►
From YouTube: IETF94-MIF-20151106-0900.webm
Description
MIF meeting session at IETF94
2015/11/06 0900
B
B
My
name
is
chill
salami
asian
para
vehicle
chair
this
meeting
and
no
to
wear
a
blind
Lisa
friday.
So
probably
you
has
already
rated
for
many
times,
so
it
make
sure
you
follow.
The
ITF
process
follow
a
PR
roads
when
you
make
a
contribution
to
this
working
group
and
what
you
opposed
and
what
you
discussed
in
this
meeting,
where
team
has
a
conclusion
to
the
air
so
SM.
B
Administrivia
and
we
need
a
minute
taker
and
when
Jabba,
squadra
and
sank,
two
horrible
in
here
we
pass
already-
has
the
meanest
acre
shadow
squadron
and,
as
a
chair,
a
solid
joined,
the
chapel
room.
Other
you
food
as
anyone
here
who
winning
to
take
to
heat
another
chopper
squad
up
that
way,
vicarage
and
in
volunteer
for
the
chopper.
C
B
B
B
This
is
a
chain
that,
with
post
for
today's
discussion-
and
we
has
already
completed
a
teachable
and
and
then
we
were
as
a
working
with
shareware-
is
five
minutes
to
discuss
a
walking
go
status
and
you
say
where
we
are
and
how
do
we
finish
this?
Where
we
have
a
two
important
part:
the
clothes
the
powder
is,
we
have
to
powder,
the
post-party
is
very
important
part,
a
that
is
chatter.
B
What
hadn't
talked
about
the
air,
so
we
can
service
model,
we
invite
also
to
them,
and
neither
the
discussion
we're
talk
about
the
what
a
change
we
has
all
made,
what
a
issuer
has
all
the
coast
and
what
is
remaining
easy?
We
still
need
to
address
and
what
a
proposal.
How
did
she
has
an
issue
and
will
focus
on
the
open
issue?
Discussion
you
can
and
it's
not
involving
me.
We
need
to
extend
the
time
as
for
this
topic,
and
so
so
second
part
is
about
as
a
worker.
B
In
this
part,
we
post
early-
this
council
liaison-
we
received
from
the
metro,
is
nana
forum
and
yes,
and
we
were
invited
to
mahadji,
and
we
have
any
efforts
to
give
a
quick
update
on
what
am
I
doing
regarding
llamada
for
the
nef
service
will
have
a
good
discussion
on
how
any
of
work
is
going
to
ikea
work
and
in
regime.
We
have
a
shadow
proposal
from
the
operators
and
we
here
we
list
the
tube
older
one
is
service.
B
Automation,
management
architecture
proposed
by
China
could
come
at
you
wrong
with
all
the
way
we
can
discuss
about
this
and
second
pathology
about
the
bus
service
model,
and
this
is
a
new
service
model
and
patent
right
now
is
not
in
school
lusm.
So
this
meeting
chato
ramon
you
and
we
are
not
a
focus
on
which,
having
so
other
way,
sinker
right
now
so
many
service
model,
/
bolster
and
there's
no
ride
home
place
for
this
abode.
Oh
and
LCM
is
the
only
working
to
talk
about
the
service
model,
so
it's
income.
B
B
You
seem
not
so
we
so
so
what
is
the
working
with
scatters,
where
we
are?
Actually,
we
have
a
laces
second
meeting
in
Yokohama.
So
since
prager
we
made
a
good
progress
on
service
model.
We
got
a
lot
of
discussion
and
we
talked
about
a
medical
support
and
given
a
very
very
supporter,
generic
of
us
and
VPN
service,
identity
and.
B
The
the
leading
editor
actually
since
the
pride
meeting
they
make
a
quick
revision
and
I
try
to
close
the
issue.
We
register
on
the
prod
meeting
and
the
key
change
actually
is
exercised.
The
moderate
approach
try
to
talk
down
the
service
model
into
several
building
block
that
can
be
used
by
the
future
service
model.
B
So,
as
working
chair,
we
we
actually,
we
set
up
the
wiki
pages
to
track
all
the
issues
that
has
already
discussed
some
of
has
already
closed,
someone's
still
open
and
we
also
plan
to
have
an
interim.
Even
before
you
cut,
you
come
out
that
the
unfortunate
on
October
is
very
busy
for
also
the
Sabbath
model
chocolate.
So
what
you
get
is
an
interim
meeting
set
up,
so
they
work
at
a
little
bit
slow
down,
but
I
think
we
still
have
to
do
this
thing.
B
We
need
to
finish
this
so
as
he's
eating
with
a
focus
on
these
to-do
lists
and
some
opening
issue,
and
hopefully
we
can
make
them
all
progress
on
this
chapter,
and
so
we
right
now
waiting
for
the
last
call
we
sinker,
maybe
after
these
Yokohama
king
and
we
were
John.
This
discussion
trying
to
cause
all
the
issue
and
hopefully
we
can
get
worked
on
as
enemies
appear
at
the
beginning
of
the
next
year
next,
so
this
is
a
current
milestone.
B
B
B
D
Hi
so
yeah.
This
is
a
brief
update
and
go
through
some
of
the
issues
that
are
in
the
current
l,
three
sm
mother
and
stefan
very
kindly
prepared
the
slugs
for
a
server,
and
we
have
to
thank
him.
So
the
context,
of
course,
is
a
bunch
of
young
models
that
we've
got
coming
together,
but
also
these
young
models
need
to
combine
to
be
able
to
actually
configure
some
form
of
service.
D
So
the
intention
of
the
original
draft
that
Kevin's,
dyfan
and
and
and
lose
myself
pulled
together
we're
to
try
and
have
a
initial
picture
ago
for
a
service
model
around
the
way
through
VPN
service,
particularly
because
the
operators
that
we
worked
for
and
I
worked.
All
the
time
was
were
most
interested
in
that
service.
That's
kind
of
a
a
pretty
fundamental
service
for
most
most
large
operators.
So
the
intention
was
to
try
and
see
whether
there
was
something
common
across
across
the
different
service
models.
Ie
is
there?
D
Is
there
a
chance
to
do
anything
that
might
be
standardized
here
and
also
to
identify
the
the
things
that
that
would
then
need
to
map
to
and
device
config?
So
to
look
at
saying:
well
what
what
underlying
models
do
we
need
such
that
we
can
actually
configure
this
service
and
then
look
at
the
set
that
we've
got
in
the
ATF
and
see
where
they're
actually
usable
the
I.
D
Think
one
of
the
things
that
we'll
we'll
have
to
debate
is
how
far
that
stunt
the
standard
model
that
we
build
goes
because
a
lot
of
the
things
that
we've
been
kind
of
discussing
and
that
my
observation
from
the
initial
work
we
did
was
that
we
took
more
things
out
of
the
Medal
than
we
put
in
I.
The
set
of
things
in
the
model
was
the
standard
things
that
was
across
all
those
the
all.
