►
From YouTube: IETF99-DMM-20170719-0930
Description
DMM meeting session at IETF99
2017/07/19 0930
https://datatracker.ietf.org/meeting/99/proceedings/
A
Today
we
have
a
full
agenda,
I
think
before
we
get
started
note
well,
this
meeting
is
governed
under
certain
rules
and
regulations.
Those
are
documented
in
this
RSC's,
please
to
read
them
and
some
few
other
things:
administrative
stuff,
blue
sheets.
They
should
be
going
floating
around.
Please
sign
them
and.
C
D
C
C
A
You
also
need
a
JavaScript,
please
somebody
who
can
be
on
the
jabber.
Oh
thank
you
so
much
all
right
and
the
links
to
the
working
group
document.
So
here
all
the
slides
have
been
not
posted,
except
for
one
bit
from
sort
of
a
song.
That's
there's
some
change.
We
will
do
that
posting
later,
but
at
least
you
can
see
the
whole
diversion.
E
A
You
so
much
yeah,
so
these
are
the
first.
You
know
we
go
the
you
know
as
agenda
is
today.
You
know
we'll
start
with
the
working
group
documents
we'll
get
the
status
from
each
of
the
author's
on
the
work
unit
documents
and
then
followed
by
the
maintenance.
Maybe
16
v6
maintenance
documents,
and
then
there
are
a
couple
of
new
documents:
new
proposals
on
the
very
interesting
proposals
on
SR
we
say
forces
and
a
few
other
things
we
will
review
them
so
I
think
to
get
started.
First,
one
on
the
agenda.
G
G
G
Right
so
I'm
here
to
talk
about
the
FPC
progress
of
the
face.
So,
as
you
guys
have
noticed
our
light,
last
version
was
version
7,
which
was
actually
prior
to
the
last
meeting,
and
we
haven't
issued
an
update.
Since
then,
however,
we've
been
working
weekly
since
Chicago,
and
then
we
decided
to
actually
increase
the
participation
to
two
times
a
week.
I
think
in
June,
if
I
recall
and
what
you
know.
If
you
recall
FPC,
we've
got
a
couple
of
things
going
on
in
terms
of
the
40
policy.
So
we
have.
G
You
know
three
different
trees
and
in
the
aspects-
and
we
decided
to
for
this
next
go-around
in
our
update
to
focus
on
two
of
the
more
neglected
aspects
of
the
model,
not
to
mean
that
we
still
don't
pay
attention
to
mobility,
but
we
wanted
to
focus
on
topology
and
policy.
So
as
far
as
work
items,
we
kind
of
go
around
robbing
and
we
try
to
work
on
each
of
the
three
subjects
and
then
run
through.
G
So
we've
been
working
diligently
on
topology
and
we've
finished
up
the
majority
of
it
in
terms
of
notes,
we'll
be
doing
a
write-up.
For
that
in
the
next
you
know
a
month
or
two
and
then
we've
did
quite
a
bit
of
work
on-site
in
Berlin
or
excuse
me
in
Chicago
and
it's
all
a
blur.
Yes,
it's
a
live,
look
in
in
Chicago
and
we'll
be
rolling.
It
updates
to
the
policy.
So
hopefully
we've
got
that
slides
up.
G
G
There
we
go
with
the
FPC
update.
Verizon
has
an
update,
it's
only
the
provisioning
model,
that's
actually
the
EM
cord
release
and
that
was
demoed
at
Mobile,
World,
Congress,
so
yeah.
So
a
couple
other
items
no
bugs
have
been
filed.
We
do
have
a
you
know.
Some
work,
also
going
on
with
internships
at
a
couple
of
different
companies.
G
They'll
talk
about
some
of
that
work
as
well,
so
there's
active
coding
on
this
with
two
separate
projects,
both
of
them
open
source,
one,
oh,
no,
space,
tandem,
cord,
one
of
them
in
open
daylight
and
then
I
get
to
be
one
of
those
few
that
get
the
torturer
interns
with
a
lot
of
this
stuff
as
well.
So
next
slide.
G
So
we
have
already
talked
about
this.
Both
of
them
are
based
off
the
version
four.
So
just
as
a
reminder,
we
did
five
six
and
seven
and
very
rapid
succession.
They
had
to
deal
with
actually
errors
and
updates
prior
to
Chicago's.
So
there's
not
a
lot
other
than
model
Corrections
in
a
lot
of
those
updates
post
version
four
next
slide,
so
this
is
when
I
was
talking
about
with
the
transition.
So
we
do
have
the
three
we
think
of
everything
in
terms
of
a
tree,
but
really
three
areas.
G
So
mobility
policy,
its
apology,
we
are
multi
tenant
under
each
tenant,
you'll
see
each
of
those
trees,
those
those
really
separate
pieces
of
information
and
we've
been
playing
the
round-robin
game.
We
spend
obviously
being
DMM.
We
spend
a
lot
of
our
time
on
mobility.
We're
now
focused
on
finishing
up
topology,
go
back
to
to
policy
and
then
finish
back
again.
So
a
lot
of
these
updates
are
in
here
we'll
talk
through
the
updates
today,
but
keep
in
mind.
There's
no
spec
out
we'll
be
sending
that
out
as
quickly
as
we
can
so
next
slide.
G
So
just
as
a
reminder,
this
is
the
the
topology
slide.
So
we
did
change
a
couple
of
terms.
So
we
had
data
play
node
and
there
was
a
lot
of
confusion
as
the
level
of
abstraction.
So
we
started
introducing
the
word
abstract
to
be
a
little
bit
stronger.
So
it's
a
little
bit
more
obvious.
Obviously,
an
ad
P
n
can
be
physical
or
virtual
or
abstract
pick
whatever
word
you
like.
G
It
can
be
composite
or
other
items,
but
we
we
did
this
because
we
want
to
remind
people
that
we're
talking
about
an
information
model
and
so
we're
really
thinking
logical.
You
can
interpret
that.
However,
you
like
this,
gives
you
an
idea,
though,
basically
of
how
we
talk
about
the
physical
realization
down
in
the
resource
plane
as
well
as
what's
happening
with
the
abstracted
layer
in
the
model
that
we're
operating
at
so
next
slide.
G
So
this
is
a
notation
we've
seen
it
before.
I
think
Sottero
first
introduced
it
so
just
notations
sake.
The
item
on
the
left
is
the
container
structure.
You
think
it
was
top-level
the
tree,
the
the
top-level
concept
and
then,
as
we
go
to
the
right,
we're
getting
into
more
narrow
items,
probably
the
biggest
change
that
we
had
was.
G
We
had
a
fairly
weak
typing
system,
and
this
doesn't
help
us
in
terms
of
trying
to
figure
out
which
data
plane
node
we
want
to
select,
and
so
in
Sao
Paulo
G
we've
strengthened
that
concept
with
DPN
type.
So,
under
the
topology
structure,
we
have
a
DP
a
list
of
DPN
types
and
each
type
actually
has
a
couple
of
things
so,
most
notably
the
access
technology
which
allows
us
to
understand
what
you
know
attribute
value
pairs.
What
what
type
of
terminology
they'll
be
using
the
role
or
function
a
lot
of
people?
G
Think
of
these
as
LMAs
or
mags,
and
then,
underneath
that
a
series
of
interfaces
that
are
tied
with
that
particular
technology
and
role
and
with
each
interface.
We
then
have
a
list
of
protocols
and
a
list
of
features
and
keep
in
mind.
A
lot
of
this
is
still
just
the
things
we
need
for
negotiation
to
try
and
figure
out
if
an
interface
is
suitable,
and
so,
in
the
middle
of
that
we
actually
discovered
a
couple
of
scenarios
where
the
features
just
weren't
good
enough.
G
G
So
when
it's
at
present
here,
it's
really
considered
a
non-negotiable
setting
as
with
respect
to
a
feature,
and
typically
just
for
example,
it's
typically
when
a
features
turn
on
or
off
and
can't
be
negotiated.
So
this
allows
us
then
to
look
at
a
particular
type
at
an
interface
during
the
selection
process,
whether
that's
on
the
FBC,
client
or
agents,
and
decide
whether
or
not
that
interface
is
suitable.
So
this
is
the
the
basis
of
the
VPN
type.
So
next
slide,
so
we
go
a
little
bit
more.
G
You
know
this
really
just
kind
of
highlights
into
that.
What's
going
on
with
this,
and
we
give
some
examples
with
the
technologies.
You'll
also
note
we
know:
what's
keys,
there's
also
display
names,
kind
of
the
administrative
stuff.
You
would
expect
for
a
human,
readable
and
type
of
information,
but
also
what
I'll
call
the
usual
suspects
what's
important
about
this
as
well
is
each
one
of
these
keys.
G
So
when
you
take
all
three
keys,
the
technology,
the
role,
the
interface
that
compound
key,
allows
you
to
build
a
reference
back
to
a
particular
type
and
that's
what
we'll
talk
about
next
slide.
So
when
we
talk
about
this,
we
actually
have
a
couple
of
other
structures
in
here.
So
we
have
a
DPN
group
list.
So,
as
you
can
tell
there's
other
dpns,
we
have
then-
and
basically
this
is
just
a
collection
of
dpns
with
a
reference
list.
We
also
then
have
dpns
in
the
structure.
G
So
these
are
that
pointers
that
I
was
talking
about
back
up
to
the
type.
So
we
should
be
able
to
look
at
a
DPN
or
a
DPN
group.
Look
at
the
type
and
figure
out
from
the
topology
information
model.
Whether
or
not
a
DPN
is
acceptable.
Now
how
you
then,
given
several
options
for
a
DPN
or
several
alternatives
for
a
DPN
that
is
suitable?
How
you
just
that?
That's
something
we
won't
go
into
in
the
spec.
G
G
So
as
a
part
of
that
right,
we
are
essentially
referencing
viable
dpns
in
this
role
and
you'll
see
that
in
the
structure
and
then,
once
again,
you
see
those
pointers
that
I
was
talking
about
back
to
the
interface
reference.
Also
right
here
and
we
talked
about
the
DPN
list.
This
actually
made
the
DPN
really
straightforward.
Once
we've
kind
of
got
to
this
design,
you
see
kind
of
this
rinse
repeat
architecture
where
basically
we're
going
back
to
interfaces,
references
and
the
pointers
back
to
the
list.
