population.
So
it's
same
thing
is
happening,
but
here
we
slightly
do
it
differently,
but
it's
a
fair
statement,
but
maybe
based
on
policy,
we
can
John.
A
B
A
A
Domains
no
I
understand
absolutely
so.
The
another
comment
is,
you
know:
well,
I,
haven't
read
it
in
full
detail,
but
just
looking
at
it,
it
seems
a
little
bit
more
like
the
Sdn
controller
you're
distributing
forwarding.
So
there
is
a
limit
to
how
much
you
can
do
that
in
some
sense,
I
mean
maybe
it's
like
a
distributed
routing,
but
this
is
more
like
yeah.
H
A
Can
do
this
selectively
right?
We
don't
have
to
do
it
for
every
flow,
every
user-
maybe
you
know
this
some
Mobile's,
which
are
rapidly
moving.
We
don't
do
this
right.
Only
for
certain
cases.
We
do
that,
but
then
what
we
realize
is
a
system
with
practically
you
know
a
very
minimal
use
of
planes.
I
think
this
is
my.
A
C
P
Did
Allen
Erickson
just
very
very
quickly?
It's,
and
this
applies
actually
to
several
things.
I've
heard
today,
and
that
is,
is
the
elimination
of
anchor
points
if
you're
purely
considering
forwarding
is
actually
relatively
straightforward.
It's
when
you
start
getting
into
requiring
virtualization
carrying
session,
made
metadata
stateful
handoff
and
in
particular
it's
that
state,
handoff
and
any
type
of
mobility
event.
That's
actually
kind
of
huge
and
you
end
up
at
classes
of
applications
that
don't
necessarily
need
all
of
these
things.
But
some
of
that
stuff
needs
to
be
there
and
is
going
to
be
inescapable.
P
N
In
1950
we
have
introduced
or
defender
through
similar
solution.
The
co
may
be
the
question
once
for
the
asymmetry
this
means,
as
in
mother
as
a
mother
suite.
Does
it
mean
when
the
UE
change
it
for
another
place,
we
can
change
the
anchor
anchor
another
solution
if
dependent
by
the
August
over
at
computing.
They
say
that
computing
means
that
they
use
the
local
area
network
or
abdication
influence.
The
user
plan
selection
a
thing
you
can
consider
or
recommend
some
some
some
weekend
in
2050.
Okay.
Thank
you.
Thank
you.
Thank.
A
I
So
very
quick
update
on
this
draft
so
well,
our
view
is
a
network
based
EMM
approach
based
on
proximal
ap
in
which
read
imagine
is
push
is
pushed
to
the
to
the
EDS
and
it's
basically
a
partially
through
the
solution.
We
have
with
a
centralized
control,
plane
and-
and
it's
rooted
at
away
so
there
are
in
terms
of
entities,
are
two
types
of
entities.
I
What
is
the
called
mobility
on
:
access,
router,
which
is
basically
the
access
router,
where
the
I
know
that
s2
and
it's
the
one
concentrating
the
anchors,
router
lmam,
Mac
functionality,
perma
one
out
/
prefix
and
it
wandering
eating
the
prefixes
that
are
the
anchor
to
the
different
mobile
nodes
and
sessions.
And
then
we
have
the
centralized
mobility
database,
which
is
basically
kind
of
an
LMA
for
the
control
plane,
so
very,
very
quickly.
I
The
operation
very
simple
when
you
have
initial
attachment,
you
connect
mobile,
not
connect
to
this
mark
and
then
mark
signals
are
to
the
centralized
salamé
since
then,
the
state-
and
it
gets
a
preface
locally
anchor
here.
So
there
is
no
tunneling
here
today
to
the
to
the
internet.
It
then
that
mobile
node
moves
to
another
mark.