D
The
operators
are
involved,
not
necessarily
trying
to
cover
every
particular
particular
eventual
eventualities,
but
to
start
with
I'll
go
through
them
where
we
are
with
the
draft.
There
have
been
some
changes
since
since
the
comments
that
I
think
we
made
in
the
session
in
Prague,
which
was
that
it
needed
to
have
some
more
groupings
in
it,
such
that
it's
more
reusable
and
put
more
modular
so
that
that's
been
done
and
Stefan's
kindly
moved.
Most
of
the
ban
Chester
two
groupings:
we've
then
had
a
look
at
the
the
VPN
policy
part
of
the
model.
D
Previously
we
had
a
native
VPN
that
kind
of
said
which,
which
site,
which
VPN
a
site
was
part
of
a
men
policies
that
allow
you
to
freaking
between
them.
Now
we
only
have
this
dystopian
policy,
which
is
slightly
more
abstracted
around
how
it
how
it
said
it
you
take
a
site
and
how
it's
put
into
various
of
a
VPN.
So
the
idea
was
here
that
you
could
just
express
things
like
a
full
mesh,
VPN
or
a
hub-and-spoke
VPN
and
give
a
site
a
role
such
that
you
can
have
multiple
hubs
or
and
multiple
spokes
it.
D
The
other
thing
we
kind
of
debated
quite
a
lot
when
we
originally
did.
The
model
was
around
whether
we
have
templating
within
the
model.
So
typically,
you
might
have
a
hub-and-spoke
VPN
where
90%
of
the
sites
of
spokes
and
they
all
have
exactly
the
same
policy.
So
it's
typical,
where
you
have
like
a
retail
provider
that
has
lots
and
lots
of
edge
sites
in
a
couple
of
headquarter
sites,
so
we
just
previously
had
this
kind
of
type
of
site
and
we
didn't
necessarily
have
a
templating
approach.
D
Now
we
have
a
separate
temporary
templating
approach,
so
you
can
say:
hey
here's,
a
template
for
the
site
and
and
then
apply
that
template
to
various
to
various
other
sites.
I
think
this
is
generally
something
we
should
consider
whether
we
want,
because
it
depends
really
who's
who
we
expect
and
how
we
expect
to
interact
with
the
model.
If
it's
programmatic,
then
there's
less
requirements
to
have
templates.
If
it's
more
that
a
human
operator
might
fill
in
a
form,
then
you
probably
do
one
templates.
D
The
British
I
think
we
discussed
that
we
had
mpls
as
a
kind
of
access
type.
That
said
that
this
was
going
to
be
a
PE
attachment
circuit.
Now
it's
a
it's
been
moved
to
being
a
leaf
from
the
VPN
basis.
So
it
kind
of
says
this
is
an
MPLS
VPN
and
and
then
we
only
have
the
signaling
type
on
the
access,
which
kind
of
describes.
What
sir,
what
the
attachment
looks
like
is
there
going
to
be
some
dynamic
routing
or
not.
D
I
think
we
had
BFD
directly
under
the
attachment.
There
was
no
convo,
am
a
separation
in
the
model
before
now.
Now
we
have
BFD
under
I
am
I
am
sure
that
that
could
be
augmented
by
different
operators
who
might
use
different
different
functionality,
but
I
think
at
the
time
we
debated
this
to
start
with
the
only
thing
that
we
commonly
used
his
BF
d.
So
that
was
why
that
that
was
there,
there
have
been
some
small
changes,
renaming
of
roofings
identities,
etc,
etc.
D
D
Stefan
has
also
been
working
hard
with
on
some
of
the
other
updates.
We
had
a
VPN
identity
and
then
a
VPN,
nameless,
separate
things
in
the
I
think
that
consensus
was
that
we
had
just
a
name
for
that
for
a
VPN,
and
that
was
used
as
the
identifier
for
it.
It's
probably
a
good
change
from
my
view,
because
it
kind
of
removes
some
of
the
the
the
inability
to
support
different
operators,
ways
about
identifying
VPNs.
We
have
dscp
matching
in
the
flow
definition.
I,
actually
don't
know
that
I
agree
with
this
change.
D
I
think
we
need
to
think
about
whether
in
your
service
definition,
you
really
want
to
talk
about
something
like
the
SCP
or
whether
you
want
to
define
some
classes
of
service.
Actually,
it's
kind
of
more
typical,
in
my
view,
that
that's
how
an
operator
service
is
sold
is
that
we
sell
bonds
of
a
gold
type
approach
or
another
multitude
of
colors,
because
you
want
have
six
classes
and
and
not
dscp.
D
In
my
experience,
so
kind
of
one
of
the
things
we
might
want
to
think
about
here
is
yes,
we
could
try
and
make
this
model
exhaustive
to
describe
multicast,
VPN
and
all
the
multicast
options
that
we
might
have
within
within
VPN
service.
But
equally
we
could
say.
Actually
this
should
be
an
extension
to
this
service
model,
and
you
should
say
you
should
augment
it
as
an
operator
or
as
whatever
and
well
you
instantiate
the
model
to
say.
Actually
these
are
the
multicast
options
that
I
support.
D
D
I
think,
every
time
we
go
through
these
things,
you
could
need
a
multicast
expert
to
come
and
to
have
a
bit
of
a
look
at
it
and
say
what
they,
what
they
think
can
it
be
I
guess
from
an
operator
that
sell
it
sells
many
multicast
VPN
to
be
great
to
hear
from
from.
You
is
to
some
input
as
to
how
how
we
might
map
that
together,
because
right
now
it's
is,
as
you
can
see,
pretty
pretty
basic
there.
D
We
have
a
few
issues
open
the
tracker
and
when
the
the
custom
and
that
address
is
used.
So
this
is
is
really
when
you
are
doing
break
out
to
it
to
a
third
party
site
and
you
need
third-party
service
for
the
internet
and
you
need
to
you
need
to
do
some
form
of
Matt.
If
we
in
some
cases
the
customer
might
provide
addressing
to
do
that
and
it's
not
provide
by
the
by
the
provider.
D
D
D
There
is
a
this
grouping
change
that
we've
made
already
talked
about
multicast,
there's
a
question
around
inventory,
so
to
say
it
should.
Should
we
have
operational
state
within
the
model
that
saves
which
sites
belong
to
VPN
or
which
sites
are
accessing
cloud.
I.
Think
this
operational
state
question
is
actually
more
general
than
that,
because
what
we,
what
we
have
to
kind
of
think
about
is,
should
we
have.
We
have
operational
state
in
all
of
the
kind
of
protocol
models.
D
Should
we
take
that
one
layer
up
and
have
an
abstraction
that
sees
something
to
do
with
the
overall
operational
state
parameters
that
relate
to
the
entire
VPN,
so
you
can,
you
could
potentially
have
that
or
on
a
per
site
basis
as
well.
That
says:
okay,
this
site
is
no
down
because
both
its
accesses
are
down
now
the
protocols
side,
if
it
doesn't
know
that,
because
it's
a
4-1
each
of
those
is
an
independent
set
of
upstate,
but
the
the
service
model
might
want
to
make
those
kind
of
abstractions,
which
are
quite
typical
right.