G
So
this
will
give
you
a
pretty
basic
setup
for
whether
you're
going
for
groups
with
a
purpose
or
you're,
actually
just
scanning
across
the
entire
DPN
list,
for
a
suitable
DPN
for
selection.
Next,
one
so
usage,
if
you
hadn't
heard
right,
it
is
about
selecting
it
and
interface
it
suitable
and
only
using
the
list
in
the
groups.
The
types
are
there
to
determine
compatibility
what's
important
in
the
spec.
Is
we
make
no
assumptions
about
who's?
Doing
this?
We
want
to
make
sure
the
information
models
available.
G
We
do
today,
for
example,
see
implementations
I
know
in
the
open
source
implementations
we
have.
We
actually
have
clients
that
go
ahead
and
select
the
data
plane
node
and
we're
experimenting
with
a
lot
of
our
other
work
where
the
agent
is
actually
doing
more
selections,
and
this
just
really
reflects
the
code.
We
built
it
very
top-down.
You
know
the
way
the
network
was
built,
you
know
today,
and
so
we
had
intelligent
clients
and
now
we're
moving
more
toward
agents
that
have
a
lot
more
autonomy.
G
So
the
biggest
piece
about
this
is
the
topology
tree
is
only
I,
wouldn't
say
only
useful,
but
it's
primarily
useful
during
the
DPN
selection
phase.
So
the
way
to
think
of
the
topology
tree
is
you
use
a
primarily
selected.
Epn
policy
is
more
about
applying
a
lot
of
the
treatments
that
we
want
and
then
obviously
the
mobility
tree
within
FPC
is
more
about
the
individual
mobile
session.
So
that
gives
you
an
easy
idea
of
how
to
break
apart
the
model
and
think
about
it
during
our
processing
stages.
G
So
one
other
note:
if
you
follow
a
Triple
DES,
which
is
35
98,
and
that
drives
a
lot
of
our
work
in
different
production
networks.
The
DPN
type
was
intentionally
set
up
so
that
it
produces
the
same
information
required
to
produce
simple,
named
pointers,
compliance
to
3598
next
slide,
so
we
actually
have
worked
through
VPN
peering,
and
what
that
is
right
is
a
series
of
configurations
within
the
groups
for
individual
relationships
at
the
interfaces
we
do
want
to
know.
We
support
multi-tenancy
I
know,
there's
been
a
lot
of
talk
about.
G
G
G
So
we
looked
at
two
things,
so
we
looked
at
pre
provisioning,
a
policy
so
we're
not
interested
in
mobility.
We
we
have
it
working.
We
know
a
lot
about
it.
In
fact,
I
think
we
were
kind
of
getting
tired
at
least
of
coding
with
it,
and
we
also
want
to
look
at
a
use
case
of
one
plus
one
availability,
composite
dpns,
and
so
we
had
two
interns
that
were
at
the
undergraduate
level
when
it
graduated
one
lonely
had
three
classes,
so
not
exactly
someone
you
would
expect
to
come
into
the
workforce
and
code.
G
Both
of
them
have
been
successful
in
their
implementations
and
so
I'm
quite
happy
with
the
results.
But
it
looks
like
just
conceptually
only
working
with
them
a
few
hours
a
week,
they're
able
to
look
at
FPC
as
far
as
the
specs
the
information
model
and
get
their
work
done
composite
VPNs,
just
as
a
notes
in
plus
k1
plus
one
configurations.
We
did
the
work
and
we
wrote
it
up
and
just
for
one
plus
one
we
had
about
seven.
We
have
a
professional
tech.
G
Writer
on
staff
is
about
17
pages
explaining
the
operational
concepts,
so
what
I
would
say
is
for
things
like
high
availability
and
this
sort
of
stuff,
maybe
in
DMM
what
we
discovered
was
very
easy
to
say
it
might
be
a
good
work
item.
Very
hard
to
explain
as
a
standard
and
would
create
really
long
documents.
So
our
general
theory
is
we're
more
than
happy
to
share
the
information
and
talk
about
it.
G
But
originally
we
thought
we'd
be
bringing
this
work
into
DMM
I,
don't
think
it's
worth
our
while,
because
it's
very
operationally
focused
and
it's
just
a
lot
of
information
to
process
and
parse.
So
so
we'll
actually
would
be
talking
about
that
may
be
sitting
stuff
out
to
the
mailing
list,
but
we
won't
be
submitting
any
sort
of
new
work
for
DMM
in
this
area
and
in
fact
we
think
the
n
plus
K
documentation
will
be
a
hundred
to
two
hundred
pages
just
operationally,
describing
it
so
next
slide.
G
So,
as
we
talked
about
in
order
of
appearance,
you'll
see
work
on
topology.
We
did
a
lot
of
work
on
policy.
We're
reviewing
that
again
this
week,
starting
again
today
and
we'll
be
back
to
then
updating
mobility
next
slide.
So
we
do
have
some
aspirational
goals.
I,
never
like
to
say
the
second
I
say
goals,
or
these
are
the
objectives,
or
this
is
the
day
we
always
fail
so
as
freshly.
G
We
would
like
to
actually
get
a
lot
of
this
updated
in
the
same
spec
prior
to
99
in
the
next
couple
of
months,
or
excuse
me
I'm,
sorry,
prior
to
100
in
the
next
couple
of
months
and
not
at
deadline,
we'd
like
to
get
it
out
to
the
point
that
people
can
review
prior
to
100,
and
this
is
really
dependent
on
you,
guys,
the
more
reviewers
we
have
prior
to
100,
the
more
likely
we'll
push.
You
know,
leadership
to
to
a
last
call
or
something
to
that
effect
in
100,
yes,
suresh
krisshnan.
G
H
One
of
the
things
I
want
to
say
is
like,
if
so
you
said,
deadlines
so
how
far
before
the
deadlines
are
we
talking
about
because
practically
speaking
right,
so
once
you
have
this
ready,
as
I
talked
to
you
before,
like
I,
want
to
get
the
somebody
from
ops
to
look
at
the
moderator
models
and
yeah
stuff
right.
So
as
a
practical
matter
right,
it's
like
three
weeks
before
the
idea,
if
I
go
to
them
with
a
review
request,
it's
not
going
to
happen
so
now,
I.
A
H
G
H
I
H
Whenever,
like
you
know,
like
you
and
the
author's
tell
me,
it's
ready
for
somebody
to
look
at
from
the
yang
perspective,
okay,
I'll
talk
to
the
off
Sadie's
like
I'll,
actually
dr.
Benoit
like
and
get
somebody
to
just
do
the
yang
stuff
right
yeah
and
we
might
get
another
obstacle
you
like
for
the
operational
stuff
after,
but
at
least
I'll
get
this
done
out
of
the
way,
because
this
is
something
like
where
this
is
the
interiors,
not
where
the
expertise
is.
In
this
probability,.
G
B
G
G
H
H
Group
mailing
list
and
lot
of
the
documents
from
DMM
like
have
this
issue-
that
there's
not
enough
stuff
happening
on
the
mailing
list,
like
a
lot
of
people
not
reviewing
and
I,
can
have
on
this
change
right,
because
there
is
like
a
lot
of
interest
in
the
room
like
these
people
here.
But
if
you
don't
review
documents,
then
right.
A
H
Everybody's
just
writing
documents.
It's
not
gonna
cut
it
because
type
documents
are
not
getting
the
review
they
deserve
because,
like
the
chats
are
gonna
read
it
because
they
have
to
like
I'm
gonna,
read
it
because
I
have
to
write,
but
other
people
needs
to
do
need
to
do
it
voluntarily
to
actually
read
stuff
right.
So
right.
A
G
Got
them
here
right,
so
we've
already
talked
about
that
a
couple
times
right.
So
this
is
really
this
week
getting
that
review
of
topology.
Seeing
how
far
it
goes.
We've
got
a
bunch
of
write-up
internally,
then
to
kind
of
proto
the
the
next
version
for
both
the
topology
and
policy,
and
then,
let's
see
what
the
impact
is
on.
That
will
then
allow
us
to
understand
if,
if
it's
a
large
impact
or
small
impact,
really
that
deadline,
how
far
ahead
of
October
first
do
we
deliver
and.
G
Then
push
out
for
reviews
and
everything
else,
but
I
expect
us
to
maintain
our
cadence,
which
is
there
will
always
be
two
calls
available.
Not
everybody
shows
up,
we
don't
always
have
corn,
but
there's
two
two
times
a
week
for
an
hour
where
we
could
discuss
this
along
with
the
actual.
You
know,
discussion
via
mail,
so
we
can
keep
this
moving.
Okay,.
A
Next
is
Charlie
proteins.
This
essentially
Des
Moines
Nile
MN
identified
extensions,
forty
to
eighty
three
extinctions,
so
this
just
to
be
an
update
on
this.
This
document
went
to
iesg
and
essentially
iesg
created
a
should
I
discuss
on
this
and
the
draft
has
been
sitting
and
but
but
no
progress
right.
So
I
requested
both
Charlie
as
the
lead
author
and
also
Suresh.
Krishna
is
a
D
to
really
you
know
see.
If
you
can
discuss
this
issue
here
and
see
what
is
the
word
is.
H
So
so
what
happened
is
this
document
went
through
the
is
G,
and
so
it
got
like
a
bunch
of
comments
and
they
are
all
centered
around
security
and
privacy
pretty
much
right.
So
two
major
items,
one
of
them
it
says
like,
are
these
things
encrypted
and
what
are
the
guidance
like
for
encrypting
them?
So
what
sensitive?
What's
not
that's
the
first
question
and
second
question
is
like
they
are
like,
like
a
whole
number
of
MIT
types
right
and
people
say
like
okay.
H
Are
these
really
needed
right
like
do
you
need
this
many
types
and
who
uses
them
and
I
didn't,
have
an
answer
to
the
second
one.
At
the
time
the
discusses
are
done
until
like
yesterday,
right
and
so
I
do
have
the
answers
to
take
back
to
the
IAC
regard
at
least
RFID
types,
because
it
was
not
clear
at
all
what
the
RFID
use
cases
for
the
immune
IDs,
but
halima
like
who
was
like
actually
proposing
it
back.
Then
again
we
didn't
have
a
mailing
list
trial
for
this,
so
that
made
it
difficult.
