►
From YouTube: IETF100-RTGWG-20171116-1330
Description
RTGWG meeting session at IETF100
2017/11/16 1330
https://datatracker.ietf.org/meeting/100/proceedings/
A
B
So
yeah
there's
one
more
document
that
that
we
didn't
review
on
Monday
in
our
first
session,
and
this
is
a
working
group
document,
that's
kind
of
dormant.
It
hasn't
been
updated
in
at
least
six
months
or
more,
and
this
is
draft
IETF
RT
DWG
policy
model.
So
this
is
the
routing
policy
model
that
you
know
it
was
originally
produced
about
two
years
ago.
I
would
say,
and
it
uses
the
open
config
paradigm
with
a
state
container
with
config
falls.
B
So
this
model
is
not
nmda
compatible,
not
compatible
with
the
new
data
store
architecture,
and
it's
generally
not
the
direction
that
the
IETF
agreed
to
go
on.
However,
it's
probably
still
a
useful,
useful
model,
but
I
won't
prejudge
things
here,
but
so
it
doesn't
show
up
under
the
documents
under
the
routing
area
working
group,
because
stuff
that
gets
archived
doesn't
automatically
show
up,
which
is
kind
of
times
out,
but
it's
still
there-
and
it's
still,
you
know
a
working
group
document
per
se,
so
we've
got
several
options
that
we
can
follow
here.
B
B
So
these
there's
a
similar
issue
in
BGP
that
that
occurred
with
one
of
the
BGP
models,
and
you
know
we
can
do
whatever
we
want.
But
but
it's
an
issue
that
if
we
find
this
to
be
a
useful
model,
then
we
should
figure
out
how
to
move
forward
on
it
either.
Based
on
these
just
document
work
just
start
over
with
any
document
that
follows
the
the
ietf
gang
modeling
versions
approach.
B
C
A
D
B
E
We
will
option
three
I,
think
it's
small
question
of
quality.
We
can
do
whatever
we
want.
There
are
plenty
of
good
things
in
the
current
models
that
we
can
reuse.
So
we
can
just
modify
the
current
documents.
Make
it.
Nama
and
MVS
are
incompatible
or
create
something
new
kind
of
accurate.
The
work
that
have
been
done
by
open,
perhaps.
A
C
The
other
point
is
the
IDR
BGP
model
is
holding
on
this,
so
I
do
have
because
I'm
an
idea,
chair,
I,
have
solicited
authors
who
were
willing
to
help
on
this
or
have
already
written
it,
but
I
don't
want
to
take.
It
would
be
indeed
using
them
number
two,
so
we
hope
to
get
the
BGP
model
out.
That
depends
on
this
in
two
weeks,
so
we
really
would
urge
quickness.
My
my
the
operator
said:
please
move
fast
and
we
had
a
prototype
one,
because
we
had
to
build
ourselves
so.
A
The
plan
is
to
contact
the
outer
Fresno
drug
and
eat
specifically
and
ask
whether
they
want
to
keep
working
and
make
version
two
and
ITF
government.
If
they
say
no
or
disagree
in
general,
then
probably
will
move
the
option.
Three
fully
acknowledged
their
input,
their
work
and
starting
new
work
in
Raju.
C
A
G
Robert
San
Cisco
I'm,
not
quite
like
option
2
in
terms
of
being
quicker
to
do
I
convert
doing
the
BGP.
Also,
the
actual
conversion
mba's
done
is
just
as
a
starting
point.
You
want
to
be
quick
if
you
had
any
pushback
on
this
is
the
option
three
and
two
starts
by
putting
the
yang
model
in
the
draft,
so
at
least
that's
there
and
then
build
the
text
around
it.
Yeah.
A
So
we
don't
want
to
go
back
to
the
because
you
used
my
step
without
my
permission
again
acknowledging.
We
fully
acknowledge
the
work
after
did,
and
we
would
like
to
be
good
citizen
and
first
approach
them
before
we
do.
She
I
know
we
can
convert
it
easily
and
we'll
probably
do
it,
but
I
would
like
you,
have
to
be
good
citizens
and
show
good
or.
A
H
H
We
have
a
couple
of
observations
to
start
with
and
then
we'll
go
through
with
a
quick
review
of
the
things
that
are
going
in
the
idea
on
this
topic
and
then
we'll
finish
with
some
discussion
of
ok.
Well,
now,
there's
an
architectural
implications
and
what
kind
of
new
work
we
should
have
to
be
doing
so,
but
at
rapid
some
of
these
observations.
H
The
first
observation
is
that
the
terminology
is
really
important,
but
it
shouldn't
necessarily
drive
like
in
innovation
of
new
new
technical
means,
so,
as
noted
like
5g
slicing,
to
be
able
to
rely
on
many
existing
technologies
or
tech
nodes
that
are
currently
development
on
DPN
space,
traffic
engineering,
data
models,
quality
of
service
and
so
on.
And
the
other
thing
is
that
there's
obviously
a
big
role
for
software
here.
H
So
obviously
many
virtualization
tasks
can
be
achieved
to
do
software
without
actually
cutting
the
protocols.
You
have
some
some
piece
of
software
that
creates
these
connections
between
networks
and
and
the
actual
protocols
that
you
run
in
the
network.
Typically,
the
networks
don't
actually
have
to
change
in
case-
that's
not
always
true,
but
but
it's
important
realize
that
that's
true
for
a
lot
of
the
cases.
H
I
should
also
talk
a
little
bit
about
like
the
different
kinds
of
technical
details
here.
So
so
it's
important
to
separate
what
what
do
we
have
in
protocol.
Sometimes
we
have
some
capabilities
like
you
can
create
this
virtual
private
network
on
top
of
some
existing
data
plane.
Sometimes
you
have
descriptions
of.
H
Virtualized
network,
such
as
data
models
that
are
important
but
they're,
not
the
ones
with
that
means
not
something
that
happens
on
the
wire.
But
it's
rather
construction
in
memory
and
it
may
be
modified
through
protocols,
but
but
it's
a
different
type
of
an
entity
than
in
the
actual
on
the
wire
protocol
Square
at
a
low
traffic
issue.
H
F
H
Try
and
understand
what
what
each
other
so
with
that
I
want
to
go
to
sort
of
a
quick
review
of
IDF
technologies
in
this
space,
and
this
is
really
quick
and
I
should
also
mention
the
diamond
personally,
although
my
co-authors
are
have
more
experience
but
I,
don't
know
much
about
this
I'm
generalist,
I'm,
trying
to
understand
this
base
and
ask
stupid
questions
and
try
and
figure
out.
Where
are
we
at
the
big
picture
level
so
that
there's
certainly
many
things
that
I'm
missing
here?
H
H
H
We
also
have
like
triple
a
protocols
for
the
mountain
case
in
profit
of
like
EAP
that
can
indicate
domains
in
some
abstract
sense
and
those
can
be
used
to
determine
which,
which
network
or
virtual
domain
you
want
to
connect
to.
Then
we
have
endpoint
VPN
technologies,
ip68
VPNs.
As
an
example,
we
have
provider
based
deviants
and
associated
technical
components.
H
Those
slots
in
this
base
problem
should
have
added
to
the
slide.
Also
data
center
virtualization
technologies
in
do
three
on
one
working
group
in
this
case.