D
So
I
think
we
should
look
at
what
the
general
op
state
approach
should
be.
For
this
model,
is
it
in
this
model?
Does
it
need
to
be
in
a
separate
model?
Is
it
common
enough
across
operators
that
we
need
to
represent
that
now
we
didn't
talk
about
this
at
all
in
the
initial
cut
of
this
draft,
so
right
now,
I
would
say
that
the
kind
of
assumption
is
that
there
is
ZERO
upstate
in
it,
but
that's
only
one
approach
that
you
can
take
out.
D
I
think
when
you
can
just
about
use
this
model
as
an
infrastructure
service
right
now,
but
if
you
start
to
add
in
vases
and
things
that
a
lot
more
kind
of
customer
focused
in
there
for
27,
then
we're
going
to
very
much
struggle
to
make
this
an
adaptable
model.
I
think
we
need
and
I
think
we
kind
of
agreed
that
we
would
have
this
cloud
access,
because
it
basically
says
a
break
out
to
something
that
something
that
isn't
this
disk
current
VPN
and
that
that's
relatively
common,
I
think
the
attached.
D
D
I
think
this
is
a
an
interesting
split
between
OSS
and
BSS
architecture
as
to
where
you're
going
to
store
things
so
I
think
you
need
a
tie
between
the
two,
because
your
operation,
operational
team,
will
always
need
to
know
where
the
thing
is,
if
they
need
to
report
it
to
a
third
party
or
actually
go
there
to
go
and
fix
something.
But
it
doesn't
necessarily
need
to
be
that
that
system
is
the
master
of
that
data.
So,
in
my
view
we
could
say:
let's
have
a
let's
duplicate
the
data
here.
D
D
Well,
so
that
new
slide
I
haven't
seen
perfect,
so
I'll
quickly
read
the
slide
and
then
I'll
present
it,
and
you
can
pretend
that
you
didn't
read
the
slide.
Whilst
I
was
reading
it.
So
yet
I
actually
didn't
read
the
thread.
That's
good.
We
have
some
discussion
as
right
now
we
kind
of
say:
oh,
they
should
be
disjoint
that
they
should
be
multiple
accesses
to
a
site,
but
we
don't
necessarily
describe
how
they
should
be.
They
should
relate
to
each
other
and
there's
a
proposal
that
we
have
a
basic
set
of
constraints
like
well.
D
We
have
a
set
of
constraints
and
I
think
the
list
was
not
that
basic.
It
was
things
like
a
latency
constraint
and
bandwidth
constraints,
disjoint,
list
constraints,
I
think
this
is
a
noble
ambition,
but
I
would
encourage
anybody
to
try
and
actually
come
up
with
a
sufficiently
generic
model
for
this,
that
you
can
use
it
in
more
than
one
operator,
because
the
the
two
twice
I've
tried
this
in
the
operators:
image
I've
words
and
had
to
model
it
entirely
differently.
So
I
think
we.
D
We
don't
necessarily
need
to
say
that
this
is
something
that
you
have
in
the
base
model,
it's
something
that
you
can
add
as
an
operator
and
it's
pretty
easy
to
add
it
right.
So
you
have
a
disjoint.
You
have
a
multiple
set
of
axis.
Is
you
have
you
can
have
some
base
identities
that
say
Holley
the
same
access
you
know?
D
Are
they
two
vlans
on
the
same
same
physical,
ethernet
bearer
or
are
they
different
and
then
an
operator
can
add
in
whatever
the
constraints
they
might
want
there
because
to
some
extent
you're
going
to
get
into
or
what
would
what's
disjoint
its
disjoint
that
that's
its
ethernet
over
sth
on
one
side
and
this
plane
Ethernet
on
the
other
side
or
what
and
then
that's
a
different
story
and
different
operators?
So
my
view,
let's
not
address
this
in
the
model.
Let's
leave
this
for
operator
extension
when
they
use
the
model.
D
If
that's
the
required,
we
then
have
a
set
of
things
that
we
haven't
yet
worked
on
security
parameters.
There
is
a
we
should
have
some
security
relief
in
the
in
the
model,
but
it
doesn't
really
do
enough.
So
it's
really
to
say
you
know
what
what
do
we
need
to
be
able
to
specify
other
things
for
ipsec
access
and
that
kind
of
thing
that
we
should
be
able
to
specify
and
yet
the
the
the
focus
of
Kevin
lose
Stefan
and
I.
D
D
But
it's
really
I
think
this
is
an
important
thing
here
that
says:
do
we
need
to
think
about
how
operators
should
use
this
model?
Should
they
use
it
to
say
that
this
is
you
know?
This
is
something
that
you
can
always
extend
to
do
any
l3
VPN
service
or
two.
We
really
expect
that
the
way
that
they
should
use
it
is
that
they
should
there
is
that
it
will
be
a
drop
in
service
model
for
their
early
3d
end
service
into
a
s.
D
I,
don't
think
we
should
include
this
in
the
modern
able
to
step
in
here.
This
is
again
a
huge
area
of
odd
complexity
that
that
actually
brings
rise
to
a
bunch
of
the
complexity
in
service
models
of
for
operators,
because
you
have
to
inter
work
your
service
with
somebody
else's.
So
do
all
these
mappings
and
how
you're
going
to
allocate
things
what
the
kind
of
interconnect
model
is
going
to
be
I
think
it's
not
it's
not
key
to
have
it
in
this
model.
D
We
don't
really
have
an
option
of
hybrid
VPNs,
so
at
the
moment
I
don't
think
we
really
have
a
good
way
to
describe
so
break
in
from
the
public
internets.
We
can
have
a
break
out,
but
we
don't
really
have
a
way
that
you
can
say
well,
here's
this
site,
and
actually
this
site
is
special
because
what
it
does
is
it
lends
a
bunch
of
hip
set
connections
there
from
n
other
sites,
which
is
a
pretty
pretty
common
model
from
workers
as
well.
D
So
do
we
need
to
do
we
need
to
model
that
is
actually
an
extension
again
that
comes
back
to
if
you
do
l2vpn
as
a
business
service.
That's
a
very
common
thing:
vol
3
VPN
as
a
service
that
Sam
that's
an
infrastructure
service,
then
you're
probably
not
going
to
use
that
and
then
is
there
anything
else,
and
that
probably
comes
back
to
the
question
that
our
steam
chairs
we're
asking
is:
are
we
ready
to?
Let
us
call
this?
D
C
You
are
that's
it
yet
right,
I
assume
that
there
would
be
something
more
in
terms
of
the
state
of
provisioning,
something
just
because
I
put
something
some
configuring,
the
yang
model
doesn't
mean
it's
actually,
provision
right
and
the
progression
and
provisioning
might
be
something
else.
You
want
to
consider
as
operational
state
right.
D
And
I
think
that
then
comes
to
debate
about.
Where
is
this
instantiated
right?
So
is
it?