H
So
I
do
have
that
information,
so
I'm
gonna
post
it
on
a
mailing
list,
like
whatever
information
I
got.
So
it
shows
up
for
everybody
in
the
future.
Okay,
but
at
least
that
part
of
it
is
clear
for
me,
so
I
can
go
and
justify
it
with
good
conscience,
because
I
couldn't
before
okay,
so
for
the
security
stuff,
I
guess
Charlie's
gonna
make
a
proposal.
J
So
just
a
slight
bit
more
context
of
this
draft
was
sort
of
like
finished
now
a
year
ago
or
more
and
then
it
sorta
had
last
call
and
has
been
noted.
We
didn't
get
comments
and
so
then
I
think
that,
let's
rush
with
absolutely
no
defense
when
me
was
talking
with
other
folks
in
iesg
and
at
actually
getting
comments
on
any
document
and
already
noted
this
morning
is,
can
be
problematic.
J
And
when
we
got
the
comments
over
the
last
few
months,
we'll
finally
I
think
we've
gotten
pretty
much
possible
resolutions
for
the
for
the
issues
that
were
raised
and
that's
why
I'm
here
and
the
format
of
the
presentation
is
a
little
unusual,
because
I
just
copied
segments
of
the
mailing
list
stuff
right
on
the
slides
for
powerpoints
and
I
thought.
Well,
you
know
it's
pretty
efficient
and
if
I
can
capture
the
nugget
of
what
was
said,
then
that's
less
work.
I
have
to
do
otherwise.
So
next
slide,
please
so.
J
Just
briefly,
there's
a
lot
of
types
in
the
draft
they
were
all
put
in
because
of
requests
from
you
know
years
ago,
and
then
then
there
was
a
little
bit
of
digestion
that
occurred,
and
then
a
comment
was
made
that
we
really
had
to
put
fuller
descriptions
in
for
each
one,
two
types.
So
that
was
a
lot
of
work
and
and
then
we've
had
too
much
descriptions.
We
had
to
pull
some
back
out
and
had
to
be
not
normative.
J
You
know
this
is
sort
of
the
you
might
say
sharpening
process
for
the
draft,
so
I
thought
by
the
time
we
were
done.
It
was
pretty
sharp,
but
anyway
here's
here's.
What's
in
there
we
have
the
v6
address
and
actually
a
Destin
one
that
I'm
not
quite
sure
we
should
have
in
there
there's
mg
identifiers.
They
could
obviously
be
put
to
use
if
someone
wanted
to
create
protocols
for
those
devices
using
mobile,
IP,
sort
of
approaches,
I
Triple
E
address
types
for
Mac,
adroit,
MAC
address,
and
actually
this
turned
out
to
be
really
useful.
J
If
you
have
somebody
set
like
multi
interface
or
unnumbered
interfaces
sources
that
you
wanted
to
handle
somehow
with
the
mobility
management
from
DHCP,
the
DUID
types
and
then
I
guess
what
has
been
you
might
say.
The
major
discussion
recently
was
about
RFID
types,
and
maybe
the
reason
is
because
there's
so
many
of
them,
so
they
look
voluminous
in
the
draft.
But
so
here,
oh
I
forgot,
I,
thought
I'd
turn
these
into
black,
but
anyway
left
us.
So
this
this.
What's
we're
in
red
right
here?
J
Was
it
directly
added
to
this
slide
from
Okemos
discussion
that
we
were
having
back
and
forth.
So
she
points
out
that
these
RFID
types
who
were
specifically
put
in
there,
so
that
you
could
use
a
mobile
IP
for
tracking
tracking
work
and
would
be
useful
for
mobile
banking
payments,
as
obviously
requires
high
security
and
also
well.
H
J
Next
slide,
oh
also,
she
mentioned
it.
Maybe
would
be
a
good
idea
to
consider
yet
more
ID
type,
since
we're
now
diving
into
the
help
you
and
world
of
different
kinds
of
identifiers,
but
I
don't
propose
to
do
that
so
and
Suresh
is
assuring
me
that
I
should
not
propose
so
can
I,
please.
Okay,
so
Vijay
wanted
to
just
summarize
what
Ben
Campbell's
issue
was,
and
so
that's
what
he
had
to
say.
J
Ben
Campbell
wanted
us
to
explicitly
identify
which
of
the
IDS
have
to
be
encrypted
and
not
just
say
that
some
of
them
may
need
to
be
encrypted.
So
that's
pretty
nebulous.
Perhaps
we
could
say
depend
on
the
context,
which
is
true
well
had
more
to
say
about
that
in
a
minute.
Another
option
is
just
to
say
that
that
they
carry
and
Emin
ID
may
need
to
be
encrypted,
over-the-air
for
all
inman
IDs.
That
might
be
the
safest,
and
so
my
answer
that
is
I,
agree,
I'm,
okay,
with
doing
them
on
next
slide.
Please!
J
So
looking
a
VJ
was
discussing
of
Stephen
Farrell's
discuss,
he
seems
to
be
saying
that
the
privacy
concerns
and
need
to
be
explicitly
justified
and
Vijay
points
out
that
there
are
privacy
concerns
with
every
kind
of
mobile,
node
ID,
and
so
the
user
can
be
tracked
using
that
under
mini-concert
of
mobile
and
Emin
IDs,
and
even
the
ones
that
are
currently
standardized.
You
could
be
easily
tracked,
and
so
it
doesn't
too
much
matter
what
kind
of
a
man
ID
you
might
be
using.
J
It
still
could
be
useful
for
tracking
if
it's
really
a
closed
identifier,
and
this
might
be
related
to
something
going
on
in
this
ID
ID
EAS
discussion
this
afternoon.
But
again
that's
way
future,
and
this
is
present
so
so.
I
believe
that
that
observation
from
Vijay
also
justifies
that
we
should.
We
must
you
encrypt
the
contents,
because
we're
operating
at
the
network
layer-
and
we
can't
know
what
the
application
is,
which
means
that
we
can't
determine
enough
context
at
the
network
layer
or
we
shouldn't
be
able
to
unless
we
have
serious
labor
violation.
J
But
we
shouldn't
be
able
to
determine
whether
or
not
the
context
requires
encryption,
and
so,
if
you're
gonna
do
this,
you
should
we
I
think
we
should
specify
encryption
and
I
came
a
and
large
as
on
that
and
I
believe
it.
Her
comment
here
should
be
taken
in
the
context
of
well.
What,
if
we
don't
encrypt
well,
then
she's
proposing
another
possible
solution
whereby
you
could
actually
ensure
privacy
by
some
sort
of
operation.
J
So
then
there
was.
This
is
intended
to
be
addressing
the
point
about
well,
why
do
we
have
all
these
MN
IDs,
and
so
the
reason
why
we
have
them
is
because
people
asked
for
them
on,
but
people
didn't,
you
must
say,
continue
to
ask
for
them
not
after
month,
and
so
that
led
to
the
perception
that
they
really
didn't
want
him
that
bad
and
so
so,
then
we
had
this
recent
discussion
about
what,
which
was
that
we
really
want
and
how
to
justify
those.
J
But
I
also
would
like
to
point
out
that
this
is
not
a
new
protocol.
This
is
a
registry
and
the
downside
of
including
possibly
useful
identifier
types
in
a
registry.
It
is
not
catastrophic,
it
just
means,
maybe
people
won't
use
it
and
so
I'm,
given
just
as
an
example,
this
particular
draft
as
a
case
in
point
how
many
years
have
we
had
this
two
years?
Well,
if
somebody
else
wants
to
register
a
new
kind
of
type,
and
we
could
foresee
that
that
would
be
useful
at
this
point
in
time.
J
Well,
why
do
we
make
dim
go
through
another
year
and
a
half
cycle
for
a
new
draft?
It
has
difficulty
going
through
getting
getting
enough
comments
on
the
mailing
list.
So
my
point
here
is
it's
not
too
hard
to
figure
out
that
these
are
going
to
be
useful,
mobile,
node
in
Erfurt
types,
and
it
is
pretty
hard
to
get
the
drafts
approved
and
so
on
because
of
the
operational
and
and
the
procedures.
J
So
so
then
suresh
says
well
no
problem,
but
actually
really
what's
going
on
is
what's
in
these
parentheses,
assuming
proper
security
measures
are
taken
well,
true
enough,
and
so
that's
what
we
are
proposing
is
that
we
should
take
proper
security
measures
that
can
be
applied
at
the
network
layer
with
now
knowledge
of
context,
and
that
means
encrypt.
So
next
slide,
please,
okay,.
J
Right
so
this
is
y'all,
guess
you
would
say.
I
already
said
most
of
this
I'll
just
point
out,
so
this
is
an
argument
that
we
really
could
have
in
the
past
entertained
Additional
Protocol
of
designs
that
would
have
been
facilitated
by
the
availability
of
some
new
kinds
of
mobile
node,
IDs,
besides
nai
and
and
so
Suresh
says
the
issues
that
nobody's
coming
forward
to
say
that
we
want
them
well.
This
is
related
to
the
known
phenomenon
of
clientÃs
on
the
mailing
list
and
hakima
is
I.
J
Could
read
this
to
you
but
or
you
can
take
a
look
at
the
slides,
but
basically
it
has
to
do
with
the
fact
that
we're
gonna
be
dealing
with
mobility
and
Internet
of
Things,
and
the
point
of
the
Internet
is
to
consider
a
lot
of
kinds
of
communicating
devices
and
particularly
sensors
and
RFID
things,
and
so
on.
So
what
else
should
I
say
about
her
coming
in.
J
J
H
So
this
is
something
I
just
want
to
like
float,
okay,
so
to
the
working
group
like
think
about
this,
because
one
of
the
things
you
said
is
like:
oh,
it's
all
difficult
and
do
a
lot
of
work
in
the
future
and
so
on.
Right,
so
I
have
a
amelioration
for
that,
so
make
this
registry
change
it
to
expert
review.
Okay,
which
means,
as
somebody
doesn't
have
to
write
a
document
through
this
to
get
an
ID
and
I
can
pick.
H
A
H
H
We
can
do
that
thing
right
and
and
I
will
pick
it
up
with
Ayanna.
So
if
the
stuff
changes
like
I
will
let
Diana
know
because
I
need
to
put
in
the
designated
experts
right
like
in
there,
so
I
need
to
nominate
them
and
I
use
UNITA
proof
right.