Goes
it
a
lot
of
traffic
engineering
work
at
the
ITF.
Yes,
working
group
was
an
example:
enhancing
the
traffic
engineering
capabilities
in
MPLS
and
keyless
networks.
H
We
have
tunneling
technology
again
planning
to
list
here
examples:
service,
chaining,
also
sort
of
a
Parton
a
component.
This
overall
picture
we
can,
we
can
build
virtual
services
and
both
real
networks
using
service
chaining.
You
also
had
management
frameworks,
obviously
tools
that
we
can
use
to
manage
our
real
and
virtual
networks,
and
then
we.
H
H
Is
this
better
okay
and
then
drawing
on
some
architectural
observations
from
from
from
this
work
and
sort
of
what's
what's
going
on
at
the
ITF
and
the
world
in
general?
So
obviously
the
increasing
role
of
software
is
is
really
important.
So
not
all
things
are
are
going
to
be
specified
by
protocols
and
and
not
not
everything
that
the
network
does
is
specified
by
an
action
of
the
routing
protocol
or
some
other
component
of
an
effort
to
be
at
the
idea
have
been
used
to
doing.
H
H
He
put
route
computation
or
some
some
virtual
network
building
functionality
or
Sdn
control
in
a
central
place,
and
that
just
makes
a
lot
of
engineering
sense.
I
just
wanted
to
point
out
a
couple
of
things
that
one
has
to
keep
in
mind
there.
So
obviously
one
one
question:
is
you
know
what
happens
if
you
have
some
kind
of
network
partition
situation,
your
your
Center
is
down.
What
would
you
do
ways
to
deal
with
that
situation
are?
Is
the
useful
there's
also
sort
of
I
mean?
K
H
K
K
H
Or
content
provider
or
social
media
network
to
tell
me
how
to
do
things
and
sort
of
a
nightmare
scenario
there
that
this
ability
to
centralize
things
and
control
things
and
and
build
virtual
networks
might
affect
that
someday.
Perhaps,
but
maybe
that's
just
unnecessary
fear,
I,
don't
know,
I
just
want
to
bring
it
up
anyway.
H
H
H
So,
let's
keep
that
in
mind.
The
other
thing
to
think
about
I.
Think
I
haven't
really
personally
done
at
all
any
work
in
data
models
and
so
on
at
the
ITF
or
elsewhere,
but
it
it
does
seem
that
there's
some
some
role
here
to
think
about
layering
between
different
models
in
this,
it's
obviously
at
least
data
or
control
models
that
you
can
use
to
control
the
behavior
of
your
individual
devices.
You
have,
you
know,
probably
something
at
the
level
of
networks
we're
also
seeing
this
service
models.
H
There
are
more
abstract
models,
but
I
want
this
kind
of
VPN
service
between
these
endpoints
and
then
you
have
to
map
between
these
different
layers
and
and
thinking
about
that.
How
they
have
the
different
pcs
in
the
idea
fit
together
is
really
important,
particularly
when
we're
thinking
about
new
new
work,
and
the
next
point
is
about
general
designs
and
specific
designs.
So
here
I
was
thinking
of
the
requirements
from
PT,
PP
and
5g
on
slicing
and.
H
You
know
it's
in
some
sense:
it's
it's
a
little
vague
at
the
moment
because
they
don't
have
their
final
requirements
set.
You
know,
of
course,
the
way
that
the
system
of
specification
work
opera.
It's
there,
it's
it's,
you
know
once
they
decide
their
final
requirements,
then
five
minutes
after
that.
Every
day,
all
the
standards
and
products
need
to
be
ready.
So
it's
kind
of
a
late
to
wait
for
that,
but
at
the
same
time
the
ITF
like
we,
we
have
our
own
work
product
and
our
own
technologies.
H
It's
our
job
to
keep
our
virtual
networking
technologies.
You
know
competitive
and
have
the
teachers
that
we
think
our
users
are
likely
going
to
need
and
I
would
actually
argue
that
it's
will
make
sense
for
the
idea
to
think
about
this
from
you
know,
take
all
the
you
know
possible
inputs,
yes
from
others
like
by
G
word,
but
but
do
our
own
analysis
of
what?
What
do
we
need
to
build
for
our
own
tools
in
terms
of
data
models
and
other
pieces
of
work
instead
of
like
narrowly
building
for
a
particular
requirement,
I
think
we.
F
H
Get
some
hints
at
least
of
where,
where
the
what
direction
requirements
are
likely
to
be-
and
this
also
should
be,
provide
more
launch.
It
want
a
bit
before
for
our
designs.
Then
then
a
particular
release
in
some
other
SDO
as
an
example,
so
coming
more
towards
the
end
possible
new
work.
What
should
we
be
working
on
in
the
ITF
in
this
space?
You
know
a
lot
of
this.
H
We
are
already
working
on
obviously
but
22,
so
the
paint
picture
from
my
simplistic
perspective
I
was
the
we're
gonna,
be
working
on
data
models
and
that
that
seems
like
a
useful,
useful
thing
to
do,
for
the
virtualization
service
and
and
and
all
the
other
layers
and
again
figuring
out
correctly
what
you
know
how
these
things
be
to
fit
to
each
other.
What
the
overall
picture
is.
H
We
also
I
think
we,
you
know
it
seems
like
natural
ITF
role
to
manage
heterogeneous
networking.
So
when
you
have,
you
know
different
kinds
of
technologies
underneath
or
your
your
announcement
tools
are
different
in
different
parts
of
the
network
to
be
able
to
provide.
You
know
some
overall
building
of
these
networks
for
those
kinds
of
situations,
I
think
that's
a
very
natural
fit
for
the
idea.
H
Other
day
that
he
would
like
to
see,
but
he
feels
that
there
are
some
cases,
particularly
in
radio.
He
was
thinking
about
the
radio
aspect,
in
particular
that
you
can,
if
you're,
trying
to
think
about
things
that
are,
you
know,
hard
guarantees
that
you
know
will
give
you
one
gig
for
ever
on
this
link
and
that's
that
that's
a
thing
if
we
can
try
and
provide
over
wireless.
H
H
Cross-Domain
doing
you
know
all
these
things
over
multiple
identity
domains
is
obviously
very
interesting
and
it's
part
of
the
requirements
set
I
think
we
were
doing
some
work
around
those
things,
maybe
at
the
ITF
I.
Remember
correctly.
The
point
that
I
want
to
make
here,
though,
is
that
that's
really
hard.
H
So
you
know
proceed
carefully
and
then
the
final
point,
this
isn't
about
the
technical
things,
data,
but
also
protocols
or
any
of
that.
But
again,
if
we
would
actually
agree
about
some
technical
things
in
the
industry.
That
would
be
a
useful
thing
for
all
of
us
to
discuss
with
our
customers
and
or
each
other
and
and
and
so
on
or
in
standards
discussions.
H
L
F
F
L
Slices
right
here
to
the
packet,
where
you
think
it
belongs
because
there's
something
it
might
be,
not
I
green
there.
Maybe
we
don't
need
the
new
protocols,
but
we
need
to
see
how
it
works.
So,
basically
how
we
can
do
our
SLA
assurance
and
they
will
monitor
that
SLA
being
delivered
and
service
practice,
because
the
services
are
so
critical
that
we
cannot
operate
Network
and
wait
for
the
client
to
say.