I
Basically,
the
same
thing
money
will
be
sent
to
the
to
the
CMD,
and
then
the
CME
will
signal
back
to
both
the
current
Mar
were
the
Mona
Lisa
touched
and
the
previous
mark,
where
the
mobile
thought
was
attached
to
allow
the
session
continuity
for
the
prefixes
that
are
anchored
at
different
marks,
in
addition
to,
of
course,
provide
a
new
anchor,
a
new
graphics
or
any
prefix
to
the
to
the
more
well-known
locally
and
Corky.
So
we
can
use
for
the
new
sessions
the
optimal
path
with
no
panel
al
for
the
ongoing
session.
I
We
can
keep
the
decision
continued.
So,
as
I
mentioned
this
running
code
for
deals,
we
even
show
in
the
in
the
ITF
in
different
occasions,
and
we
are
actually
keep
on
using
this
even
for
teaching
purposes
at
the
University.
So
it's
running
code
for
the
returning
code
for
this,
so
they
are
very
quickly
going
to
the
last
slide.
So
there
was
Russian
zero
zero.
That
was
the
first
one
merging
a
couple
of
previous
solutions
raft
that
was
presented
in
Singapore.
I
We
got
some
good
consensus
for
adopting
the
document
back
there.
Then
we
have
the
Russian
one
that
it
has
been
posted
before
London
we
address
uncommitted
when
you
sit
in
Singapore
from
Daniel
from
three
we
have
been.
We
have
had
some
discussion
on
the
Middle
East.
We
have
received
comments
from
and
abused
on
the
same
panel,
exact
forum,
Daniel
and
I.
Think
hi
also
resistors
and
commissions
are
total
dude
in
the
previous
to
this
meeting
and
I
also
got
some
offline
discussion
with
with
Danny
Moses.
G
K
Charlie
Perkins
from
Norway
in
future
way,
so
there's
been
a
lot
of
work
in
this
general
space
and
this
is
a
nice
version
of
it.
But
some
of
the
previous
work
had
important
things
like
pre-registration
and
context
transfer
and,
and
there
was
a
document
called
fast
mobile
IP,
which
I
think
should
be
considered
relevant
to
this
so
I'm
wondering
well,
I
mean
I
guess.
One
of
the
features
of
this
is
a
creation
of
something
logical
interface,
basically
and
I.
K
Wonder
if,
if
you
consider
this
to
be
an
intermediate
point
along
a
general
design
or
something
and
this
it
has
a
lot
of
resemblance
to
what
Shri
was
just
saying
in
certain
ways
and
so
I
think
what
I'm
trying
to
say
is
it.
It
opens
the
door
to
a
significant
design,
space
and
I'm
wondering
if
you
agree
that
there
could
be
some
lot
of
modifications
in
the
future.
Yeah.
I
I
mean
mr.
Baum
I
started
to
be
like
kind
of
direct,
straightforward,
dmn
solution
from
pimmit,
but
of
course
we
agree
it's
the
document,
I
mean
if
the
working
group
part
of
the
document
and
they
weren't
going
to
believe
that
we
should
improve
our
other
functionalities
and
going
a
different
I
mean
evolved
direction.
Yeah.
Of
course,
we
would
agree
to
do
so
yeah
to
consider
other
other
things
and
to
to
extend
the
solution
with
that,
but
the
I
mean
the
kind
of
the
trigger
for
the
document
was
like.
Okay,
there
is
this
diamond
race.
K
The
mm
well
I,
guess
part
of
my
motivation
for
asking
the
question:
is
that
we're
you
know
focusing
a
lot
of
attention
on
5g
and
5g?
Is
nothing
except
you
know
if
it's
not
high-performance
and
so
adopting
a
new
approach
for
local
mobility.
I
think
should
take
a
lot
of
effort
to
make
sure
it's
high
performance
and
fulfill.
Some
of
that
need.
I
A
So
in
the
last
20
or
500,
we
pull
the
working
group
and
there
was
consensus,
but
at
that
time
I
was
not
ready,
because
I
mean
I
did
not
see
much
shot.
Many
discussions
in
the
mailing
list
right
but
looks
like
lately.
There
were
some
discussions.
I
saw
a
bunch
of
people,
commenting
on
it
right
and
so
I
think
just
a
rough
poll
right.