So
that's
a
given,
but
it's
just
something
for
the
working
group
to
consider.
So
it
was
like
I.
Don't
want
like
bloat
in
this
like
for
potential
applications
right.
That's
the
idea
so
because
people
say
it's
difficult
to
get
new
idea.
H
A
H
Water
comes
in
the
future,
because
I
don't
want
to
throw
in
like
another
bunch
of
things
in
there,
like
you,
so
I
just
write.
It's
a
good
idea
right,
but
that
means
like
I'm
gonna,
have
to
take
the
draft
through
everything
again
right,
like
voting
plus
call
IETF
last
call
balloting
and
everything
right.
So
if
you
add
new
types
in
here,
but
if
you
change
the
IANA
policy
is
like
a
policy
change
for
me
that
I
need
to
take
through
the
iesg.
H
J
A
J
J
Right
so
I
maybe
already
said
all
of
this
sort
of
thing.
What
else
should
I
say?
Think
yeah
I
think
I
already
said
all
these
things
next
slide.
Please
and
I
wanted
to
listen
to
show
how
cameras
work
that
she
is
doing
and
searching
for
funding
and
getting
some
programming
done
and
so
on.
But
it's
some
cool
work
and
you
know
you
might
take
a
look.
So
if
you
have,
you
know,
get
the
slides
and
take
a
look
at
these
URLs.
You
can
see
what's
going
on
with
this
track.
J
J
A
A
J
K
K
We
had
version
free
last
time
and
number
of
people
have
agreed
that
who
we
are
to
review
the
document.
So
we
had
received
a
review
at
different
times
and
we
try
to
update
that
as
soon
as
we
received,
we
will
shut
out
shortly
after
review,
so
the
sort
of
so
it
ghost
from
version
3
to
version
4
after
we
will
and
two
additional
vehicles
at
version,
5
and
then
another
vehicle
to
version
6
and
and
Marcos
we
will
is
coming.
K
A
A
K
A
K
There
are
lots
already
adding
changes.
Well,
I'm,
not
going
to
mention
and
parade
I
comment
about
the
network's
lies
neither
mention
what
at
least
or
deleted,
so
the
charge
was
to
at
the
reference
at
that
time,
and
then
the
other
comment
was
that
it
was
a
very
long
document,
photography,
some
pages
for
informational
documents
or
trick
I
tried
to
condense
some
of
them,
so
it
shortened
a
little
piece
to
40
something
and
after
that
the
next
night
would
be
would
be
fellow's
comments,
and
this
time
he
deleted
the
neck
sizing
and
Heidi.
K
A
K
K
They
will
be
clear
later
when
I,
when
I
go
for
the
different
sections.
Okay
and
then
give
examples
who
are
doing
that
so
and
that
that
percentage
was
there
in
the
in
the
description,
but
it
was
not
spelled
earlier
so
I
move
it
early
and
move
some
of
the
text
to
make
that
better
scope
out
that
more
clear
and
then
let's
go
to
a
table
content
to
see
how
that
scope
in
the
next
slide.
So
in
section
3
we
already
enacted
hectic
job.
K
Did
you
find
the
contraband
anchored
in
a
panic
entrepreneur
and
so
meet
that?
Okay?
But
then
so
this
is
going
that
you
further
water,
the
function
that
are
in
terms
of
product
function
that
that
are
having
their
elected.
Those
functions
are
so
in
Section.
301
are
where
those
functions
are
in
which
egg
are
those
function
and
so
on
which
no
and
then,
and
then
what-
and
this
is
specific
on
distributor
anchoring
because
they
keep
on
here-
is
that
you
don't
want
to
be
a
centralized
angle.
You
want
to
be
able
to
move
anger.
K
So
if
I
had
a
teachable
anger,
what
are
the
function
that
you
need
over
there?
Now,
that
is
a
very
long
list
of
functions,
are
and
I
am
okay,
if
I
initially
that
you
don't
need
all
that.
Have
you
got
a
solution?
There's
many
different
solution.
You
can
do
tendering.
You
can
do
some
command
to
change
the
routing.
You
can
do
many
other
succession
in
so
there
was
extend
to
be
are
what
are
the
different
functions
that
you
can
put
over
there
and
it
turns
out
that
you
don't
need
out.
K
Then
it
depends
on
what
you
should.
This
user
may
need
some
of
them,
so
you
don't
get
a
plane
out
and
then
with
those
phones
and
other
operations
of
those
phones
and
then
what
that
a
parameter
is
going
to
do
it
pricks
that
into
the
later
you
have
a
location,
you're,
the
management,
okay
and
we've
that
are
that
work
that
that
description
of
the
angle
because
disappear.
K
Angling
is
new
in
the
sense
that
this
is
not
a
sense
as
anchoring,
but
you
still
need
a
lot
of
the
function
that
you
had
before
if
the
same
time
same
thing,
but
what
else
do
you
need?
So
if
the
description
of
those
functions
and
knock-
and
that
was
basically
the
description
and
now
how
did
that
description?
What
are
you
going
to
do
a
bit
then?
The
next
step
would
be
a
solution,
but
this
HIV
is
not
trying
to
propose
any
specific
solution,
because
you
can
go
to
other
job
to
look
at
the
different
solutions.
K
This
is
the
information
ajar,
so
in
so
sir,
it's
giving
an
example
of
how
to
put
this
user
together
without
going
into
other
detail
all
other
steps.
They
are
comment
that
this
is
a
long
document,
but
you,
but
for
people
using
a
you
really
don't
need
to
read
four
of
them.
It
give
it
give
a
very
rapidly
example
of
the
DIMM
function.
For
example,
inception,
for
you
start
with
internship
you're
anchoring.
K
First
of
all,
you
don't,
if
you
don't
need
to
change,
keep
the
origin
or
address
you
can
just
use
a
new
address,
and
that
is
an
acceptable
mo
for
the
mm,
so
that
description
was
insufficient
for
one
one.
But
if
you
do
need
to
change
the
IP
adjust
and
you're
presented,
and
if
you
do
need
to
change
it,
then
it's
in
section
five
is
describing
examples
that
try
to
have
different
way.
K
A
five
on
one
would
be
afraid
Network,
so
you
don't
have
to
anchor,
and
then
you
don't
know
and
then
type
on,
and
then
you
move
to
anchor.
And
then
you
introduce
the
habit
of
network.
Besides
angle,
you
have
to
know
down
there
and
then
you
switch
the
anchor
switch
to
know
and
then
find,
and
then
you
also
have
a
scenario
in
which
you
do
the
never
mobility,
so
they're
different
application
that
you
you
had
so
for
the
reader.
You
don't
have
to
read.
K
K
That's
why
the
the
document
seems
to
be
quite
long
in
that
sense,
because
it's
it's
more
like
a
guideline
on
to
put
a
solution
together:
okay,
so
that
they
can
fight
the
scope
of
the
document,
and
it
also
clarify
how
do
we
do
basically
don't
have
to
be
before
the
whole
thing,
depending
on
what
you
an
interest
on
what
you
need?
Okay
and
then
another
major
common
was
that
difficult
was
very
difficult
to
understand.
K
So
in,
on
the
left
hand,
side
is
the
figure
one
in
version
5
and
Sakura
University.
In
six,
on
the
left
hand,
side
just
put
down,
basically
is
trying
to
say
you
had
anger.
You
had
the
control
thing
that
they're
hanging
and
then
what
at
the
function
you
add
on
to
it
and
what
we
are
acting
out,
the
the
location,
the
location,
the
information
please
and
also
the
faulting
function.
K
These
two
functions
also
suffer
in
the
control
and
theta
plan.
They
were
putting
in
twins.
On
the
right
hand,
side
is
the
figure
one
to
make
it
more
clear.
So
we
draw
a
network
and
in
turn
they
go.
You
have
conjurer
parent
function
and
you
had
eleven
function.
So
there's
a
separation
contravention
and
it
also
shows
that
the
distribute
distributor,
develop
and
use
distributor.
K
So
that's
the
architecture
of
the
network
and
what
the
function
you
put
down
there
and
and
separate
city
you
had
the
location
function,
suing
location
function,
it
cause
there's
a
location
such
as
a
server
kind,
server.
Okay,
that
you
are
communicating
with
each
other
and
then
the
note.
The
mobile
node
at
the
bottom
is
just
a
mobile,
no
attach
the
network,
so
that
was
trying
to
explain
their
figure
to
be
more
intuitive.
K
Yes,
hello,
oh
the
next
one,
it's
so,
for
example,
in
the
next,
in
the
rest
of
the
fact
that
I'm
not
going
to
sort
of
sort
that
whole
thing
as
figure
one
in
the
in
that
manner.
But
just
put
them,
you
had
the
same
control
pan
and
in
the
fret
network
you
had
to
depend
angel.
So
you
more,
she
moved
the
control
and
help
them
to
move
okay
and
they
were,
and
there.
A
I
think
we
want
to
make
sure
first
thing:
L,
don't
really
shoes
are
okay,
but
technical
issues.
First
thing
step
one:
we
want
to
make
sure
there
are
no
technical
issues
with
the
draft.
Okay,
we
don't
want
the
dissing
the
document
to
good
iesg
and
come
back.
That's
goal
one
right,
then
eight,
oh
really,
it
should
be
readable
right.
If
somebody
cannot
understand
it,
there
is
a
problem
right,
I
think
all
the
four
people
who
have
you
know
will
need
it.
A
A
A
A
M
You
okay,
so
in
face
to
face
in
Chicago
there
was
some
discussion
about
the
nature
of
blocking
said
sock.
Adopt
I
was
told
that
I'm
not
allowed
to
do
that.
It's
not
a
good
thing
and
I
was
requested
by
the
chairs
to
come
with
some
alternatives
for
a
resolution
of
that
problem.
I'll
describe
that
in
a
moment
and
another
small
additional.
A
M
So
I'm
just
saying
we
gave
people
in
the
list
a
chance
to
comment.
We
didn't
get
any
comments,
just
meditate
and
we
chose
one
of
them,
so
I'll
describe
all
three
and
then
I'll
describe
what
we
chose.
Thank
you
want
to
say
something:
yes,
was
it
on
the
list
or
privately
with
me?