Is
it
in
something
that
does
workflow
that
does
that
provisioning
and
then,
and
then
is
that?
Is
it
this
model
that
has
to
do
that?
Is
there
enough
commonality
between
our
workflow
systems
that
we
can
actually
store
the
the
common
states
earth
here
be
dragons
and.
A
So
Rob
thank
you
for
stepping
up
and
doing
this.
What
I
think
we
should
try
to
do
in
the
room?
If
we
can
it's
just
get
back
through
these
slides
reach
of
the
issues
that
are
still
not
as
unresolved
and
see
whether
there's
anyone
in
the
room
who
wants
to
take
a
position
on
each
one
and
see
if
we
can
get
in
progress.
D
So
I
think
we
need
to
clarify:
what's
it
what's
there
right,
so
that
the
flag
is
not
here
are
the
parameters
that
you
shall
use
for
this
OEM
protocol.
It
should
be
judicious
OEM
protocol
be
used.
So,
if
you
imagine,
when
I,
when
a
customer
oil
as
a
site,
they
have
some
lightness
detection
requirement,
actually
how
that
depends
quite
a
lot
on
what
the
CP
they
put
on
the
other
end
of
the
connection
it's
so
it's
really
a
case
of.
D
Are
you
going
to
run
be
BFD
true/false,
rather
than
I
want
to
run
BFD
with
this
multiplier,
etc,
etc?
So
I
think
there
is
some
abstraction
you're
not
going
to
get
to
a
state,
in
my
opinion,
not
going
to
get
your
stage
where
we
say
generically.
Is
there
some
liveliness
detection,
because
simply
because
customers
need
to
be
able
to
specify
something
to
do
with
it?
So
would
the
the
achievable
level
of
abstraction
is
an
interesting
debate
and
I
think
the
way
we've
got
it
right
now,
it's
a
reasonable
balance
of
that.
Ok,.
G
I
thinka
data
listener
may
be
necessary.
I
agree
this.
The
user
Kissin
your
poster,
but
my
comments
aren't-
is
a
walk.
I
asking
the
days
of
the
first
one:
either
you
could
use
a
kiss
you,
the
necessary
I,
seen,
for
example,
the
OEM,
because
you
for
this
is
a
necessary
but
I
think
maybe
you
can
defend
the
user
user
friend
language
instead
of
directly
you,
the
professional
technology,
yeah
I
think
this
may
be
a
challenger
for
the
user
because
they
also
hear
the
net
I.
Think
this
view
you,
the
device
APIs
the
language.
G
D
D
So
if
we-
and
that
the
same
is
true,
is
if
you,
if
you
say,
there's
actually
a
BSS
system
that
interacts
with
this
thing
to
instantiate
it
on
the
network.
At
that
point,
the
name
is
a
little
irrelevant,
but
actually
it's
quite
important
because
the
the
user
needs
to
be
able
to
sufficiently
accurately
tell
what
this
thing
is
when
it's
going
to
go
onto
the
network
because
they
actually
do
know
what
what
their
talk.
D
G
So
I
say
this
is
my
comment:
I'm
feeling,
but
I,
think
I.
I
hope
at
the
Soviets
model
is
another
good
developer
as
a
last
use
a
similar
as
the
device
a
base
model.
They
see
you
first
in
my
water
and
the
second
one
if
we
think
that
may
be
that
different,
the
other,
they
had
a
different
requirement
on
the
service
model.
Maybe
split
this
one,
maybe
to
some
the
augment
of
a
man
oughta
mental
work
instead
of
a
different
or
since
the
user
service
model,
but.
D
D
I
actually
would
really
struggle
to
get
to
the
stage
where
we
have
a
sufficient
level
of
abstraction
and
for
it
to
be
user
presentable,
whilst
keeping
it
common
across
more
than
one
service
provider,
because
different
service
providers
sell
in
a
different
way
to
different
people
and
they're
going
to
all
those
users
are
all
going
to
have
different
views
of.
What's
simple
and
what's
not.
E
My
hair,
shaitan
and
learning-
I
actually
agree
with
Europe
that,
as
it
comes
to
abstraction
of
these
service
levels
at
least
the
way
we
have
done
it
in
F,
which
I'll
talk
a
little
bit
about
next,
is
the
level
of
abstraction
that
we
are
providing
is
really
between
what
we
call
business
applications
and
the
orchestrator
level,
where
you
really
don't
talk
about
VFD
intervals.
So
you
talk
about
NAT,
addresses,
you're,
really
talking
about
service
level
parameters
and
leaving
kind
of
the
rest
of
the
intelligence
in
the
levels
below
that,
and
some.
D
D
D
So
that's
where
things
get
interesting,
because
that
good,
so
my
I
would
call
that
product
and
because
the
that's
what's
independent.
Actually,
when
it
comes
down
to
her.
In
my
view,
most
operators
don't
have
a
product
that
sufficiently
abstracted
from
the
technology
and
and
that
that's
fine,
but
this
is
going
to
be
one
of
the
tensions
of
getting
to
a
standardized
model
right
if
we
want
to
have
a
sufficiently
abstract
model,
a
we're
going
to
kind
of
go
away
from
the
same
of
what
are
the
underlying
models.
D
We
need,
because
we're
going
to
have
this
really
generic
thing
that
doesn't
map
to
technologies
and
then
doesn't
tell
us
what
the
underlying
models
we
need
our
and
would
I
actually
don't
believe
they'll
be
operationally
useful
because
I
still
don't
know.
It
I
think
that
you're
going
to
end
up
saying:
oh
here's,
some
fluffy
things
that
haven't
necessarily
necessarily
delivered
anything
useful
for
an
operator.
C
My
name
is
Michael
sharp.
I
guess
this
issue
as
well
as
others.
It
really
comes
down
to
the
definition
of
service
model
and
you
can
use
this
model
in
very
different
ways
and,
as
my
suggestion
is
at
least
document
in
the
draft,
what
you're
about
what
the
auto
syncing
is
than
what
this
model
is
about?
Who
interfaces
to
what?
Because
this
is
not
clear
and
a
lot
of
the
issue
is
really
come
down
to.
Who
is
using
this
model?
C
For
what
and,
as
my
suggestion
would
be,
try
to
leave
at
least
document
this,
and
they
talked
that
current
draft
is
very
back
on
this
piece,
and
this
causes
a
lot
of
confusion
in
my
point
and
I
think
you
can
try
to
describe
it
at
least
one
architecture
where
this
fits
in
and
then
a
couple
of
these
issues
are
very
clear.
It's
very
easy
to
sort
them
out,
because
then
it's
Q.
E
Miyashita
man
actually
I
will
second
that
opinion
from
Michael,
but
on
a
slightly
different
common
and
I
haven't
liked
the
crowd
I.
Just
let
read
the
abstract
to
be
able
to
speak
a
little
intelligently
right
now.
The
one
of
the
issues
that
means
struggled
in
math
is
to
figure
out
when
we
define
the
service
level
model,
and
you
say
that
in
your
draft
that
you
are
not
talking
about
device
level
configuration.