You
know
anybody,
you
know
how
many
people
you
know
support.
The
adoption
of
this
document
is
a
working
group
document.
A
Okay
Suresh:
what
do
you
say?
Okay,
well
like
I,
think
so,
roughly
there's
consensus,
but
you
know
we'll
discuss
with
the
ad
and
we
will
pull
them
again
will
confirm
the
same
in
the
mailing
list
and
if
there
is
sufficient
interest,
if
there
are
no
objections,
we'll
take
this
up
as
a
working
document.
Okay,
thank
you.
Next
is
wuji's.
You
know,
I
think
there
are
a
extensions
for
on-demand
mobility
at
this
is,
you
know,
be
requirement
hochi,
my
apologies,
but.
Y
Y
As
you
know,
router
advertisement
rars
is
one
of
the
ways
for
an
IP
know
to
configure
its
routing
address.
One
of
the
things
with
distributed.
Mobility
management
is
that
there
may
be
a
need
to
allow
throw
up
options
with
particular
set
of
prefixes,
as
mentioned
earlier
by
three
there's,
a
LS
statement
from
3gpp
regarding
our
a
metadata
related
work
and,
in
particular,
they've
defined
three
SSC
modes,
and
it
was
mentioned
in
Kalyani's,
talked
a
little
bit
earlier,
and
what
sa-2
is
looking
for
is
a
mechanism
to
deliver
information
regarding
literacy.
Y
Its
helper
I
forget
his
last
name,
jegging
and
Danny
and
and
company,
and
in
that
draft
there's
these
four
modes
of
operation
session
lasting
IP
address,
fixed
IP,
address,
graceful
replacement
and
non-persistent.
The
starred
ones
are
the
ones
that
are
mentioned
in
the
SSC.
They
mapped
to
SSC
modes
for
3gpp,
alright,
and
so
there's
a
nice
mapping
between
what's
going
on
in
beyond
demand
draft
and
what's
going
on
in
3gpp,
we
presented
something
at
IETF
100
and
we
saw.
Z
Y
A
A
On
a
pillion
PDU
basis
on
how
do
you
do
a
video
based?
Okay
and
now
this
control
plane
is
going
to
advertise
in
our
a?
What
had
ad
address
type
is
right:
I
guess!
My
comment
was:
are
you
going
to
look
at
what
happens
if
the
3gpp
could
program
advertise
has
one
mode
and
they
are
rare
because
is
another
one.
A
H
A
E
Well,
once
curry
I
think
you'll
have
a
very
hard
time
getting
bits
from
six
men.
You
might
want
to
look
at
the
PVD
work.
That's
going
on
in
the
interior
work
group
which,
which
is
looking
at
expressing
this
stuff
in,
is
a
little
more
extensible
format.
I,
don't
know
that
will
do
exactly
what
you
want,
but
we've.
E
E
And
then
you
basically
match
it
so
I
yeah,
because
I,
because
realistically,
like
six
months
in
a
in
a
position
to
block
this
forever,
if
they
don't
like
it,
because
it
says
that
the
Charter
says
that
it
can
right
says
it
came
like
recommend
that
the
document
not
be
published
if
they
don't
like
it.
So,
yes,
we.
AA
A
Danny
in
a
solution,
I
had
a
draft
long
back.
You
know
we
never
managed
to
push
it
through.
You
know
he
started
in
DHCP.
Then
we
there
was
some
work
happened
in
this
working
group
right.
It
went
through
all
and
find
it
no,
but
maybe
you
know
tying.
Sometimes
you
know
right.
You
know
the
coloring
work
is
taking
off
it's
just
with
a
different
name
right.
You
know
it's
sometimes
you
know
timing,
you
know
yeah.
So
if
you.
A
E
E
So,
first
of
all,
if
you
haven't
gone
a
6-man
with
it
like,
you
should
go
right,
ASAP
and
I.