I
think
it
was
privately
with
me
and
my
understanding
of
your
comment
was
not
you
didn't
select
one
of
the
three
you
just
said:
I,
don't
think
this
work
is
necessary.
A
M
Yeah
we
weren't
aware
of
that.
Okay,
so
the
first
alternative
was
practically
described
in
the
session
in
Chicago,
and
that
is,
you
cannot
block,
sets
set
sucked
up
so
add
another
function
that
does
the
block
that
was
alternative
number
one.
I
must
say
that.
Originally
we
tried
and
I'm
talking
about
Albers
intent.
We
tried
not
to
define
a
new
function
because
we
thought
that
would
be
too
difficult
for
people
to
accept
and
it
would
be
much
easier
to
say:
hey.
We
just
were
just
adding
another
field,
but
not
a
new
function.
M
But
after
that
discussion
we
got
to
the
conclusion
that
we
can't
avoid
that
we'll
have
to
add
another
function,
so
that
was
alternative
number
one.
And
then
we
came
back
home
and
said:
wait
a
minute.
Why
do
we
need
two
functions
if
we
are
blocking
and
before
adding
a
new
function?
Let's
just
not
change
set
soccer
options
but
add
a
new
function
that
requests
a
service
and
that
function
can
block
until
the
service
is
provided.
M
It
will
be
provided
by
providing
back
an
IP
address
and
the
application
can
then
bind
to
this
IP
address
and
that's
basically
is
the
solution.
The
second
alternative
and
the
third
third
alternative
is
quite
similar,
but
we
were
saying
why
request
the
application
to
bind
this?
This
new
function
can
implicitly
bind
the
socket
with
a
new
address,
and
then
we
save
a
call.
So
these
are
the
three
alternatives.
Next
slide,
please.
M
So
this
is
a
more
detailed
description
of
alternative.
One
call
set
socket
option.
The
call
will
trigger
may
trigger
some
action
with
a
network,
but
it
will
not
block
call
a
new
function.
We
called
it
wait
for
service
and
that
will
block
until
we
get
the
IP
address
that
we
were
expecting.
So
that's
next
slide.
That's
alternative
one!
That's
a
code
sequence!
If
someone
is
if
it
helps
people
to
understand
and
it
works,
but
as
I
said
before,
but
wait
a
minute:
why
do
we
need
such
socket
options?
M
We
can
call
wait
for
service
without
that.
So
the
next
alternative
next
slide
says:
okay,
we
will
define
a
new
function
set
session
continuity
in
which
the
application
will
request.
The
type
of
session
continuance
and
sets
a
socket
set
SC
will
return
with
an
IP
address
and
then
we
will
bind
to
that
IP
address.
So
can
we
see
in
the
next
slide?
This
is
the
code
C
code
sequence
we
open
socket.
We
call
sets
SC
when
we
get
the
address,
we
call
bind
and
then
connect
and
we're
all
set.
M
So
no
problem
with
set
socket
options.
Third
alternative
is
very
similar
to
the
next,
the
next,
the
next
code
sequence
like
next
one.
Next
exactly
so
it's
very
similar
to
the
first
one,
I
grayed
out
the
call
to
bind
because
set
se
will
implicitly
bind
to
that
selected
address.
So
these
are
the
three
alternatives
and
next
slide.
M
M
A
N
N
M
A
E
N
M
N
N
N
V6
we're
worried
about
okay
as
long
as
it's
as
long
as
that
function
is
documented
to
that
this
blocks
right,
yes
and
the
then
that
seems
fine
to
me
like
for
POSIX.
He
kind
of
people
you
might
want
to
have
a
like
an
async
version
or
like
okay,
call
me
back
with
a
completion
when
this
is
done
so
that
you
can
write
yeah.
M
A
M
N
Yeah
I
understand
yeah,
but
okay,
so
yeah.
If
it
makes
sense
right.
This
is
there's
a
documented
blocking
call
version
and
there's
there's.
Hopefully
we
can
request
that
people
implement
a
an
asynchronous
version
as
well.
That
makes
sense
how
one
gets
notification
in
various
operating
systems
is
a
different
entry.
Humanness
yeah,
it
seems
fine.
Acef
recalled
it's
documented
to
say
this.
This
can
block
for
an
unknown
amount
of
time
and
might
fail
when
it's
done.
H
N
M
O
O
M
O
See,
okay,
but
there
can
be
more
than
one
v6
address
with
these
properties,
so
who
makes
that
decision
set
a
C
or
C
will
return
with
one
IP
address
and
if
the
application
chooses
wants
to
make
a
different
choice
between
the
two,
for
example,
use
two
different
pvds
or
whatever
then
set
SC
has
to
take
those
parameters.
It
has
to
know
about
that.
O
I
mean
another
approach
would
be
to
say
just
set
the
set.
The
address
preferences
like
we
have
now
and
just
say,
I
want
to
address
of
this
forum
right.
So
what
you
want
is
you
want
to.
You
want
to
say
to
the
network.
A
give
me
give
me
a
dress
that
looks
like
this
and
you
know
get
a
callback
when
something
arrives.
Maybe
you
already
have
one
in
which
case
you
get
called
back
immediately.
O
Maybe
you
don't
right
and
then
you
say
well,
I
said
like
I
said
buying
this
usually
not
a
good
idea,
but
like
you
basically-
and
you
say,
look
you
know
set,
it
said
something
on
a
socket
that
says:
I
only
use
an
address
of
this
type
and
then
call
connect
and
then,
if
you
don't
have
an
address
of
that
type
or
whatever
you
can
get
e,
okay
yeah
turn
on
avail.
And
if
you
want
to
do
something
more
complicated,
you
can
always
call
bind
yourself
right.
A
B
E
Michael
Abramson,
so
I
was
going
to
ask
but
the
whole
API-
and
this
is
you
said
we
don't
yeah.
We
don't
dis
owners.
Api
is
here,
and
this
is
abstract.
That
is
not
perfectly
clear
from
the
text.
The
initial
like
I,
if
I
you
read
the
introduction
and
so
on,
Proctor
would
be
good
to
make
an
explicit
statement.
Just
basically
saying
what
you
just
said:
I,
don't.
M
H
M
M
E
G
P
G
Q
He's
camera-shy
I
can't
can't
Liang
I'm
Hachi,
sorry
I,
haven't
read
your
sauce
say
that
ahead
of
time,
so
I
shouldn't
ask
a
question.
Probably
so.
I
know
that
this
says
that
you
should
get
an
IP
address
back.
How
do
you
deal
with
the
ipv6
privacy
aspect?
So
if
you
call
again
technically
you
get
the
same
IP
address.
Is
that
correct
or
if
you
want
I'd,
be
privacy?
Do
you
do
you
get
a
new
one?
So
that
way
you
know
for
this.
M
Is
a
socket
API,
so
it
it's
about
the
interaction
between
the
application
and
the
tcp/ip
stack
in
the
host.
So
nothing
goes
out.
In
that
sense,
privacy
is
important
for
I
mean
this
may
trigger
some
communication
like
DHCP
or
our
SRA,
and
if
we
need
to
deal
with
privacy,
we
need
to
be
able
to
deal
with
privacy
over
there.
But
here
there
are
no
privacy
issues
so.
Q
A
H
H
M
H
Home
I
understand,
but
it's
a
it's
an
API
extension
and
it
actually
talks
about
privacy
addresses
as
well.
So
so,
as
I
said
right
like
it's,
multiple
access
and
all
the
cross
points
don't
make
sense
right.
So
you
cannot
have
session
continuity,
so
you
cannot
use
like
a
privacy
address
as
a
home
address
it
kind
of
defeats
the
purpose
right
right.
H
M
R
M
In
the
original
versions
of
this
draft
we
had
three
types
of
accession
continuity,
fixed
session,
lasting
and
our
persistence,
and
that
was
quite
how
would
I
say
similar
to
the
work
that
is
being
done
in
3gpp
3gpp
had
another
type
of
session
continuity,
SSE
mode
3,
which
what
it
does
is
it
tells
the
net
I
mean
the
network
tells
the
the
host
you
can
receive
this
IP
address
on
the
event
of
mobility.
You
can
still
use
this
address,
but
a
limited
time
frame
and
then
you'll
have
to
move
to
a
new
address.
M
So
we
are
not
guaranteeing
session
continuity
forever.
We're
guaranteeing
that
for
a
certain
limited
amount
of
time
and
then-
and
that
gives
you
a
graceful
period
to
move
to
the
new
IP
address
and
I
wasn't
sure
whether
I
want
to
introduce
that
here
or
not,
but
eventually
I
found
a
good
use
case
for
that.
So
I
said:
okay,
let's
add
this
to
and
I'm
not
strong
about
that.
If
people
object
I'll
take
it
out,
I
think
it's
important
to
know
that
it
will
exist
in
3gpp.
So
that's
why
I
thought
well.
M
O
F
H
Description
on
so
I
think
what
Lorenzo
means
is
like
I,
think
you'll.
Correct
me
if
I'm
wrong
is
that
you
can
already
signal
this
in
RA
right.
So
what
you
do
is
you
you
set
the
valid.
Let
come
to
zero,
okay,
your
preferred
life
down
to
something
which
means
that
your
stack
will
not
select
the
address
for
any
outgoing
connections,
but
will
keep
all
the
connections
that
are
on
already
on.
So
what
are
you
saying
is
this
is
not
required.
Am
I
right
Lorenzo
yeah,
so.
M
So
so
here's
the
confusion,
I
understand
that
the
confusion
here
is
it's
actually
with
these
api's
the
application
requests
for
this
type
of
service
and
for
a
long
time,
I
didn't
see
why
an
application
would
want
to
request
for
such
a
serve,
and
this
is
why
I
didn't
add
it
to
to
the
api's.
If
what
you
are
describing
is
not
a
specific
request
from
the
application
you're
saying
hey,
the
network
supports
that.
M
H
M
H
Understand
so,
okay,
the
only
kind
of
thing
you
recommend
in
v6
is
graceful
replacement.
So,
like
no
flattery
numbering
right,
so
we
always
like
to
placate
the
addresses
because
of
v6
like
we
can
have
multiple
addresses
on
the
same
interface.
So
we
always
like
gracefully
duplicate.
So
we
put
the
previous
one
into
non-preferred
with
but
valid
right,
I'm.