Oh
I
was
my
service.
H
L
H
M
H
I
just
wanted
to
finish
with
with
next
steps,
so
I
kind
of
view.
This
exercise-
and
this
is
like
from
our
parts-
it's
a
really
tiny
exercise.
We
haven't
done
much
late
submitting
the
graph
as
well
this
morning,
but
I
think
will
be
useful
to
connect
sort
of
a
top-down
approach
that
some
people
are
taking
in
to
like,
slicing
and
and
then
the
bottom-up
approach
figure
out
where
they
meet.
H
And
then
you
have
a
bigger
or
cooler
picture,
and
then
you
know
we
determine
what
concrete
thinks
we
need
to
do
over
existing
technology,
and
maybe
it's
very
little.
Maybe
it's
more
I
don't
know,
and
then
my
dream
is
actually
maybe
between
this
idea
and
the
next
I
see
if
you
actually
figure
this
out
on
lists
and
together
and
then
we
know
that
we
need
to
do
this
data
model
and
this
enhancement
in
that
other
protocol.
H
And
then
we
push
it
to
the
existing
working
groups
and
then
that
that's
the
work
that
we
need
to
do
so
I
think
success
is
actually
reasonably
easy.
If
you
take
that
approach
instead
of
like
this
big
bang,
totally
new
concept,
very
complicated
approach
and
the
data
model
development,
obviously
with
virtual
networks,
seems
like
it
should
be
a
continue
to
the
IETF
and-
and
this
isn't
just
for
5g.
H
This
is
for
other
other
purposes
as
well,
and
and
it
isn't
only
between,
like
ITF
and
IDs
consumers
and
other
organizations,
it's
also
within
IETF,
like
how
do
you
our
different
pieces,
fit
together?
How
does
the
service
model
fit
with
traffic
engineering,
or
you
know
how
all
these
things
relate
to
say,
deterministic
network,
so
many
questions
are
now
you
go
to
discussion.
M
The
end
of
Dunwich,
so
when
you
mention
the
SLA,
that's
a
look:
there's
a
slippery
slope
with
that.
The
you
have
technical
isolates
and
you
have
business
SLA
and
the
question
is
which
ones
we
want
to
define,
and
we
had
the
discussion
already
before
in
depth
and
we
didn't
really
know
the
correlations
between
technical
and
the
business
ones.
And
then
we
said:
oh,
we
will
just
leave
it
inside
the
service
model,
but
you
will
not
go
into
deep.
You
know
KPIs
and
then
try
to
map
it
into
that.
M
So
this
is
one
thing
where
we
will
need
much
more
input
from
the
operators
and
then
maybe
no
reach
out
to
nano
ripe
peach.
You
know
and
get
input
on
on
that
from
there
because
they
know
that
and
there's
something
is
want
to
work
with.
Just
that's
where
we
would
have
to
create
some
new.
You
know
liaisons
in
order
to
be
able
to
change
that
information
and.
H
That's
a
very
good
point:
I
I
just
wanted
to
say
that
when
I
mentioned,
SLA
I
was
mostly
thinking
SLA.
In
the
context
of
you
know
some
industry
alignment
across
some
terminology
and
concepts
which
would
be
referenced
by
by
SLS,
but
not
not
that
we
necessarily
have
to
develop
SLA
standards
here
at
idea.
If
that's
that
may
be
an
easier
task
and
then
developing
them
yeah.
H
M
N
Oh
you're,
playin
UCL
I
have
a
couple
of
confusions
that
been
brought
up
by
your
slide,
set,
which
I
wonder
if
you
could
clear
up
for
me.
You
mentioned
virtualization
quite
a
lot
to
mention
VPN
quite
a
lot,
and
you
mentioned
slicing
quite
a
lot.
It
wasn't
clear
to
me
if,
in
your
mind,
these
are
fully
interchangeable
or
you
see
them
as
separate
concepts.
Somehow
that's
my
first
point
and
my
second
point
confusion.
The
second
one
is
that
you
start
off
by
saying
there
are
some
frustrations
people
not
building
on.
N
What's
there
and
you're
near
the
end,
you
say:
well,
this
cross-domain
staff
is
very
difficult.
Beware,
but
for
many
people
slicing
is
the
cross-domain
issue
as
opposed
to
the
single
domain
elements
that
already
exist.
So
I
wondered
again,
you
can
crank
their
up
any
confusion.
I've
got
from
that.
Yes,
some
noise.
H
N
H
O
A
Layer,
1
layer
2
in
there
too,
for
us
20
year,
but
they
haven't
been
doing,
is
inter
domain
connectivity
across
layer
right
so
this
yet
to
be
solved
back
to
your
question:
if
service
provided
networks
or
network
or
service
service
model
or
API
for
the
service
model
might
be
the
right
level
to
interest.
If
it
was
upper,
maybe
interrupting
the
radio
system
service
level
API,
it
could
be
ditto.
So
it's
we
need
to
figure
out
right
level
of
abstraction.
That's
going
to
be
exposed
to
a
persistent,
that's,
not
bad.
Okay,.
A
If
you
look
at
today's
work
with
idea,
we
see
service
model
layer,
interacting
with
network
layer,
interaction.
You
device
layer,
so
well
established
terminology
done
in
a
CTU.
Is
it
defined
in
the
city
on
architecture
down
TST
come
some
other
working
group
for
this
particular
use
case.
It
feels
like
write
all
of
drafting
for
complete
slice.
It
might
not
be
yet
to
figure
out
how
to
properly
and
do
the
mapping
between-
and
this
goes
kind
of
tomorrow,
abstract
things
such
as
graph
theory.
A
O
H
B
Only
because
it's
sort
of
tied
up
with
other.
Q
So
my
name
is
Jeremiah
I'm.
The
switch
EVP
liaison
person
to
right
here
and
I
mean
we
had
a
lot
of
discussion
on
slicing
at
the
last
idea,
also
in
regards
or
in
relation
to
switch.
Evp
I
think
this
discussing
this.
In
this
way,
it's
a
very
good
approach.
It's
it's
very
open
and
it
gives
us
possibility
to
bring
things
in
in
3gpp.
I
can
only
say
you
know
we
have
the
tight
schedule.
We
see
that
the
slicing
issue
effects,
as
we
have
really
an
end-to-end
view
to
slicing.
F
Q
Is
logically
one,
this
affects
the
radio.
This
affects
the
core
network
and
this
also
affects
the
operation
and
maintenance.
These
are
three
areas
that
are
capable
of
talking
with
each
other
in
our
standards
body,
but
it's
not
always
the
easiest.
So
we
we
need
to
sort
out.
The
internal
communication
needs
already
how
much
we
can
participate
in
that
work
soon.
This
is
a
question
I'm
not
able
to
answer
and
I
think
you
were
some
of
the
last
meetings
you
might
have
seen
the
rush.
Q
R
Thank
you.
Don't
from
Holly
first
I
like
to
say
that
a
good
pointer
is
about
the
Taylor
network
and
the
general
purpose
Network
cause
to
me
recently,
there's
some
emerging
surveys,
which
would
require
guaranteed
or
more
predictable
performance,
and
they
will
like
to
be
carried
in
a
common
infrastructure
network
and
with
together
with
the
existing
services
we
have
so
into.