E
But
what
would
be
interesting
as
an
observation
is
then
l3
we
can
service
is
defined
to
you.
Are
you
able
to
find
all
the
device
level
configurations
in
the
models
that
are
there
today
or
would
there
be
a
recommendation
for
you
to
say
we
need
to
have
a
corresponding
models
developed
in
the
devices
also.
D
D
F
I
agree
and
what
I
want
to
say:
that's
what
what
I
personally
she
here?
Yes,
what's
is
just
a
position
like
from
my
point
of
view
v,
everything
in
one
bucket
right
now,
like
we
mixing
service
configuration
in
mixing
equipment,
configuration
it's
in
as
a
result,
we
have
just
one
big
ugly
tree
about
configuration
and
that
what
we
should
do
we
should
do.
F
D
Right
and
that's
repeat
myself,
I
think
thing:
the
problem
is
some
customers
come
and
say:
I
journal
l2vpn.
With
these
two
end
points
some
say:
I
want
this
LG
VPN
hands
of
this
empty
you.
Mr..
This
is
the
thing.
These
are
the
characters
most
of
this.
Could
this
have
don't
want
p
mappings,
so
we
can.
What
you
need
to
have
is
a
model
that
can
take
all
of
the
possible
inputs,
but
you
can
use
your
subset
of
them
right.
F
It's
is
impossible
to
beautiful
how
to
say
some
model,
which
include
everything
yes
at
once,
but
it
possible
to
build
some.
You
know
reusable
models
divided
into
some
logical
blocks,
as
you
know,
and
what
else,
and
also
what
I
want
to
say.
Even
if
you
talk
about
oil
shale,
oil,
free
VPN,
yes,
what
what
problem
I
can
see
right
now,
we're
actually
service
comes
our
service,
starts
and
read
our
service
ends
well
and
what
I
mean,
because
right
now,
the
demarcation
line
for
our
services.
As
I
understand
it
will
be
e
or
c
device.
F
D
D
This
is
probably
what
I
would
put
in
it
and
we
found
a
subset
that
was
that
was
probably
acceptable
from
things
that
we
capture
from
our
custome
customer
or
captured
from
our
customers,
that
that
would
build
the
service
as
soon
as
we
get
into
where's
the
service
demarcation.
You
know
we
are
not
going
to
get
to
a
standard,
because
so
I
went
to
a
British
Telecom
till
very
recently,
and
we
haven't
had
another
company
that
did
the
access
link.
D
So
the
service
demarcation
really
had
this
odd
thing
that
the
Seas
run
by
one
bit
of
the
business.
The
P
is
run
by
another
bit
of
the
business.
A
third
P
that
actually
just
the
service
delivery,
is
run
by
a
third
bit
of
the
business.
If
you
try
and
describe
that
you're
going
you
already
baked
in
regulatory
things
with
an
environment
that
you
wrote
in
company
structure,
so
the
debate
here-
and
actually
this
is-
is
one
of
the
reasons
I
think
we
should
be
debating.
D
This
is,
is
there
a
and
is
there
a
model
that
actually
works
for
more
than
what
operator?
Because
if
there
isn't?
Let's
not
standardized
this
right,
let's
let's
say
fantastic:
we
had
a
look
at
it,
it
didn't
work
off,
we
go,
but
but
the
value
of
standardizing
it
is
all
the
mappings
that
we
do
to
underlying
yang
models.
If
ninety
percent
eighty
percent,
sixty
percent
that
the
stuff
is
the
same,
it
means
that
anybody,
that's
writing.
Mappings
can
write
once
and
then
extend
their
stuff,
but.
F
Probably
I
believe
it
based
on
their
buyers
because,
in
my
in
my
point
of
view
is,
I
would
say,
probably
easy
to
do,
because
you
know
if
you,
if
you
have
as
a
level
of
abstraction
yes,
it's
it's
for
OSS
and
BSS
is
obvious.
You
just
need
some
service
level
of
abstraction
which
customer
face
and
you
need
/
service
like
become
positive
service,
the
company
and
service
components,
and
they
did.
B
H
You
very
quick
comment
about
having
two
different
types
of
services
where
they
might
be
very
complicated
to
set
them
up
or
simple
service,
and
hence
the
thing
you
said.
The
Mon
tearfully
covers
the
complicated
case.
There
is
the
choice
potentially
to
have
a
simplified
model
and
orientations
and
extensions
to
do
more
complicated
ones
as
well.
Yeah.
D
H
D
It
makes
the
base
model
more
usable
to
more
people,
so
in
in
BTW,
then
you
use
NL
through
VPN
to
broadband
backhaul
and
you
use
an
l-3
VPN
for
as
a
business
service.
Those
things
actually
have
sixty
percent
at
the
same
config,
but
the
other
forty
percent
is
entirely
different.
And
if
you,
if
you
do
this
kind
of
model,
then
you
end
up
with
one
one
big
blob
and
then
you
need
to
rewrite
it
to
do
the
other
service.
So
I
think
it's
a
good
suggestion.
G
But
I
think
the
because
we
have
a
many
if
you
member
I
work
at
our
need,
but
it
seems
for
your
the
service
model
and
you
now
use
not
only
related
with
the
VPN
and
also
with
the
with
the
net,
with
the
OEM,
with
the
queues
and
the
width
and
also
maybe
I.
Remember
the
next
slide.
Also
with
some
days
the
constraint
may
be
related
with
tunnel
I
think
this
will
need
a
very
experienced
people
to
understand.
Almost
all
features
in
the
network,
so
I'm
not
sure
how
you
work
so.
D
If
your
network
in
an
operator-
and
you
have
to
only
serve
this,
you
have
to
understand
all
the
bits
of
your
service
right.
Actually,
it's
it's
a
bonus.
If
you
do,
but
there's
always
that,
but
you
don't
really
understand
but
yeah
and
that's
I
think
it
comes
down
to
you
in
an
operating
probably
going
to
understand
most
of
the
things
that
in
the
service
or
not,
maybe
not
in
the
same
detail.
You
need
to
do
the
device
level
model,
but
you
have.
D
G
D
Entirely
related
with
the
device,
because
the
only
reason
we're
possibly
capturing
the
things
that
are
in
this
model
is
to
configure
them
on
devices
to
deliver
a
service.
So
if
it
doesn't
relate
to
the
device
config
or
something
to
do
with
the
way
that
that
device
config
fits
into
into
the
service,
then
we
shouldn't
be
capturing
it
the
model.
I
guess.
H
And
robertson,
I
wonder
whether
isn't
effectively
two
different
models
being
discussed
here
wonders
the
male
year,
you're
sort
of
discussing,
which
is
one
for
the
operators
to
provision
services
and
one
that's
been
discussed
here,
which
is
almost
one.
That's
client
facing
or
customer
facing
that's
a
different
level
of
complexity
in
detail
and
that's
a
unity
so
promoter
that
I
think
try
to
get
one
team.
Five
is
too
complicated.