H
So
there's
nothing
in
the
cellular
profile
right,
like
the
7059,
whatever
right
like
that,
like
you
know,
bunch
of
us
wrote
together,
which
says
this
part
of
VC
exists
is
overridden
over
there
right.
So
there's
like
some
things,
we
say
like
okay,
we
don't
do
like
you
know
NS,
for
example,
right
like
RR
s,
right
like
and
we
just
say
like
we
don't
do
it.
H
So
the
that
specification
says
like
where
we
are
deviating
from
the
IETF
shirts,
pretty
much
okay,
so
I'm
not
sure
like
what
would
be
the
network
equivalent
of
saying
I,
don't
want
graceful
replacement
like
I
I,
don't
see
why
somebody,
you
would
say
I,
don't
want
graceful
replacement
right
like
so
that's
the
use
case.
That's
not
clear
to
me.
So
why
would
an
application
come
and
say:
I,
don't
want
graceful
replacement,
and
why
would
the
network
honor
that,
like?
What
is
the
benefit?
That's
the
part.
Okay,.
M
A
B
R
When
I
recall
and
now
it
is
describing
that
a
team
and
terminology
and
functional
elements
in
the
other
and
and
also
there
are
some
mappings
with
Jamie
a
function
to
the
the
specific
service
primitives
that
how
demian
functions
is
also
stated
with
which
service
primitives,
like
mobility,
anchoring
and
of
example,
like
a
message
classifier.
So
this
current
version
hazard
as
addressed
all
the
open
issues,
and
now
we
think
that
now
in
a
good
shape,
and
but
we
couldn't
get
some
much
feedback
from
the
working
group
and
on
the
list.
R
So
another
one
important
point
is
that,
as
you
know,
that
here
is
that
we
are
defined
that
the
four
major
a
team,
a
function
so
CPAC
PN,
TPA
TPN.
So
based
on
this,
that
we
try
to.
You
know
specify
and
present
it
at
how
it
can
be
deployed
in
some.
You
know,
deviant
function
can
be
combined
together
or
separated,
depending
on
that
deployment
model
option,
flavor
so
important
things
that
we
need
to
ensure
that
there's
any.
There
is
a
consistency
between
that.
R
This
draft
and
with
other
working
group
draft
and
so
I
hope
that
some
also
of
each
working
group
documents
also
is
kindly
you
know
that
check
and
is
their
definitions
and
thumbs
are
in
the
world
consistency
and
they
were
distracted.
So
now
we
are
suggest
moving
this
to
ADA,
leave
you
and
we
are
hoping
to
go
and
buy
the
next
a
single
meeting.
Okay
thanks.
It
is
update
of
the
dis,
diplomat,
motor
working
group
document
and
any
comments,
and
the
feedback
will
be
welcomes.
A
A
B
R
B
A
S
This
is
a
thermal
suma,
I'm
I'm,
going
to
present
some
discussion
of
the
applicability.
It's
a
basic
I'm,
not
sure
how
many
people
are
aware
of
the
water,
it's
our
basic
ease
but
I'm
talking
about
the
basics
of
it
and
then
how
to
apply
a
ps4
basic
to
the
mobile
user
friend
okay.
So
this
will
show
the
just
one
example
of
the
current
mobile
network
deployment.
S
I,
don't
talk
much
about
this
its
itself,
but
a
civil
pain
point
there,
such
as
the
the
network,
is
well
fragmented
to
several
parts
of
the
road
and
the
session
from
mobile
user
or
mobile
terminal
bihaan.
During
for
the
session
basis,
that's
really
hectic
operation
require
in
the
each
of
the
mobile
node
and
we
adopt
the
toning
solution
between
the
node
of
the
mobile
data
frame.
So
it's
not
so
much
Optima
optimized
for
in
terms
of
the
data
path.
S
S
So
if
we
could
see
a
service
had
our
extension
header
to
the
pocket,
the
ipv6
network
can
integrate
all
of
the
network
not
only
for
mobile
use,
the
frame,
but
also
the
list
of
the
part
of
network,
the
transport
metropolitan
area
core
network,
not
mobile
core,
but
also
the
backbone.
Sir
layer
of
the
network,
can
be
integrated
both
of
the
moa
but
with
the
mobile
user
brunette
work
and
if
the
service
ID,
which
is
called
segment
ID,
represent
data
prime
role
and
function
its
enable
us
to
deploy
anywhere
within
the
service
network.
S
S
So
that's
the
benefit
of
the
SR
basic.
When
we
achieve
D,
that's
idea
into
the
our
use
opera
Network
next
slide,
please
so
I
talked
a
bit
about
us,
our
basics.
In
nutshell,
it's
a
basically
utilize,
the
one
of
the
extension
header
of
the
routing
I,
think
the
peat
person
who
sits
in
the
room
really
aware
with
the
routing
hitter,
because
the
nothing
types
to
hitter
is
really
favored.
S
The
average
use
it's
not
a
Shariah
sign,
but
using
type
full
as
the
ID
of
the
hitter
routing
header
type,
and
this
header
can
include
the
multiple
segment
ID
which
cone,
which
is
the
128-bits
face.
So
the
in
terms
of
the
routing
data
type
tool
are
the
two
using
just
one
segment,
but
it's
our
age
can
include
much
roof,
not
only
one.
It's
a
segment
ID
to
indicate
the
older
of
the
segment
so
too
that
helps
us
to
implement
the
sorting
stuff.
S
S
S
So
in
this
case
we
have
segment
B
and
C
and
D
D
is
actually
a
final
destination,
but
the
the
first
segment
is
indicated
by
a
segment
left
as
a
pointer
and
then
the
exit
point
node
rewrites,
the
decision
header
followed
by
the
Signet
ID,
to
pointed
by
a
little
segment
drift,
and
then
the
node
B
nor
B
receive
the
packet
and
then
the
decrements
a
segment
draft
and
then
the
rewrites
a
destination
address
by
the
is
ID.
That
is
a
second
addressed
point
to
the
segment.
S
S
So
this
is
a
little
bit
busy
side,
but
this
is
kind
of
the
extension
definition
extend
the
definition
which
the
ESL
v6
trying
to
standardize
the
a
60
original
source
routing.
It's
just
kind
of
the
sorting
point
to
the
the
meter
past
midpoint
of
the
path,
but
it's
our
basics
network
programming,
I'm
trying
to
introduce
the
some
semantics
of
this
function
for
the
segment
ID.
So
we
pick
up
the
some
segment.
S
S
To
go
through
the
mobile
node
to
also
outside
network
in
upland
and
then
go
down
to
the
mobile,
no
in
down
link
side.
So
in
uplink
side,
access,
node
put
this
our
header
in
to
the
customer
pocket.
Just
in
case,
you
in
capsaicin
is
available.
How
it's
a
little
bit
controversial
agenda
but
for
make
explanation,
is
I
use
the
insertion
so
once
the
segmented,
a
segment
wrist
input
it
to
the
custom
pocket
the
layer
to
incarnate
its
I,
define
this.
S
S
You
so
that's
the
the
this
notation
means
and
in
down
link
layer,
Sri
Lanka
known,
for
example,
a
P
gateway,
push
the
SL
header
to
the
pocket
and
in
the
middle
layers
on
kernel,
such
as
like
escalatory
process,
T
segment
and
then
the
for
to
the
pocket
to
the
access
point
which
is
ID
indicated
and
then
access
node
do
the
process
of
the
kind
of
the
cross
connecting
to
the
right
phrase
of
the
frickin
spectrum
and
time.
So
that's
how
the
six
men
hunting
works
for
the
mobile
user
train.
S
S
S
S
S
S
Thank
you
so,
okay,
this
is
the
one
of
the
network
sizing
consideration
with
SL
v6,
so
other
I
mentioned
before
a
set
of
the
segment
ID
can
represent
the
network's
rise.
So
the
set
of
D
is
a
network
resource
can
be
aggregated
by
s
ID,
which
is
very
much
same
with
the
ipv6
numbering,
so
the
ones
we
use
the
same
prefix
for
those
s
ID
within
the
size.
That
would
be
really
handy
to
indicate
the
s
rise,
specific
size.
S
So
then
Italy
I'm
not
sure
it's
impossible
or
not,
but
it
will
be
one
of
the
idea.
Then
the
the
application
in
the
mobile
node
can
direct
three
point:
the
each
with
a
thrice
by
using
s
ID
in
the
air
pocket.
So
that's
the
another
possibility
to
handle
the
network
are
select
the
network
sizing
from
the
dominant
side,
even
Tamara,
but
also
the
application
in
the
mobile
node.
S
A
S
Okay-
okay,
okay,
next,
sorry,
okay,
so
this
is
next
step.
This
is
a
very
premature
war
right
now,
but
I
want
to
study,
wrote
more
to
cover
the
more
user
brain
function
to
be
supported
by
us,
our
basics.
So
after
the
least
a
study,
we
update
the
ID
with
you
outcome
with
a
study
and
also
the
figure
out.
Yes,
the
set
of
the
sort
of
basic
function
form
I
will
use
a
friend
which
could
be
standard,
so
the
spring
working
group
trying
to
standardize
a
service
function
but
see
if
it
will
be
a
suitable.
S
S
S
S
I
Clarifying
question
I
think
your
address
it
in
the
slice,
but
it
in
right.
So
in
today's
network
it
runs
as
an
overlay
of
mobility,
bankers
and
chain.
So
taking
these
anchors,
do
you
think
about
using
segment
rods
to
define
the
hops
between
these
anchors
when
you
consider
having
way
more
hops,
given
in
my
new
segment
rod,
I
find
more
fine,
granular
steering
and
customized
steering
yeah.
S
I
Optimize,
the
path
or
other
network
function.
How
many
hops
do
you
think
about
specifying
is
a
tool
like
today
or
way
market
depends
on
the
IETF
they
operate
at
Porsche
and,
if
you
take
it
as
replacement
of
today's
channeling,
there
is
some
information
about
us
and
whatever
yeah,
it's
just
patient
segment
routes
to
in
the
in
the
routing
header
to
incorporate
such
information,
like
you
see,
is
or
yeah
I
think.
S
The
mention
the
my
idea
in
this
sorry
about
I
wrote
a
draft
to
put
some
semantics
of
the
QoS
or
data
limiting
stuff
that
can
be
combined
within
the
function
segments
which
combine
using
the
one
128
bit
space.