P
My
name
is
George
Anderson
I'm
working
for
four-way
yari.
Thanks
for
the
presentation,
I
have
a
question
regarding
I
assumed
that
the
alignment
terminology
and
concepts
is
not
all
I
assume
that
that
is
also
the
case.
So
if
that
is
the
case,
then
looking
to
3gpp,
you
know
think
it
is
very
much
focusing
on.
P
Maybe
the
terminal
should
be
different,
but
there
are
some
concepts
that
are
very
similar,
and
my
question
is
related
to
the
first
of
all.
The
data
models,
you
might
note,
did
you
with
PSA
5
is
working
on
templates
social
I,
see
templates
that
are
somehow
you
know
similar
to
probably
what
we
like
to
do
with
the
data
models.
So
how
could
you
ensure
that
you
know
that
we
are
not
competing
with
with
the
slicing
templates
that
did
you
page
a5.
P
A
Be
sorry
there's
nothing
because
they
get
a
time.
The
point
being
again,
we
need
to
pick
same
language
in
order
to
make
work
and
partially
idea
behind.
This
drop
is
to
provide
enough
thinking
to
people
for
not
familiar
with
ADF
technologists,
to
understand
what
can
be
done
and
really
listen
technology
and
technology
rather
than
trying
to
even
do
it.
H
A
To
head,
we
expect
a
lot
of
interest,
your
communication,
the
lifecycle
management,
but
to
think
we've
been
literally
betted.
It's
happening,
someplace
else
and
very
important
stuff.
Someone
have
to
manage
the
lifecycle
resource
allocation,
de-allocation
of
the
crowd
operation
across
net
refugio
faculty.
S
S
S
In
some
sense
for
a
purpose
of
managing
clothes,
obviously
not
more
for
other
and
that's
an
additional
element.
However,
the
reserve,
probably
around
220
methods,
to
build
virtual
networks
based
on
surveys
done
in
academic
world.
That's
basically
what
happened
so
far
before
it's
time
to
move
on.
You
said
bottom-up
link
with
a
top-down,
not
one
or
the
other,
and
that's
where
I
would
like
to
be
to
highlight
a
little
different
from
virtualization,
where
slightest
are
substantially
different,
conceptually
impractical,
which
is
basically
led
to
the
fact
that
slices
are
driven
by
services.
S
B
S
S
H
H
H
L
B
A
A
B
T
A
T
B
T
T
Okay,
this
work
is,
is
the
new
work,
so
the
pasta
is
a
problem
so
for
every
protocol
learn
several
vendors
have
their
own,
provide
young
modules
to
implement
the
F
function.
I
just
spoke
to
nifer
Anwar
we
set
fried
young
module
can
well
applied
unto
their
own
devices.
However,
you
choose
the
different
function,
which
bodies
provide
via
module
is
hard
to
use
across
a
DD
but
across
devices,
and
cannot
meet
the
demand
of
common
requirements
abuser.
T
So
in
this
case
he
imagine
a
scenario
is
that
a
service
provider
managing
the
thousands
of
devices
and
the
disturb
others
are
from
many
different
vendors.
It
must
be
a
disaster
new
if
the
l
Yahoo
is
no
way
to
control
this
two
bodies
in
the
stand
way.
So
the
solution
can
be
extract,
some
common
properties
within
devices
implementing
the
eruption.
So
this
method
in
a
two
and
two
meanings.
But
why
is
a
provide?
A
unique
identifier
for
a
pecan
configuration
function,
for
example
the
AG?
T
It's,
and
sometimes
you
in
the
provided
module
we
decode
the
time
out
or
the
expand
time.
So
you
need
to
find
a
wheel.
Somebody
summarized
in
a
unique
or
identify,
and
the
similar
is-
the
interval
is
sometimes
called
they
detect
in
about
all
the
air
interval,
so
we
can
give
a
unique
name,
so
the
user
cannot
be
confused
by
these
similar
names
and
provided
this
master
can
provide
the
common
properties,
for
example,
the
static
RP
country
under
the
tip,
the
dynamic
IP
interest.
T
Here
you
in
in
our
traps
that
we
summarized
the
perfect
cheers
of
Arabic
of
configuration
functions
the
static
IP
entry,
if
you
perform
this
task,
is
to
define
the
static
matching
between
a
few
addressed
and
MSG
MS
address
of
the
host
is
that
to
not
spot
dynamic.
A
addressed
resolution
put
calm
so
a,
however,
because
most
a
hosta
suppose
a
dynamic
address
resolution
so
defining
the
static,
IP
cache,
which
interest
is
usually
not
required,
and
the
second
is
the
dynamic
learning
of
happy
answer.
T
Here's
the
host
Alice,
which
itself
has
a
function
and
the
default
state
of
the
device
art
lab
she
learning
dynamic.
We
do
not
need
the
you
do
not
need
to
use
the
command
to
start
in
this
venture.
However,
according
to
the
network,
needs
users
can
adjust
the
stand
parameters
of
dynamic
therapy,
the
crocks,
the
a
P
it
is
enabled
by
default.
So
from
this
task
a
future,
it
is
about
the
proxy
option
on
interface
and
then
finally,
is
a
graduate.
T
So
we
can
see
the
tree
diagram.
The
bottle
is
a
is
a
container.
Have
you
static
table
it
initial
or
busted'?
To
get
IP
configuration
expiry
time
is
a
lift,
so
is
the
lift?
Hritik
describes
the
edging
time
of
the
dynamic
at
the
entry.
Also,
the
appeal
name
is
the
leaf.
It
describes
the
way
the
dynamic
IP
learning
is
disabled
in.
F
T
To
the
enabled,
by
default,
and
also
the
IP
proxy
duties,
rather,
it
is
enabled
by
default,
and
we
have
the
corruption
of
every
prop.
So
this
Covina
he
described
as
a
common
configuration
of
every
prop,
so
the
live
gives
crop
interval
means
the
interval
pod.
Detecting
the
dynamically
appear
interest
under
the
prop
time
is
the
number
of
aging
problem.
Attempts
for
autonomic
dynamic
at
the
entry
and
the
unicast
unit
has
two
means
whether
use
unique
aasta.martin
averaging
a
problem
message.
T
Query
happy
entries,
including
static,
dynamic
and
interface
tester,
a
few
inches
it
has
the
BFG
we
BR
I's
named.
Vietnam
is
the
name
of
a
VPN
instance
to
which
are
happy.
Entry
belong
under
the
container.
Happy
statics
is
the
list
of
active
package,
a
package
static.
It
has
two
subparts
the
club
or
under
the
local.
So
this
means
you
can
display
useful
at
the
app
is
testicle
globally,
all
all
for
specified
specify
the
in
the
test.
T
T
The
first
is
about
the
publishing
model
I
HAF
IP.
He
also
called
it
I've
c7
2007,
with
overlaps
data
being
provided
in
this
module.
So
for
this
question
we
can.
We
can
somebody's
surprised
if
you
need
to
to
to
pass
the
pasta.
Pies
is
a
for
in
the
best,
invest
a
bit
for
a
labor
for
scanner,
which
it
as
it
does
describe
the
entries
of
matching
or
mapping
from
ipv4
address
to
link
layer
addresses.
It
is
similar
to.