C
D
Tunnels
no
overlap
so
that
the
the
comment
that
was
it
that
we
mentioned
the
the
common
we
mentioned
was
around
sites
that
are
not
connected
via
physical
wires.
Basically,
so
the
only
reason
when
you
have
a
site-
that's
not
connected
by
a
physical
wire,
it's
probably
connected
by
simple
virtual
tunnel
right,
but
it's
not
we're
not
going
to
describe
all
of
the
the
parameters
that
are
required
for
it
is
that
it's
the
key
things
that
are
required
like
say
it's
GRE
connected.
D
We
probably
need
to
know
the
GRE,
the
IP
of
the
endpoint
or,
if
it's
not
here,
if
it's
not
add
something,
that's
did
I
namak,
Lee
discovered
and
any
kind
of
key.
Those
are
the
opening
parameters
that
you
really
need
to
capture
in
the
service
model.
It
the
device
config
model,
has
tab
a
lot
more
in
it
and
when
you
do
the
mapping
between
a
service
model
on
your
device
config
model,
what
you
really
do
is
buy
the
programmatically
or
transfer
through
translations
or
lookups
fill
out
what
the
device
configs
are
going
to
be
I.
A
Think
there's
a
bit
of
a
sort
of
up
level
there,
which
is,
if
you
were
making
a
service
model
for
l2vpn
or
if
you
were
making
a
service
model
for
tunnels
or
whatever.
Would
there
be
elements
of
this
service
model?
You
might
borrow
or
generalize
or
common
eyes,
and
that's
a
really
interesting
question
to
be
had
outside
this
room.
F
What
ball
the
I
believe
model
should
be
divided
into
baths,
while
parts
should
be
a
customer
faced
and
the
other
parties
should
be
resource
based
because
for
the
customer
customer
don't
care
for
how
I
provide
service
for
him.
He
just
needs
some
basic
parameters
from
the
series.
Yes
and
the
research
part
it
can
refer
to
exact
technology.
A
So
I
see
heads
in
the
room
shaking
one
of
them
is
mine,
but
I'm
too
lazy
to
get
up
and
go
to
the
mic
depends
on
your
customer.
If
you
go
and
hire
a
car,
sometimes
you
just
want
a
car.
Yeah
click
I'll
have
I'll
have
one.
Sometimes
you
want
a
red
one
or
you.
You
want
one
with
a
certain
amount
of
trunk
space
or
in
fencing
not
having
a
VW
diesel.
A
Okay,
if
you,
if
you
take
your
view,
then
the
abstract
interface
you
present
to
the
customer
is
probably
as
simple
as
how
many
wheels
does
it
have
an
engine
can
many
seats.
If
you
want
to
worry
about
facilitating
the
advanced
customer,
you
have
to
be
extensible
to
some
of
these
other
parameters
in
a
key
thing
around
that
in
my.
D
Opinion
is
actually
what
we're
saying
here
is
what
we're
building
a
paradigm
that
says
we
map
between
levels
right.
We
map
between
this
layer
of
surface
model,
which
is
results
facing
potentially
to
device
config,
but
there
is
no
reason
you
can't
infinitely
stack
that
mapping
and
actually
in
the
systems
that
I'm
currently
working
on.
That's
what
we
do.
D
We
have
fragmented
the
configuration
for
devices
into
into
into
abstractions
and
that
are
much
more
and
then
we
have
services
that
use
sets
of
those
kind
of
bread
crumbs
to
builder,
build
a
service,
and
then
we
actually
have
a
service
that
sits
on
top
of
it,
which
is
which
is
much
more
meaningful
to
the
user.
So
we
just
bill
introduced
an
N
layer
mapping
to
be
able
to
say
take
this.
This
is
level
of
detail,
but
the
consumer
wanted
to
give
us
and
then
we've
map
it
down
into
wherever
it
tends
to
go.
D
And
yes,
somebody
has
to
write
the
way
that
says
when
you
ask
one
of
those
I'm
going
to
automatically
fill
out
these
300
options.
If
it
happens
to
be
that
complicated,
but
the
we,
if
you
say
actually,
we
have
to
have
one
more
model
and
it
has
to
be
the
for
this
customer
facing
it
has
to
be
this
abstract.
D
When
you
end
up
with
exactly
the
problem
that
Adrian
described,
which
is
everybody
has
to
have
the
same
thing
and
in
this
area
one
of
the
big
problems
is,
you
have
to
remember
that
operators
have
hundreds
and
hundreds
of
salespeople
they're
going
out
and
telling
people
that
they
should
buy
this
unique
snowflake
of
a
model
because
that's
our
product
and
that's
what
differentiates
us.
So
operators
are
not
going
to
adopt
something
that
where
we
overly
abstract
it
and
don't
give
them
the
freedom
to
be
able
to
do
to
differentiate
within
it,.
A
We've
got
a
number
of
issues
in
the
tracker,
some
of
which,
as
we
are
presented,
we
have
proposed
resolutions
and
some
of
those
resolutions
need
updates
to
the
document,
which
is
relatively
simple.
The
two
things
we've
got
to
look
at
are
those
issues
that
are
not
yet
in
the
tracker
that
people
have
and
I
think
some
of
those
popped
up
are
at
the
mic
and
a
couple
of
them
actually
on
the
last
slide
in
the
presentation.
A
I
A
A
That's
possibly
a
few
more
hands.
Ok,
so
I
think
that's
a
fairly
clear,
marching
order
to
the
chairs,
and
since
the
chairs
are
keen
that
we
try
to
deliver
on
our
agenda,
we
may
be
looking
for
a
date.
I
guess
in
early
December
I'm
a
work
with
the
author's
about
which
issues
they
can
drive
to
closure
on
the
list.
Obviously,
don't
wait
to
the
meeting.
Talk
on
list
do
work
and
we'll
try
to
push
through
some
of
these
I.
Don't
come
up
with
an
agenda
for
an
interim
hello,
Scott
hi.
I
Scott
man
Scott
again
just
a
suggestion
having
dealt
with
trying
to
organize
modeling
discussions,
I
would
have
two
virtual
meetings,
one
set
up
for
one
set
of
time
zones
and
another
set
for
a
different
set
because
we
constantly
bring
in
west
coast
and
China
and
East
Coast
in
Europe,
and
it's
tough
to
find
one
that
works
for
everyone
unless
somebody's
up
in
the
middle
of
the
night.
So
just
just
a
suggestion:
ok.
E
I'm
almost
glad
that
Rob
then
before
we
so
he
answered
most
of
the
questions.
I
was
expecting
to
field
so
meth
last
week
approved
a
yang
project
to
work
on
yang
models
from
neff
services
and
as
part
of
that,
they
decided
to
send
alies
on
statement
to
both
ITF
and
my
triple
e
to
kind
of
let
the
other
rescuers
know
about
the
work.
That's
going
to
happen.
That
is
going
to
be
happening
in
math.
So
let
me
talk
a
little
bit
about
that.
E
So
I
know
give
a
lot
of
questions
with
respect
to
water.
What
is
the
service
model?