That's
not
really
handy
to
put
many
semantics
in
that
Raj
bit
space,
so
I
mean
so
if
we
define
the
segment
ID
Mart
prime
much
proof,
but
that
can
be.
You
know,
push
into
this
one
over
the
128
bits
that
come
using
masking
something
I'm.
B
A
Right,
thank
you.
Sorry,
son
I
think
a
couple
of
things
is
a
very
good.
You
know
document
I
think
you
know
at
least
a
good
work
item.
Currently
it's
not
in
the
Charter,
but
you
know
fairly
it's
a
new
work
and
you
know
no
discussions
and
will
let
others
you
know,
do
some
more
work.
Also
get
some
feedback
from
the
working
group
will
discuss
with
the
ad.
If
there's
sufficient
interest
to
take
this
up.
Definitely.
A
Happened
if
initely,
it's
very
good
good
work
item
that
hopefully
will
take
it
up.
Yeah.
Thank
you.
So
much
so
next
is
your
Cutlass
network
based
on
client
base
diem
solutions
using
mobile
IP
mechanisms
so
godless.
Unfortunately,
you
had
to
cut
down
your
time
little
bit.
How
do
we,
okay
with
that?
Okay.
T
So
this
is
a
presentation
about
some
graphs
that
have
been
there
for
a
while,
as
you
will
see,
we
have
implementations
back
in
two
thousand
thirteen
or
fourteen.
So
next
slide,
please
so
I'm
gonna
just
discuss
about
AI
and
Molly
PVA
solution
that
we
specify
some
years
ago
and
at
the
same
for
network
base,
DMM
solution
and
an
additional
item
for
network
based
solution,
which
is
what
we
call
the
list
with
that
vertical
interface.
So
next
slide,
please!
T
So
what
motivation
I
think
we
have
been
working
in
the
in
this
working
group
for
a
while
on
on
the
four
main
charter
items,
but
I
still
think
that
it
may
be
good
to
discuss
on
on
solutions
based
on
well
understood
protocols
like
Molly,
P
and
proximally
P.
There
have
been
a
lot
of
proposals
out
there.
T
Those
that
I'm
presenting
here
are
the
ones
that
we
work
on,
but
there
are
similar
alternatives
and
I
think
that
may
be
now
good
time
to
see
whether
there
is
interest
in
in
specifying
this
kind
of
mobile
IP
proximally
P
based
solutions,
40mm
s1,
please
so
those
are
just
pointers
for
the
drafts
and
some
academic
papers.
We
have
we
work
on
that
quite
extensively
in
the
last
few
years,
so
there
are
plenty
of
kind
of
papers
with
more
details
and
that
perform
an
analysis
discussion
on
deployment.
T
These
kind
of
things
I
will
skip
just
for
the
sake
of
time
in
next
one,
please.
So,
let's
go
into
the
brief
summary
of
the
client.
First,
a
client
based
EMM
solution.
So
it's
basically
an
extension
of
Molly
P,
with
some
very
simple
considerations
for
how
to
make
that
more
deployable
or
easy
to
deploy.
So,
basically,
what
we
do
is
to
propose
moving
the
home,
the
home
alien
to
the
edge.
So
basically,
we
are
co-located
in
the
functionality
of
the
home,
a
lean
on
basically
the
access
router,
as
we
will
see
later
next
one.
T
Please
then
we
call
that
distributed
access
router,
but
terminology.
We
have
been
playing
around
with
that,
so
you
can
call
this
root.
Access.
Router
is
routed
gateway,
it's
basically
the
same
thing
is
the
access
router
with
uncommon
functionality,
so
Anthony
was
discussing
previously
so
that
access
router,
that
home
alien
will
assign
topologically
valid
IP
prefixes
to
the
attached
in
mobile
nodes.
T
So
on
attaching
mobile
node
gets
an
IP
and
IP
prefix
configuration
IP
address
that
can
be
used
for
the
time
the
guy
is
attached
there
and
then,
if
the
guy
damn,
why
not
moves
another
access
router
to
another
distributed
access
router?
That
guy
will
obtain
a
new
address,
the
politically
valid
on
that
access
router,
but
you
will
be
able
to
still
use
the
IP
address
that
it
obtained
while
visiting
the
former
the
previous
access
router
so
and
in
that
case,
that
access
router
will
play
the
role
or
will
function
as
a
home
agent.
T
This
next
one
you'll
see
that
in
a
very
simple
diagram,
so
first
attachment
know
a
note
gets
an
IP
address
here.
You
can
call
that
care
of
address,
but
at
this
point
in
time
is
the
APR
is
the
only
one
that
the
guy
has
and
is
topologically
ballad
here
so
can
be
used
to
connect
to
any
node
in
the
internet.
It
again
moves
to
another
access
router.
T
You
will
obtain
a
new
address
here
that
can
be
used
for
new
communications
directly
with
any
node
without
any
tunneling,
but
that
can
be
used
also
as
a
care
of
address
to
still
use.
The
aggregate
was
configure
here
so
for
sessions
that
are
going
while
the
guy
moved.
This
will
be
using
the
tunnel
with
the
previous
access
router
to
communicate
with
any
any
system
note
that
it
has
a
communication
ongoing
and
for
new
communications.
You
can
use
the
address
configure
on
the
new
access
point
of
attachment
next
slide.
Please
over
here.
T
T
Yes,
but
for
a
minute
sorry,
if
we
move
again,
we
can
configure
a
new
address
and,
on
the
same
thing,
can
combo
and
of
course,
as
long
as
the
sessions
stop
being
a
lie
for
all
unsigned
addresses,
those
can
be
just
remove
and
stop
being
used
next,
one
please
next,
one
okay,
so
that
was
a
very
brief
summary
of
the
client
base,
so
very
simple
extension
to
more
ipv6.
Then
we
also
did
the
same
for
network
based
mobility
management.
T
So
we
understand
proximal
ipv6
again
we
push
the
uncle
into
the
to
DX
to
the
edge
at
the
access
water
level,
and
we
come
here.
We
came
up
here
with
a
partially
distribution.
The
sense
that
we
have
a
data
plane
is
routed,
but
the
control
plane
centralized
in
in
a
kind
of
control,
only
LM
a
role
which
is
what
we
call
a
CMV,
as
we
will
see
in
the
in
the
next
slide.
Next
one,
please
so
again,
you
can
forget
about
terminology,
because
we
are
using
different
terminology,
because
that
was
a
different
draft.
T
Again,
we
have
a
distributed
gateway
access,
router,
that
is
the
access
router,
where
the
mobile
not
attached
to,
and
this
guy
in
this
case
concentrates
the
exit,
router
functionality,
Dell
ma
and
the
Mac
functionality,
because
again
it
will
be
delegating
prefixes
to
the
mobile
node.
That
will
be
optimal
because
are
topologically
valid
there.
But
then,
if
a
guy
moves,
the
guy
will
obtain
new
addresses
at
that
location
and
we'll
be
able
still
to
use
the
previously
allocated
addresses.
T
So
in
that
case,
those
as
a
router
will
be
playing
the
role
of
Mac
and
LMA,
depending
on
the
on
the
different
addresses.
We
have
some
additional
terminology:
we're
not
go
into
the
details,
because
I
think
is
easier
to
just
see
in
an
example
without
going
into
much
into
all
the
terminology,
and
the
only
thing
I
wanted
to
highlight
here
that
we
have
this
central
mobility
database,
which
is
this
control
only
LMA,
that
stores
the
different
allocations
of
the
prefixes
and
the
location
of
the
moa
nodes.
Next
one
please.
T
So
here
is
an
example:
we
have
these
two
two
guys
we
have
the
database
control
plane
here.
So
if
I
know
that
touches
to
one
of
these
access
routers,
it
will
obtain
a
prefix
topologically
valid
here
that
can
be
used
for
communication
with
the
internet
without
any
tunneling.
At
the
same
time,
the
guy
here
will
communicate
that
mapping
or
that
allocation
of
traffic's
with
the
mobile
node
to
this
central
mobility
database.
That
will
know
then
the
mapping
of
the
address
of
the
mobile
node
and
the
location.
T
So
if
the
guy
moves
next
slide,
please,
basically
or
here
we
explained
that
there
are
different
ways
of
updating
this
database
different
operational
modes
conceptually
are
similar,
but
we
are
going
to
focus
for
a
sec
of
time,
also
because
it's
the
one
that
we
implemented
first,
what
we
call
the
proxy
the
proxy
mode.
So
next
one
please
so
continuing
with
example.
That
I
think
is
easier.
T
T
You
know
that
there
is
a
prefix
being
used
on
this
guy,
so
you
have
to
establish
a
tunnel,
and
this
guy
will
signal
the
same
to
to
this
guy
so
that
it
can
be
established
from
the
point
of
view
of
the
data
communications
communication
already
established
will
go
through
this
tunnel
between
the
two
axis
raptor
new
communications
will
directly
go
without
any
tunnel
in
between
the
denote.
The
router
on
the
internet
so
basically,
is
applying
proximally
p
with
something
running
to
enable
that
action.
T
Please
and
then
another
piece
of
that
solution
that
is
independent,
can
be
used
or
not,
but
I
provide.
Some
advantages
is
what
we
call
the
technical
interface.
So
basically,
what
we
allow
is
to
configure
logical
interfaces
on
the
on
the
access
routers.
So
when
the
mobile
node
moves,
it
doesn't
see
any
change
because
it
will
always
be
connected
to
the
same
logical
interface
from
the
point
of
view
of
the
rotor.
So
if
a
guy
moves
from
one
gateway
router
to
another
one,
this
logical
interface
will
be
constructed
set
up
on
the
new
distributed
Calaway.
T
So
we
will
see
the
same
MAC
address
the
same
ip
address
as
in
mobile,
approximately
P,
but
with
additional
advantages,
because
we
can
have
different
logical
interfaces.
So
it's
like
the
guy
as
it
moves.
It
will
be
connected
two
new
routers
but
I
still
see
in
the
old
ones,
and
that
allows
to
play
with
a
few
specific
routes
and
these
kind
of
things
that
we've
just
playing
proximal
ipv6
to
you
cannot
do
it.
Cutlass.