B
B
F
T
B
B
B
So
so
this
is
Chris
Powers
speaking,
one
of
the
working
group
chairs
and
I
wanted
to
just
kind
of
jump
in
on
I.
Think
this
this
model
is
actually
a
pretty
good
fit
for
Archie
gwg.
B
That
is
we're
tasked
with
doing
yang
models
for,
for
you
know,
standards
and
and
work
for
which
there
are
no
that's
in
the
routing
area,
but
there
are
no
other
working
groups.
You
know
specifically
tasked
with
that
now,
technically,
you
know,
one
might
think
of
ARP
is
not
being
in
the
routing
area,
but
in
the
in
Taria,
but
it
the
aspects
of
this
is
really
more
service.
You
know
proxy
ARP
and
other
features
that
are
going
on
on
routers.
B
You
know
large
scale
devices,
not
the
art
functionality,
necessarily
on
a
host,
that's
sort
of
fixed
and
but
but
you're
really
talking
about
I.
Think
in
this
draft
more
the
art
functionality,
the
sort
of
complex
art
functionality
to
on
on
routers,
so
I
think
it's
a
good
fit
to.
You
know
eventually
consider
for
adoption
in
the
routing
area
working
group
I.
You
know
I'm
not
in
a
position
to
judge
consensus
on
that
or
anything
yet,
but
I
think
it's
a
good.
It
you've
brought
it
to
the
right
place
and
presenting
it
here.
B
T
Unopposed,
a
in
the
best
state
in
the
best
ipv4
labor
force
can
enter
the
Lister
describes
a
few
cash,
so
that
is,
we
can
obtain
the
happy
entries
for
in
the
base.
However,
in
our
fgfe
tables,
we
cannot
only
obtain
the
every
entrance
for
interviews.
We
also
can
obtain
the
other
entries
for
a
VPN
yeah.
T
The
admin
can
be
used
to
children
practice,
the
module
of
the
pasta
in
the
past
and
in
the
festival
in
the
past
in
the
FC
72
swing,
and
this
last
request
I
left.
The
comment
is
the
about
the
I
can
network
instances
for
implementing
they
be
asked,
as
just
discussed
in
the
first
part
we
have
director
and
tree.
We
can
do
the
unity,
a
dictating,
every
app
stack,
stack
tables
and
container
so
best
of
the.
B
In
here,
so
so
we're
running
out
of
time
for
this
this
presentation,
but
it
would
be
good
to
take
the
you
know
the
comments
and
your
potential
responses
to
them.
And
you
know
at
this
point
it's
an
individual
draft,
so
go
ahead
and
make
the
changes
that
you
think
are
appropriate
to
respond
to
these.
That
draft.
F
F
F
F
F
U
This
draft
is
basically
version
zero,
but
it
has
little
bit
of
history
involved,
so
we
presented
so
we
presented
this
drop
in
ITF
93
with
there's
a
point-to-point,
VFD
solution
and
at
the
same
time
another
use
case
was
presented
to
multi-point
PFD.
So
in
that
meeting
we
decided
we'll
merge,
both
the
rocks
and
in
the
next
idea,
ITF
95.
U
We
presented
the
version
four,
including
certain
comments,
and
after
that
there
was
a
little
bit
of
you
got
an
ITR
issue
from
a
direction
that
is,
we're
working
committee
decided
that
we
should
stop
working
on
the
drop
and
we
abandon
it
and
now
since
that
idea
is
open-
and
we
have
moved
out
of
that
point
to
multi-point
PFD
use
case
from
this
job
and
you
have
got
the
new
graph
in
picture
so
I'll
just
in
order
to
make
sure
that
everyone
is
in
the
sync
with
the
doc.
So
this
rock
was
meant
for.
U
If
you
see
nowadays
in
the
data
centers
another
world
geography
detection-
usually
it
goes
with
our
sub
second
timers.
So
if
we
implement
vrrp
in
the
control
plane
those
sub
second
timers,
they
they
become
bit
aggressive
and
they
can
cause
CPU
delays
and
all
those
things.
So
if
it
implies,
if
it
uses
the
BFD,
which
is
meant
to
serve
that
purpose,
then
we
can
move
forward.
But
in
order
to
use
the
BFD,
the
error
T
has
some
intrinsic
of
limitations
where
vrrp
itself
them
only
the
master
vrrt.
U
It
sends
out
the
packet,
whereas
the
backups
they
don't
know
about
the
backups,
as
a
result
of
which
master
also
they
don't
send
out
a
packet.
So
master
also
don't
know
about
the
backups.
There
is
for
the
VAP
to
him
to
work
on.
Bfd
needs
the
IP
address
of
the
source
and
destination.
That
is
where
we
brought
this
truck
to
bring
a
new
Edward
Edward
Packer
type.
Along
with
it.
He
brought
a
new
thing
called
peer-to-peer
peer
learning
model
there
there
once
we
will
receive
the
packets
from
the
backup.
U
We
will
decide
what
can
be
the
critical
backup
and
based
on
that
will
form
a
session
which
will
be
known
as
the
critical
DMV
session.
So
the
table
will
look
like
all
the
PR
notes.
They
will
interact
with
each
other
and
they
will
form
a
table
based
on
the
highest
priority.
Backup
and
I
hear
the
master.
They
will
form
a
critical
point,
the
FD
session,
if
the
critical
VFD
session
goes
on
the
tentacle.
U
So
in
this
trap,
what
we
have
done,
we
have
brought
all
the
comments
which
Jeff
has
given
Max
and
Chris
and
venkata.
They
have
given
us
comments
on
this.
So
with
this
job,
we
are
bringing
an
extension
to
the
VRP
protocol
where
we
are
adding
a
new
packet
type,
then
we
are
defining
a
peer
learning
model
and
defining
a
critical
BFD
session
or
the
faster
convergence.
U
Also
with
this,
since
we
are
bringing
a
new
packet
and
Edward
packet
type,
as
a
result
of
which
we
are
updating
the
vrrt
protocol
state
machine
to
accommodate
two
new
timers,
then
we
have
considered
the
lot
of
things
like
security,
like
the
scalability
considerations,
that
backups
should
not
send
that
word.
At
the
same
time,
they
should
set
at
a
lower
frequency
or
the
operational
considerations
like
the
error
free,
should
it
should
interface
only
with
vfe,
when
VFD
is
capable
to
send
aggressive
type
aggressive
packets.
U
U
B
So
Chris
parish
here
so
since
we
had
reached
the
point
of
working
group,
adoption
and
I
think
consensus
was
to
adopt
adopt
the
other
draft,
except
for
the
IPR
issues.
I
think
you
know,
I
haven't
read
this
draft
yet
so
I
need
to
I'll
finish
this.
So
I
need
to
look
at
it,
but
I
think
we'll
do
a
quick
evaluation
and
then
you
know
pull
for
the
consensus
for
working
group
adoption
and
we
can
have
that
that
discussion
in
that
poll
or
Greg
may
have
some
other
no
I.
L
I'm
very
glad
and
I
appreciate
your
consideration.
I
know
that
either
issue
is
very
sensitive
idea
and
then
share
that
concern.
I
just
want
to
reiterate
what
was
said
in
presentation
in
the
first
session
when
we
presented
updated
document
that
opera's
believe
again
to
the
best
of
our
knowledge
that
their
stated
disclosed
IPR
is
not
covered
new,
updated
version
entirely.