What
are
the
service
models
trying
to
target?
Well
methods
try
to
do
a
maybe
a
little
better
job
of
trying
to
describe
exactly
what
they
are
trying
to
target.
They
have
what
is
here
and
LSO
a
reference
architecture
and
I'm
going
to
talk
a
little
bit
about
where
exactly
the
yang
models
that
we
are
looking
at
going
to
focus
on.
E
E
E
What
are
the
services
that
meth
is
looking
at?
Meth
has
a
whole
list
of
DC
based
services
that
they
have
defined
in
f6,
start
to,
let's
say,
initial
phase,
1
scope
of
that
project,
and
then,
following
that
obviously
basis
is
that
knife,
as
defined
in
net
51,
so
annex
some
examples
of
the
services
that
we
would
be
looking
at
as
part
of
trying
to
model
in
the
yang
we're
talking
about
an
e
line.
A
point-to-point
EVC
based
service
elong,
a
multi-point
multi-point
service,
iching
routed
multi-point
service.
E
So
these
are
some
of
the
examples
of
what
the
gang
molds
are
going
to
try
to
cover.
So
the
question
is
why
why
am
I
here?
Why
am
I
presenting
this?
Of
course
it's
to
make
sure
it
to
let
this
stand
body
know
of
the
work
that
Neff
is
looking
at,
but
also
possibly
looking
at
some
of
the
work
that
might
be
all
lapping
between
what
work
gets
done
here
and
maybe
gets
done
in
meth
and
see
what
it
is.
We
could
be
collaborating
that's
pretty
much
it
any
questions.
D
Shakir
jive
I
think
a
really
interesting
thing
to
hear
from
the
mef
side
of
the
house
would
be
how
people
envisage
mapping
and
the
interrelation
of
models
in
the
orchestration
type
layer
to
work
and
what
they
think
the
requirements
of
the
models
that
the
service
models
are
to.
Do
that,
because
I
think
there
are
some
nuances
where
you
might
want
particular
annotations
to
be
able
to
help
your
mapping
work
and
as
more
people
work
on
that
I
think
it'd
be
cool
to
not
have
to
everybody,
have
everybody
to
find
their
own
thinking.
D
I
Hi
Scott
an
answer
like
Ericsson
I,
can
address
at
least
do
some
pointing
to
where
to
find
some
information
about
that.
This,
the
meth
is
working
a
lot
with
other
groups
like
the
onf
in
the
ITU
one
on
this,
especially
in
the
area
of
defining
the
guidelines
on
how
to
use
UML
and
the
tool
sets
to
use
UML
and
how
you
move
from
one
to
another.
So
it's
all
work.
That's
in
progress.
There
is
a
lot
of
effort
being
put
into
place
there.
The
framework
that
we're
trying
to
use
is
this.
I
This
life
cycle
service
orchestration
model,
which
is
more
than
just
a
model
that
describes
the
lifecycle
associated
with
all
of
service
management,
but
it
provides
us
a
place
where
we
can
talk
about
product,
catalogs
and
service
catalogs
and
how
you
interrelate
products
and
services
and
resources
and
the
different
levels
of
models
that
that
need
to
be
created,
so
there's
lots
of
different
things
that
are
going
on
in
different
places.
So
if
you
have
any
questions
about
that,
you
can
contact
my
hedge,
you
can
contact
me.
I
E
Would
probably
start
back
at
raising
it
as
informational
and
then
maybe
as
a
data
point
to
say.
Well,
if
there
are
any
services
that
this
working
group
defines
that
needs
to
work
with
any
of
the
models,
I
I,
don't
see
anything
quite
obvious
as
yet,
but
if
there
are
that
would
be.
There
is
a
con
set
of
contact
points
to
you
for
you
to
reach
out
to
do
and
get
that
collaboration.
E
I
It's
got
again,
I
I
agree
with
what
Mahesh
said,
I,
don't
think,
there's
anything
in
particular.
This
group
needs
to
do
with
this
particular
liaison,
but
just
keep
in
mind
that
there
is
a
lot
of
it's
mainly
for
information.
So
there's
lots
of
information
going
around.
There
is
a
need
to
support
the
ability
to
do
this
type
of
orchestration
and
controlling,
and
there
are
other
standards
bodies
that
have
been
dragging
their
heels.
A
So
you
may
recall
that
we
have
precisely
one
deliverable
on
our
charter
so
discussing
other
documents,
whether
or
not
they
have
l
three
sm
in
the
name
means
we're
stepping
outside
our
charter
and
we're
just
making
a
bit
of
space.
At
the
end
of
this
meeting
for
people
to
do
advertisements
come
read
my
draft.
This
is
what
it's
about
and
and
come
and
play
with
me.
A
J
Grow
good
morning,
everyone-
and
actually
these
authors
on
the
list
on
the
page
she's
not
being
cannot
be
here.
So
I
present
these
slides
Gordon.
If
you
have
questions
that
I
cannot
answer.
So
please
Rah's
up
your
questions
on
the
on
the
mailing
list.
This
is
some
sorts
of
how
to
use
service
models,
defining
this
group
and
maybe
in
the
future,
and
how
to
use
these
service
models
combined
with
the
network
element
models
that
we
have
developing
ittf.
J
So
the
first
way
is
what
is
now
in
IETF
and
there
are
on
the
left's.
There
are
requirements
from
the
service
provider
and
the
right
side
is
something
efforts.
Hell
is
already
being
done
in
idea.
We
have
metal
metal
worker
service
models
and
there
we
have
net
conf
to
configure
such
kind
of
models.
You
know
we
also
have
network
element
models,
developing
many
groups,
so
the
question
here
now
is
how
to
deploy
them
in
the
real
network
and
what's
the
relationship
with
other
management
systems,
so
this
slide
is
current
model
status
in
IETF.
J
So,
according
to
the
classification
document
in
net
mode,
the
the
models
can
be
divided
into
two
types.
The
on
the
top
is
the
service
model,
which
are
abstract
some
kind
of
customer
requirements,
and
the
next
thing
is
the
on
bottom
line:
is
the
network
element
models
which
is
some
working
other
groups?
J
So
the
question
here
is
no
job
to
discuss
the
service,
automation,
architecture
that,
based
on
this
network
service
model
and
network
of
element
models
gesture
during
their
first
discussion,
the
relationship
and
how
to
use
this
kind
of
network
element
models,
which
is
specific
models
and
the
service
models.
This
is
proposed
architecture,
so,
on
the
scream
that
c1
okay
on
top
the
c1,
is
where
service
model
I
used
and
the
sieve.
Perhaps
it's
the
c3
is
the
instantaneous
of
series
connect
to
the
device,
which
is
some
specific
models.
So
is
this
architecture
they
sought
search?
J
Should
it
be
some
kind
of
c2
in
this
architecture
on
the
top?
It's
the
customer
okie
a
service,
ok
Streeter,
which
is
our
G
I,
which
is
used
to
map
the
service
models
into
some
technology,
independent
models
and
on
the
next
level,
which
is
the
network
service
orchestrator.