Think
if
you
know
I
feel
that
if
it
was
cast
as
like
a
link
layer,
specific
option
or
something
that
wasn't
Universal,
because
I
think
those
bits
are
basically
used
on
every
link,
type
on
every
network
right
and
so
I.
Think
that's
why?
E
AA
A
A
H
A
AB
You
find
yourself
yeah,
actually
they're,
always
back
into
the
more
times,
but
anyway.
So
thank
you
and
it
will
be
a
some
quick
update
and
party
diplomat
model
and
architectural
conservation.
Actually,
that
updates
was
given
in
the
last
meeting
and
based
on
the
comment
in
the
list
and
so
that
new
version
0
for
version
will
be
submitted
and
with
the
following
update,
but
update
will
be
bit
small
and
the
last
model
will
be
named
and
there's
some
missing
interfaces
and
the
models
will
be
society
and
actually
no.
AB
We
are
still
wait
for
comments
and
opinions
based
on
the
stabilized
that
it
is
it
wrapped.
But
here
what
the
stabilized
means
here
is
that,
as
the
opposite
draft
is
still
progress,
it's
based
on
the
implementation,
so
we
are
just
living
a
spaces
subject
to
change,
but
looking
at
this
you
know
that
FSC
draft,
and
also
looking
at
this
architectural
consideration,
deployment
that
what
you
have
adjusted
own
is
about
the
pointing
out
that
we're
F
interfaces
will
be
a
will,
be
applied
and
so
I
terms
of
that.
AB
I
think
that
there's
nothing
will
be
conflicted
with
this
current
episode,
wrapped
ongoing.
So,
anyway,
that
if
this
episode,
sir,
gives
some
comment
that
we
will
you
know
post
the
new
update
in
the
list
and
then
what
we
are
very
importantly
identifying
is
that
new
work
stream
is
being
done
and
proposed
for
use
of
plane
of
the
mobile
ethics.
And
there
were
two
following
a
title,
and
there
also
will
be
welcome
to
get
some
the
corners
of
the
some
relevant
ulcers
and
particularly
to
identify
the
impacts
of
those
works
on
the
existing
deployment
models.
AB
AB
Actually,
also
additional
some
you
know
trend
we
are
identifying.
Is
the
five-day
service
based
architecture
and
disappear
in
the
last
year
that
December,
that
you
know
this
version
of
fear
so
and
also
today,
that
we
could
have
status
on
a
couple
of
you
know
the
presentation
talking
about
the
SBA,
so
also
we
saw
that
these
people
are
models.
AB
He
almost
you
know
stabilized
with
believe
so,
but
we
also
saw
that
there
needs
to
be
considered
based
on
this
s5g
SBA,
but
actually
that
the
last
model
given
in
the
draft
icing
is
well
reflect
this
SBA
concept
and
some
in
a
in
a
big
picture.
So,
and
what
we
have
just
done
is
that
just
that
you
know
that
the
adding
does
some
interfaces
between
the
mobility
controller
and
the
control
plane
function
proves
and
also
between
these
control,
plane
functions
and
the
routing
controller.
AB
So
nothing
has
been
changed
and
only
except
adding
that
this
interfaces,
and
so,
but
actually
that
what
we
are
also
asking
is
about
that.
The
what
else
or
any
additional
zon
any
opinion
or
comment
to
be
considered,
and
if
not,
that
it
is
good
enough
or
not.
So
maybe
that
yeah,
so
in
terms
of
that,
we
in
a
solicit
that
you
are
commenting,
opinions
and
tested
Shelly.
Thank
you
see.
K
Charlie
now
quick
question:
I
guess:
quick,
I'm,
Charlie
Perkins
from
Huawei
just
curious
if
it
makes
sense
to
show
some
of
the
5g
architecture
and
how
this
is
embedded
in
that
I
mean
with
n9
and
SML.
As
I
said,
what
is
it
user
accounts
and,
though,
we've
seen
several
discussions
about
if
nagi
architecture,
yeah
and
I
think
this
is
relevant
to
that,
and
it
should
be
some
connection
made
between
those
two
yeah.