I
try
not
to
be
very
legal
wine.
All.
B
Right,
so
so,
so
this
is
a
new,
a
new
draft
right
part
of
the
working
group,
adoption
poll
and
I
guess
in
this
case
we'll
separate
the
adoption
poll
in
the
IPR
poll.
Okay,
there
is,
this
is
a
new
draft
and
if
there
is
any
IPR
that
applies
to
this
new
draft,
the
point-to-point
draft,
then
we
will
expect.
We
expect
it
to
already
be
disclosed
if
and
then
we'll
require
all
the
authors
to
to
attest
to
that
or
you
know
to.
L
Yes,
well,
the
thing
is
that
what
happened?
I
don't
know,
I
I
think
it
was
quite
clear,
stated
in
presentation.
This
presentation
that
drafts
now
split
again,
so
we
have
point-to-point
separate
draft.
Both
point-to-multipoint
leave,
the
drab
and
IPR
was
disclosed
a
war
against
old
version
of
proposed
mechanism
using
point-to-multipoint
BFD
in
geography.
Yes,.
B
U
B
L
B
We
have
effectively
withdrawn
the
disclosure
by
creating
a
new
draft.
The
draft
was
related
to
a
previous
draft.
This
is
a
new
draft.
It
will
have
no
IPR
disclosed
against
it.
Presumably,
if
that
remains
the
case,
then
we've
communicated
to
the
working
group,
the
you
know
we
follow.
It
I
believe
the
correct
intention
here,
Lou
Berger,
our
IPR
consultant-
will
give
us
some
comments
on
that
as
well.
V
B
B
W
Is
Dolly's
like
with
Huawei
whoever
submitted
the
disclosure
can
submit
an
updates
and
disclosure
saying
it
doesn't
apply
anymore,
so
you
there's
no
way
to
cause
there
not
to
be
any
disclosure
at
all
associated
with
it,
with
the
disclosure
associated
with
it.
Everyone
that
says
that
the
previous
disclosures
are
no
longer
applicable
and
there's
no
IPR.
B
X
That's
right,
thank
you.
It's
there,
so
VPNs
have
been
widely
deployed
to
support
multi
tendency
in
public
operating
that
much
higher
me.
That's
what
their,
what
I
think
they
well
know
for
they're
now
expected
to
provide
emerging
new
services
and
your
customers
with
more
stringent
performance
requirements,
for
example,
bandwidth,
latency
and
jitter
were
on
this
shared
infrastructure.
So
what
we
think
we
need
to
do
is
do
provide
an
enhanced
VPN.
F
X
X
We
then
divide
up
the
resources
and
make
separate
resource
units
available
on
the
rooters
and
on
the
links,
and
then
we
allocate
in
the
overlay
those
resources
to
a
particular
customer
VPN.
So
at
the
top
we
got
a
VPN
plus
instance.
Then
we've
got
the
underlay
and
then
at
the
bottom
we
got
the
physics
of
we
think
we're
going
to
have
to
do
that
isolation.
X
So
traditional
overlay
mechanisms,
each
overlay
network,
now
its
own
route
or
a
forwarding
table
and
separate
address
space
overlays
compete
for
resources
with
each
other
unless
every
connection
in
the
overlay
is
mapped
to
a
dedicated
traffic.
Engineered
LSP
for
bandwidth
reservations.
The
problems
with
this,
as
we
increase
the
number
of
overlay
tenants
asking
for
guarantee
performance.
This
results
in
a
increase
in
the
number
of
traffic
engineered
LS
period.
Ultimately,
we
believe
leads
to
scalability
problems
and
that's
the
signal.
Routing
argument.
X
X
X
X
So
all
the
problems
well,
MTR
is
really
still
a
best-effort
forwarding
service
and
adding
it
with
segment.
Routing
doesn't
entirely
solve
the
problem,
because
segment
routing
doesn't
provide
resource
reservation,
which
is
necessary
for
the
performance
guarantee.
Also,
if
we
build
our
control
plane
based
on
MTR,
we
think
we
hit
a
topology
ID
limit
problem
right,
I.
L
X
Well,
the
way
we
see
things
I
think
is
that
there
are
two
candidate
solutions.
Flexing
is
one
packet
solution,
but
that
doesn't
give
you
the
statistical
multiplexing
advantage
that
yes,
so
you
burn
through
your
fan
with
by
hard
allocating
it
services,
but
you
get
the
best
get
the
best
Apple
the
physics
you
can
get
the
packet
yeah.
Y
X
L
X
X
So
we
think
we
need
to
extend
segment
routing
to
write
the
right
resource
reservation
in
a
way
it
doesn't
have
at
the
moment.
So
we
need
perhaps
that
our
path
reservation
follow.
The
paradigm
of
s
are
following
the
paradigm:
s
are
achieved.
Resource
reservation
with
much
less
state,
now
view
controller
based
reservations.
X
We
need
to
use
a
much
finer
grain
system
than
we
have
before.
So
let's
look
at
the
next
slide.
So
this
is
the
proposed
mechanism.
I
think
I've
shown
this
before
from
now
shown
this
before.
So
this
is
showing
you
can
see
in
the
underlying
physics.
Then
we
carve
this
up
into
the
virtual
networks
using
it.
X
It's
it's
and
link
seeds
per
per
slice
or
for
a
VPN
instance
enhance
you've
units
and
then,
if
we
can
build
VPNs,
as
you
can
see
on
the
right
using
that
technology-
and
we
can
do
this
for
both
strict
paths
and
we
can
do
it
when
all
the
resources
are
absolutely
dedicated
to
the
packet.
We
get
the
loose
sort
of
loose
part,
but
this
path
has
got
a
number
of
interesting
issues.
X
Then
it
currently
runs
over
the
data
plane.
That's
or
the
substrate
is
unmodified
from
the
existing
substrates
would
be
with
the
deck
net.
People
are
going
to
put
I
talking
about
changing
the
queuing
model.
There
were
some
sort
of
discussion
on
that
today,
but
it's
essentially
running
over
an
unmodified
substrate.
A
About
terminology,
so,
if
you
look
at
death
net,
they
specifically
provide
point-to-point
service.
Yes,
you
call
this
virtual
network,
wouldn't
call
it
plus
of
service
or
something
really
relevant
to
the
way
network
behave
or
avoid
better
description.
What
it
does,
because
you
cannot
do
multi-point
connectivity
right
if
you
second
Trouty.
A
I
X
X
We
need
to
integrate
the
service
function,
trains
because
service
function
chains
are
an
integral
part
of
the
types
of
network
that
this
is
going
into.
So
we're
looking
under
need
to
look
for
a
holistic
sort
of
approach
to
both
VPN
plus
and
network
slicing,
whether
we
do
this
with
segment
routing,
which
is
one
of
the
candidates
or
whether
actually
it
forces
us
to
revisit
that
decision
is
the
next
thing
we
need
to
study.
X
X
Now,
here's
the
master
and
really
quite
interesting
one.
In
order
to
get
hard
slicing,
you
need
to
do
disruption,
free
reconfiguration
of
your
network
and
it's
to
say
it's
quite
easy
to
add
a
service
expose
you
take
another
service
away,
but
modifying
a
service
is
going
to
intrinsically
affect
the
path
that
goes
on
and
the
net
result
is
going
to
be
a
potential
disruption
to
quite
sensitive
traffic.