A
T
F
L
T
So
there
are
more
details
on
the
distributed,
logical
interface
and
the
other,
slides
and
I
just
wanted
to
highlight
that
we
had
demos
already
in
ITF
in
2012
and
2013.
The
code
is
available
next,
one
on
it's
open
source,
so
you
can
download
the
code
for
the
decentralize
at
the
client
base,
the
dmn
there
sorry
the
network
base,
and
even
though
I
even
cover
here,
we
even
have
an
SDS
solution.
Also,
the
code
is
also
available
there.
T
T
S
S
Mc,
as
I
have
said,
the
embassy
can
be
used
to
reduce
the
engine,
the
latency
and
but
for
some
scenarios,
for
example
the
critical
or
latency
requirements
the
deployment
of
application
instance
as
the
agile
network,
maybe
not
such
a
not
enough
to
satisfy
this
critical
latency.
So
when
the
mobile
node
moves
away
in
the
in
the
network
and
to
guarantee
the
load
latency
requirements,
we
can
manage
the
application
instance
between
the
MSE
servers,
which
was
falling
on
mobile,
not
know
this
movement
and
the
ETS
s
specification.
S
S
And
here
is
what
the
admin
couldn't
could
do
for
a
missus
matter
location.
We
want
to
propose
a
network
based
solution
in
DMM
workgroup
to
satisfy
the
service
community
in
this
embassy
scenario
is
a
gap,
a
gap
for
the
current
IDI
memo
solution.
In
the
on-demand
mobility
scheme,
the
rotary
dancin
redundancy
can
be
reduced
at
a
certain
level,
but
it's
not
quite
enough
for
the
ambitious
Mathur
relocation
scenario
because
for
the
ungoliant
ongoin
session,
the
amount
of
scheme
couldn't
eliminate
routine
returns.
Redundancy
at
the
same
time
providing
service
continuity,
and
here
is
our
proposal.
S
Solution
and
honestly,
our
solution
or
other
concepts
of
the
identity
and
the
locator
separation.
Accept.
We
define
a
network
location,
independent
host
ID
to
identify
the
communication
session
under
the
ID
will
be
assigned
by
the
network
to
the
host
and
the
it
will
remains
unchanged
during
the
movement
of
the
mobile
node
and.
S
The
the
try
figure
will
be
route
still
rooted
on
the
IP
address,
a
base
down
which
is
based
on
the
network.
Location
information
here
is
an
example
when
the
mobile
node
I
am
mobile,
node
and
the
application
instance
both
have
their
own
ID,
which
is
ID
admin
and
IDM.
When
the
mobile
node
moves
to
the
new
location,
the
the
application
instance
will
be
migrated
to
the
new
location,
and
here
is
the
third
three
three
paths:
the
black
one.
The
black
line
is
the
path
before
the
way.
Emigration
and
the
blue
lines.
S
A
So
there's
a
new
ID
right.
There
won't
be
any
discussions
right.
It's
you
posted
the
document,
yeah,
okay,
so
for
hopefully
you
know
folks
can
now
what
are
here
in
provide
feedback.
Next
we
have
the
last
topic.
Actually
your
tiny
ways,
your
okay
I
think
you
just
rushed
up.
Carlos.
Sorry,
everybody
that
I
management,
yeah.
C
U
All
right
in
Chicago,
we
had
talked
about
having
router
advertisement
and
router
solicitation
in
support
of
3gpp
SS
emotes.
This
would
be
a
single
example
and
after
the
discussion
we
had
a
pretty
in-depth
discussion
with
Suresh
and
three,
and
we
came
to
some
enlightenment
about
router
advertisements,
so
we're
kind
of
continuing
along
this
line.
U
U
U
There
is
four
well
three
that
are
currently
described
in
Dani
added,
this
graceful
replacement
in
the
latest
version,
so
we're
looking
at
providing
session
lasting
IP
addresses,
fixed
IP,
addresses
non
persistent
and
then
this
graceful
replacement,
and
so
what
we're
looking
for
is
a
way
for
the
ability
of
a
mobile
hostess
like
the
appropriate
IP
prefix
from
an
RA
next
slide.
Please.
U
And
so,
in
order
to
add
on
demand
to
RA,
there
are
a
number
of
ways
it
exists.
We
started
off
with
thinking
about
modifying
RI
itself
to
create
a
vendor
specific
option
and
after
hearing
all
the
worst
war
stories
and
scars
from
Suresh,
we
decided.
This
is
probably
not
a
good
way
to
go,
and
then
we
start
looking
at
well.
What
are
the
actual
things
we
need
to
get
done
and
what
are
the
right
ways
to
do
this?
And
ultimately
we
came
to
the
decision
that
you
know
the
bottom.
U
Two
ways
are
kind
of
the
ways
we
can
handle.
This
one
is
creating
something
at
the
prefix
option
level,
and
so
you
know
the
prefix
option
gives
you
you
know
for
a
persist,
specific
prefix.
You
know
various
options
that
you
can
use
and
we
decided
that
it
might
be
useful
to
do
this
in
the
prefix
option
itself.
The
only
drawback
there
is
that
it
would
require
the
modification
of
reserved
fields.
So
if
we
create
a
prefix
option,
that's
kind
of
alongside
the
prefix
option,
it
feels
like
it's
at
the
wrong
level,
but
slightly
cleaner.
U
But
if
we
put
it
inside
the
prefix
option
itself,
then
we
need
to
modify
one
of
the
reserved
fields
to
specify
the
prefix
type.
So
this
is
kind
of
where
we
are
now,
and
you
know,
one
of
these
options
will
most
likely
need
to
be
considered
in
order
to
provide
the
different
SSE
mode
specifications
in
our
8
next
light.
U
E
E
I'm
worried:
okay,
yes,
no!
No,
there's!
No,
but
I,
don't
think
there!
It's
good!
You
know
we're
not
talking.
There
is
the
the
whole
orbit.
That
is
not
in
your
draft,
so
there
is
actually
an
RC,
that's
defining
one
of
those
that
is
reserved
in
yours,
okay,
so,
but
right
now,
I
just
think
that
there's,
if
we
take
our
our
proposal
annual
proposal,
no
there's
one
bit
left
so.
H
Krisshnan
you
so
like
this
bits
are
not
gone,
so
we
we
didn't,
define
an
extension
mechanism
for
the
flags
right,
and
so
we
can
do
like
a
extension
mechanism,
for
this
is
the
last
bit
right.
If
we
are
worried
about
it
and
I,
have
you
been
kind
of
inconsistent
with
like
how
we
use
updates
in
the
idea,
so
it's
not
specific
to
like
int
area
or
anything,
but
generally
we
don't
have
a
consistent
definition
of
updates
on
anything.
So
I
think
that
would
be
one
way.
A
H
Usually
caught
by
somebody
who's
vigilant,
not
by
any
anything,
that's
gentle
in
process,
so
I
found
a
few
of
these
mistakes
where
somebody's
using
the
same
blade
for
different
things,
and
so
like
I,
just
like
say
if
I
would
notice
it,
but
I'm
not
gonna
notice.
All
the
drafts,
but
I
do
think.
We
need
like
a
more
rigorous
way
of
doing
this,
so
like
Michael,
is
ready
for
something
else,
but
I
think
it's
like
similar
like
we
need
to
do
this
for,
like
most
of
the
bit
options
and
everything
I
agree.
L
A
L
S
U
U
L
L
M
M
N
N
N
N
A
V
So
I
looked
at
this
draft
quickly
and
this
is
something
we
have
actually
implemented
using
forces
in
the
past,
so
it
was
an
easy
fit.
I
put
these
slides
quickly,
so
quick,
maybe
just
jump
to
the
next
slide
right.
So
in
forces
you
have
a
data
model
I'm.
Assuming
that
this
draft
is,
you
have
an
information
model,
but
we
also
have
some
reason.
You
also
define
the
yang
models,
so
he's
he's
young.
Have
you
decided
on
yang
as
the
only
data
model
or
no?
V
V
However,
you
can
define
any
resources
you
want.
We
define
constant,
we
constantly
use
Linux
resources.
Anything
is
defined
as
such.
Proc
etcetera
write,
a
document
for
it
and
we
start
talking
to
so
from
the
data
model
document.
You
can
define
an
interface
for
applications.
Applications
is
a
very
simple
protocol
with
very
a
handful
of
protocol.
V
V
So
the
proto
semantics
are
very
simple.
Like
I
said
it's
not
an
RPC
based
well,
the
RPC
you
in
the
sense
that
there's
only
a
few
verbs
infinite
amount
of
nouns,
which
look
like
your
eyes,
designed
for
very
high
throughput,
so
binary
encoding,
which
is
formally
defined
in
the
data
model,
traffic
sensitivity,
optional,
high
availability,
both
of
the
controller
and
the
data
path.
Next
one
protocol
semantics
very
simple
publish/subscribe
is
built-in,
as
well
as
a
request
response
next
and
I
love.
The
class
like
I
said,
is
object-oriented.
V
You
can
build
a
Bayes
model
and
then
inherit
from
it.
When
you,
when
you
instantiate
these
objects
in
the
resource
path,
you
can
have
instances
of
them
so
good
fit
for
multi-tenancy.
We
always
put
multiple
containers
each
with
its
own
LP
class
instance
of
the
same
resource.
Next
I
just
keep
going
all
the
way
to
the
end.
I
think
keep
to
the
go
to
the
last
slide.
V
Okay,
so
there's
a
very
good
fit
with
DNA,
because
we
really
have
defined
the
data
path
and
control
separation.
There's
any
the
information
model.
When
I
took
a
quick
look,
it
they're
not
exactly
one
to
one
mapping
but
there's
very
close
mapping
to
what
we
have
already
implemented
and
the
protocol
semantics
have
a
good
fit
a
slide.
G
Well,
Burt
sprint,
so
my
group
actually
has
the
primary
open
source
implementation
for
open
daylight's,
so
the
FPC
agents,
just
my
general
comment
right
now,
we're
using
an
agent
heavily
to
get
more
benefits
of
common
provisioning,
other
items
with
respect
to
forces.
Is
it
something
that
fits
well
with
the
agent
model,
or
is
it
very
much
something
this?
It's
very
much
on
the
data
plane
note
itself,
but
we
decline
the
data
so
that
you
could
this
you?
Could
the
controller.