It's
used
to
map
the
some
kind
of
technology,
independent
models
on
to
the
technology
specific
models,
so
their
question
is
how
to
define
the
c2
interface
between
the
customers
of
this
orchestrator
and
network
service
orchestrator.
J
Also,
the
next
lie
is
something
they
think
what
can
be
wrong
on
the
c2
which
is
in
the
middle
is
so
we
have
already
have
c1
for
the
service
models
and
a
c2,
maybe
some
technology
independent
models
just
to
describe
what
I
want
from
the
network,
but
a
regardless
of
the
specific
technologies
used
in
different
domains
or
different
network
segments.
So
on
the
bottom
layer,
it's
the
technology,
specific
models
which
we
are
tell
the
infrastructure
does
pacifica
our
technology.
How
to
configure
so
I
think
that's
the
sorts.
J
Yeah.
Do
you
think
it's
a
useful
work
to
to
have
this
kind
of
architecture
and
tell
people
how
to
use
the
service
model
and
how
to
connect
it?
Two
wins
the
network
element
models
and
how
to
connect
with
some
technology
independent
and
dependent
models,
so
I,
yeah
I,
think
that's
the
last
side.
So
questions.
C
J
C
So,
under
on
this
year's
all
functional
game-
and
you
can
argue
a
lot
about
what
functions
are
there,
but
the
one
that
really
strikes
me
is
on
this
year's
or
number
three,
where
you
state
that
the
Sears
or
translate
service
parameters
to
network
element
configuration.
So
they
really
assume
that
the
seals
overseas
individual
network
elements.
J
C
A
Sir
Adrian
Powell
as
a
contributor
I
think
I
agree
with
that
point.
I
think
we
we
need
to
separate
talking
about
Network
elements
and
networks,
yeah
I,
think
having
an
architecture,
a
picture
2.2
and
some
separation
of
the
models
in
two
different
layering
just
in
our
heads
is
really
really
useful,
very,
very
useful,
I,
don't
think
it
particularly
belongs
here.
A
I
think
this
working
group
needs
to
be
looking
at
it
and
we
should
talk
with
the
ad
about
where,
where
we
should
position
is
work
within
the
area,
whether
it's
net
mod
or
whether
it's
in
the
ops
area
working
group
or
something
like
that,
I'm
sure
we
can
find
a
home
for
talking
about
it
more
widely
than
just
this
group.
Thank.
C
F
I
A
J
K
Hi
good
morning,
everyone
I'm
from
China
bio-
and
I
think
my
presentation
has
some
relationship
with
the
preview
presentation,
because
my
topic
is
about
the
younger
model.
For
then
you
added
a
service
and
last
shot
and
he
represents
the
vas
service
model
and
before
I'm
describing
the
last
service
model,
I
think
it
can
be
divided
into
five
questions
and
the
first
two
to
question
is
about
to
the
service
model,
and
why
do
you
want
to
a
service
model
and
how
to
use
a
service
model?
K
And
I
think
we
can
find
the
answer
from
the
preview
presentation,
because
the
service
model
is
the
abstract
way
to
express
the
users
can
requirements
and
how
to
use
the
service
model
and
the
service
model
by
macaroon.
The
service
model
and
the
specific
data
model
service
can
be
easily
provided
here.
The
service
model
is,
if
you
see
in
a
in
a
figure
and
with
the
red
star,
and
it
connects
the
application
and
the
orchestration
and
a
pouch
on
the
practice.
K
Real
question
is
about
the
Rancic's,
the
model
what's
last
fall
and
the
words
the
master
is
model
for
and
how
to
use
the
vast
model
and
amount
of
our
service.
I
think
all
of
you
know
was
mass
and
what's
the
last
games
model,
it's
the
one
of
the
service
mode
of
folkestone
last
service,
and
it
also
provides
a
common
structure
to
make
the
configuration
and
the
operation
of
the
right
service
and
how
to
use
it.
K
K
In
realizing
the
rascals
model,
we
use
the
young
module
the
first.
The
young
model
defines
the
advanced
service
components
here
is
the
roster
is
components
and
it
has
five
parts.
The
first
part
is
the
pouch,
the
adminstration
information,
such
as
the
canon,
ID
and
and
the
last
service
name,
masters
ID,
and
the
second
part
is
about
the
type
of
the
last
component.
K
Something
should
be
noted
is
that
we
don't
make
the
detailed
policy
and
configuration
here
that
we
just
use
the
policy
ID,
which
can
be
referred
to
some
policy
modules
and
such
as
in
ocean
power
group
or
any
other
policy
module,
workgroup
and-
and
the
first
part
is
about
the
last
availability
configuration
to
identify
whether
the
service
is
of
high
availability
of
not
the
five
has
unpacked
is
the
budget
of
management
information,
and
then
we
also
go
to
the
young
model,
are
defining
and
providing
to
our
cacique
months.
K
After
the
main
content
of
my
chapter
here,
I
want
to
show
you
a
child
of
the
service
model.
In
our
clouded
same
her
under
here,
we
decided
to
plan
five
kinds
of
services
from
service.
A
2
Series
united
center
on
the
service
is
the
power
or
load
balance
only
and
from
the
service
be
under
own
to
the
service
II.
There
are
some
service
function
channel
here
we
have
realized
the
service
function,
I
use
the
service
model.
In
the
bottom
teacher.
K
We
can
see
that
the
load
balance
is
adopted
here
and
so
and
then
come
to
my
question
and
as
we
all
as
I
know,
on
the
bus
service
model
is
out
of
the
charter
of
this
world
group.
So
I
want
to
ask
everyone:
is
the
vast
service
model
at
your
action
to
go
on
if
the
vast
model
is
that,
if
the
last
of
its
model,
if
the
direction
should
go,
then
you
like
to
participate
and
contribute
to
this
work,
because
we
think
it
is
valuable
and
then
come
to.
K
My
third
question
is:
how
can
master
this
model
work
with
on
the
recent
eviction
service
models
and
then,
before
my
next
step,
I
want
to
share
your
teacher.
This
is
a
teacher
I
find
out
in
the
best
book
on
this.
Is
this?
Is
the
story
about
the
banana?
The
banana
wants
you
to
be
happy.
Look,
it's
even
smiling
at
you,
and
this
is
the
last
day
of
the
IGF
94.
So
don't
be
sad.
G
I'll
Romeo
hobby,
I'm,
okay,
you
need
an
auto
changer
I'm,
a
hug.
Ok,
I!
Think
that
I
think
that
this
a
worker
you,
the
useful
yeah
but
I,
say
that
data.
You
did
because
now
the
vast,
because
in
the
data
center,
maybe
they
have
the
different
VPN
technology
will
be
taking
yeah,
not
only
the
the
airstrip
here
and
also
the
EVP,
and
also
maybe
the
excellent
like
the
way.
Yes,
I
think
this
may
be
also
related
with
the
many
Vivian
technologies
and
I
think
this
York
was
started
to
have
a
discussion
here.