The
SR
is
quite
good
at
doing
that.
At
least
strict
SR
is
quite
good
at
doing
that.
X
The
loose
path,
though,
are
still
subject
to
looping,
unless
you
have
some
conversions
controlled
technology
such
as
one
of
the
IPR
IDF
RoR
techniques,
but
really
that
one
of
the
hardest
problems
is
to
figure
out
how
we
do
disruption.
Free
defragmentation
of
the
network
now
is
how
we
recover
the
resources
without
affecting
any
of
the
other
services,
because
you
can't
move
anything
around
without
changing
the
delay
and
jitter
characters.
Y
X
X
Y
You're
defeating
the
purpose
of
segment
trusting
it
means
you
are
third
service.
You
are
putting
state
in
the
network.
I
am
kinda
missing,
so
let's
say:
I
have
a
hundred
services
yep
in
order
for
them
not
to
compete,
I
need
to
create
a
hundred
slice
yep
on
the
top
link.
If
I
have
a
thousand
I
need
to
create
a
thousand
slice,
so
I
end
up
doing
RSVP
right
now,.
Y
Y
Y
Look
like
a
number
syphilis
I'm
talking
about
a
simple
scenario:
I
have
a
thousand
like
I.
Have
a
thousand
services
yep.
We
are
competing.
I
need
to
create
thousand
colors
on
this
link,
you're
saying
no,
you
don't,
but
then
you
say
is
say:
go
ahead
and
use
the
label,
stacking
techniques
and
SRT's.
Why
do
I
get
colors?
I
can
use
the
label
stuck
in
without
having
colors
I.
Think
my.
F
R
C
C
R
For
Maui,
basically,
this
the
this
kind
of
mechanism
is
for
the
emerging
new
services
which
require
them
much
high-end
amount
of
the
performance
guarantee.
So
for
this
kind
of
services
you
need
them
to
reserve
the
resource
and
each
hub
and
link
a
father,
normal
service.
You
still
can
abrogate
state
into
a
one-seed.
R
X
You
need
tools
you
can
pop
in
to
an
ounce
of
aggregation
depending
on
the
relative
load
of
the
system
force,
but
by
using
a
seedless
you
can
you
you
RSVP
would
require
a
single
label
that
went
end-to-end
right,
a
that.
You
swap
attorneys
no,
but
it
will
actually
one
for
service.
This
allows
you
to
have
fewer,
because
if
you
can
share
a
source,
then
you
don't
need
to
have
it
have
a
dedicated
label
for
it.
X
X
What
we're
in
the
process
of
working
out,
because
we
don't
want
them
to
invent
anything
we
don't
need
to.
So
this
is
sort
of
a
difference,
different
slant
on
the
stuff
that
your
is
doing
and
it's
to
say,
look
at
the
technologies.
We've
got
see
what
we
need
to
add
to
it
and
see
what
we
need
to
build
some
some
working
models
and
then
figure
out
what
missing
bits
are
so.
V
If
you're
doing
like
traffic
engineering
architecture
work,
you
don't
work
a
group
for
that.
If
you
are
thinking
about
how
to
leave
some
state
behind
on
with
traffic
engineer
des
our
turns
out,
there's
a
document
on
that
inside
being
adopted
inside
MPLS,
so
you
can
go
there
and,
if
you're
doing
something
that
specifically
to
go
there,
although
I,
don't
think
you
have
any
hope
you
don't
go
there.
I
think
this
is
good
stuff.
It's
just
not
clear
to
me.
It
should
be
discussing
this
well.
M
K
I
X
X
Flow
filtering,
as
was
discussed
in
depth
in
the
deterministic
Network
earlier
on,
but
without
that
we
we
do
need
to
have
some
separation
of
the
traffic.
All
they
were
all
the
packets
will
bump
into
each
other,
and
this
isn't
really
about
a
much
harder
degree
of
isolation
than
we're
normally
used
to.
L
Gregg
misty
city
I
think
that
what
your
intention-
and
that
will
be
very
helpful
if
we
said
common
dictionary,
that
already
being
used
in
other
areas
that
work
on
network
slicing,
for
example,
and
that's
not
not
very
specific,
but
it's
with
as
the
end.
There
is
such
thing
as
the
lifecycle
orchestration.
L
So
what
we
talk
about
when
Commission
on
the
Commission
entity
on
networks
wise,
it's
a
part
of
their
life
cycle,
so
you
don't
draw
it
and
hope
that
it
works.
So
we
have
a
certain
process
that
you
need
to
follow.
You
can
characterize
this
process
for
the
data
model
and
use
it
so
in
there
is
part
of
this
process.
Is
a
service
activation
testing
in
terms
of
what
we
talk
and
continue
to
talk,
talk
and
dead
net.
L
Yes,
in
some
scenarios,
especially
where
there
is
a
very
stringent
requirements
to
latency
in
deuteron,
Beckett
was
care
on
a
bandwidth
on
an
edge
may
not
be
sufficient.
That's
what
answer
to
tell
and
again
I
think
that
there
might
be
that
will
not
find
their
satisfactory
technology
in
the
packet
with
a
statistical
multiplexing
that
will
address
this
requirements
of
ultra
reliable.
In
the
world
latency
communication,
because
other
scenarios
for
Network
slicing,
they
don't
have
the
strict
requirements.
I
I
would
like
us
to
understand
that
network
slicing
not
necessarily
set
forth
there.
L
X
L
X
A
X
A
F
J
J
Because
I
think
that
is
the
new
thing
potentially,
which
is
coming
out
of
this
world
right
and
it's
more
include
de
and
collapse.
Large
teams-
I,
don't
have
anything
about
that.
But
what
you're
actually
going
to
introduce
here
is
that
if
you
have
a
virtual
link,
which
has
specifically,
is
also
like
a
definite
specific
interface,
you
have
to
semantically
segment
routing
today
to
allocate
a
city
and
the
DC
circuit
or
whatever,
based
on
that
virtual
interface
right.
F
J
A
Agree,
I'll
take
a
turn:
I
think
what
you're
trying
to
achieve
is
to
show
how
a
label
could
be
used
method
data
to
derive
particular
behavior.
This
is
the
purpose
of
the
drug.
It
doesn't
really
introduce
new
functionality
as
of
now
see
so
yeah
I
think
as
a
work
in
a
group,
you
should
be
looking
at
it
as
a
solution
to
solve
particular
problem,
except
now,
because
there
are
implications
to
dark
the
control
plane
in
segment.
Routing
of
the
path
computation
is
happening.
A
X
X
Y
Y
Y
R
Z
N
Z
So
so
one
of
the
requirements,
what
I'm
missing
here
is
to
take
into
consideration
some
of
the
physical
resource.
Like
a
certain
note
or
the
link
in
case,
you
want
to
provide
a
service
which
you
know
needs
to
send
traffic
over
pots,
which
cannot
be
separated
from
a
layer.
Three
thing:
for
example,
like
you
wanna
send
traffic
on
one
end,
a
service
over
to
different
parts
going
over
to
set
up
completely
different
components
within
the
network.
X
Z
B
AA
From
the
agenda
months,
maybe
I
can
start
my
incision
pages
from
the
agenda.
I'm
lucky
to
be
the
last
one,
to
give
the
presentation
and
I
think
they
were
meaningful,
because
it
is
the
ITF
100,
and
this
is
the
last
rotation
of
the
larger
work
group
as
well.
I
think
my
topic
is
minifides.
Will
my
topic
is
about
the
information
model
of
the
control
plan
and
use
of
plans,
depression,
dmg
divide?
My
name
is
Beryl
from
China
Mobile
and
my
co-authors
are
Victor
Michael
and
shipping
fee.
AA
Before
going
to
the
details
of
the
information
model,
we
have
several
pack
questions
to
be
answered.
The
first
one
is
that
why
we
need
the
steel
suppression
EMG
device
and
why
we
need
to
the
information
model,
the
contact
of
the
BMG
divided
with
control
plant
and
the
use
of
length
depression
come
from
the
operators.
We
think
that
the
PNG
device
want
to
and
can
be
utilized
such
as
some
function
can
be
uploading
to
centralized
processor,
which
can
be
deployed
in
the
reconstruction
based
in
heart.
Considering
of
the
PNG
device.
AA
There
are
many
functional
modules
in
our
device,
such
as
the
information
management
module
and
the
loud
heat
and
forwarding
modules,
and
when
we're
talking
about
the
PNG
device
with
the
control
plan
and
a
user
plant
oppression,
we
mean
that's.
The
Benjie's
and
control
plan
focused
on
the
information
management,
such
as
the
users,
information
and
resource
information
which
can
be
distributed
and
centralized
and
virtualized
we're
always
talking
about
the
BMG's
for
audience
or
user
plan.
AA
We
think
that
it
includes
the
forwarding
and
the
routing
modules
they
keep
in
the
tradition
of
BMG
divided
and
due
to
the
actual
deployment
in
our
network.
We
have
found
out
that
much
for
you
much
more
than
those
device
on
if
now,
if
this
is
in
the
network,
so
we
have
the
distributed
user
plan
belonging
to
multiple
vendors
and
we
have
the
unified
control
and
last
we
need
the
scan
type.
The
management,
and
that
is
the
reason
why
we
proposed
our
information
model
of
the
PNP
device,
with
the
control
and
and
user
plantarflexion.
AA
The
information
our
model
is
going
to
present
the
attributes
or
information
transmitted
transmitted
between
the
control
plant
and
the
user
plan,
and
if
the
send
of
the
question
is
right
now
we
give
out
the
information
model
and
I
think
it
is
the
right
time
to
come
up
with
the
standard
information
model.
The
first
thing
important,
the
reason
is
facts
about
the
operators:
implementation,
wait.
China
Mobile
have
found
the
trial
from
last
year
in
further
provinces,
with
the
PNG
device
of
the
control
plan
and
user
pan
suppression
for
providing
the
home
province
service.
AA
Besides,
we
have
also
deployed
the
DMZ
device
with
the
fewest
operation
in
our
no
not
no,
not
means
the
whole
network
based
on
the
action
and
unless
a
technology
that
is
the
most
important.
The
reason
why
we
needed
a
standard
information
model
and
it
decides
in
other
organizations
that
has
the
broadband
forum.
AA
Neither
are
we
needed
a
control
plan
to
be
dynamic
enough
to
handle
all
these
users
of
death.
We
think
it
is
the
right
time
to
come
up
with
the
on
few
suppression
information
model.
That
is
the
reason
why,
by
now-
and
this
is
not
the
first
time
to
presents
the
few
separation
youngji
device
information
model-
we,
the
information
model-
is
firstly
introduced
in
ITF
98
in
chakra.
We
present
the
information
model
in
in
the
face
2000
system
workgroup
and
the
later
way
modified
li
information
model.
AA
Due
to
some
comments,
and
in
this
ITF
after
the
discussion,
wisdom
in
the
face
to
routing
service
workgroup
chairs
and
the
80s,
they
think
that
women,
not
it's
sufficiently
review
in
the
recent
work
group,
so
they
just
asked
to
handle
this
walk
up
to
the
mountain
work
group.
That
is
the
reason
why
we're
here
and
decide
this
month
on
this
week.
Among
this
this
week,
are
the
few
suppression
protocol
requirements
are
proposed
and
also
in
the
large
team
work
group.
AA
C
Idea,
because
I
tourists
here
the
reason
why
we
encourage
them
to
do
an
information
model
is
they're
actually
looking
for
the
dynamic
data
store
portion
where
they
can
be
there
and
then
go
away
and
not
use
the
configuration
line.
So
that's
why
they
were
first
a
tire
to
us,
as
they
were
one
of
the
first
dynamic
data
store
candidates.
C
We
were
looking
at,
so
that's
why
them
MDMA
and
NBA
reference
was
made
because
they
were
actually
working
on
it
when
they
do
that,
we
suggested
an
informational
model
first,
because
we're
still
working
through
how
old
that
works.
If
the
topology
model
who
have
grib
models,
this
was
a
new
case
notice
that
in
the
RIP
model
we
went
for
an
info
model
as
well,
so
that
that's,
why
you've
got
that
background
and
that's
why
we've
sent
it
here.
C
M
F
M
AA
M
I
AA
AA
The
end
of
the
information
model
contains
the
rules
and
the
entries
generated
by
the
control
plan.
We
think
the
information
model
is
a
foundation
and
based
on
a
standard
information
model,
then
we
can
generate
the
corresponding
data
model
or
govt
and
about
them
about
this
new
SP,
the
protocol,
it's
aimed
at
defining
and
find
it
aims
at
finding
a
suitable
protocol
to
carry
the
attribute
or
the
information
and
described
in
our
information
model,
as
we
think
that
the
recent
universe,
such
as
Netcom
or
is
not
sufficiently
enough
to
carry
these
attributes.
AA
AA
If
you
are
interested
in
in
this
work,
you
can
keep
an
eye
on
the
mail
list,
because
we
will
show
the
demo
of
the
PMG
device
with
control
and
induce
a
plan
suppression
in
the
next
IDF
action,
and
if
you
are
interested
in
that
you're
welcome
to
join
us
and
last,
please
read
the
document
and
give
us
some
feedback
or
comments.
See
I
really
enjoyed
see
the
email.
AA
The
next
slide
is
about
the
overview
of
the
model.
This
is
just
the
information
model
and
ignacio's
depression.
Emg
device
on
the
information
model
is
divided
into
two
parts.
One
is
the
control
name
from
each
modal.
Justify
the
controlling
devices
and
another
is
the
user
plan
information
model
justify
the
use
of
plantomycin
the
control
plane.
Information
mobile
focuses
on
the
information
model
related
away,
such
as
the
user
information,
the
service
interface
information.
AA
And
the
used
our
plan
information
model,
it
contains
the
some
available
natural
resources,
such
as
the
prod
resource
information
and
the
traffic
statistics.
If
you
want
to
go
to
the
details,
you
can
refer
to
now,
adjust
yes
next
slide,
maybe
the
stance-
and
we
think
we
think
this
model
enjoys
the
journey
idea,
because
we
first
meet
stuff
on
interface
to
routing
system
and
then,
when
we
start
out
a
large
work
group
will
meet
new
friends,
so
here
I
wish
everyone
to
enjoy.
The
journey
is
like
the
same
as
our
chopped
thank.